Welcome to AI Academy
Your comprehensive journey from AI novice to proficient practitioner
29
Modules
7
Tool Deep Dives
50+
Activities
∞
ROI Potential
What You'll Achieve
- Understand how AI works - transformers, models, and capabilities
- Master prompting techniques and advanced frameworks like PDCA
- Use multimodal AI: text, images, speech, and video
- Become proficient with 6 essential AI tools
- Automate data analysis with AI-generated scripts
- Apply 12 manufacturing-specific AI prompts to real operations data
- Build and pitch your own AI pilot project
Course Structure
- Getting Started (1-2): Foundations, model landscape, capabilities
- Core Skills (3-6): Prompting, PDCA framework, multimodal AI, productivity hacks
- Tools Deep Dive (7-13): Perplexity, Claude, Cowork, Excel, Shortcut, NotebookLM, Nano Banana
- Power User (14-17): Templates, guides, projects, iteration & fine-tuning
- Advanced (18-22): Data automation, Perplexity dashboards, Claude artifacts, workflows, safety & ethics
- Use Cases (23-26): Writing, data, research, meetings
- Capstone (27): Build your own AI pilot
- Manufacturing (28): 12 copy-paste AI prompts for operations
🚀 The BlackArc AI Mindset
1
Augment
AI enhances, not replaces
2
Iterate
Conversation beats one-shot
3
Verify
Always check AI outputs
4
Scale
Once it works, systematize
Module 1: AI Foundations
Understanding how AI works and what it can do for knowledge workers
Learning Outcomes
- Understand the transformer architecture at a conceptual level
- Know how LLMs process text and generate responses
- Recognize capabilities and limitations of current AI
- Identify AI-suitable tasks in your workflow
The Transformer Revolution
Modern AI is built on transformer architecture - a breakthrough from 2017 that enabled machines to understand context and relationships in text.
Key Concepts
- Tokens: AI breaks text into chunks (~4 chars = 1 token). "Understanding" → ["Under", "stand", "ing"]
- Embeddings: Each token becomes a vector capturing meaning
- Attention: The model weighs which words matter most - this is the magic
- Generation: Output is created one token at a time
Why This Matters
- Context dependency: AI needs context to perform well
- Specificity helps: Clear prompts create clearer attention signals
- Hallucinations: AI generates plausible but sometimes false tokens
- Cost: More tokens = more processing = higher cost
What AI Excels At
- Summarization: Condensing long documents into key points
- Drafting: Creating first versions of emails, reports, documents
- Analysis: Finding patterns, extracting data, identifying issues
- Brainstorming: Generating ideas, alternatives, edge cases
- Translation: Between languages, formats, or technical levels
- Code: Writing, explaining, debugging, converting code
- Research synthesis: Combining information from multiple sources
AI Limitations
- No real-time knowledge: Training data has a cutoff date
- Hallucinations: Can generate confident-sounding but false information
- No true reasoning: Pattern matching, not genuine logic
- Context limits: Can only "see" a fixed window of text
- No persistent memory: Each conversation starts fresh
- Biases: Reflects biases in training data
Module 2: Model Landscape
Understanding different AI models and when to use each
Current Model Landscape (2026)
Different models have different strengths. Choosing the right model for your task can dramatically improve results and reduce costs.
| Model | Best For | Context | Speed | Cost |
|---|---|---|---|---|
| Claude Opus 4 | Complex reasoning, research, nuanced writing | 1M | Slower | $$$ |
| Claude Sonnet 4 | Balanced - everyday tasks, coding, analysis | 1M | Fast | $$ |
| Claude Haiku 4.5 | Quick answers, simple tasks, high volume | 200K | Fastest | $ |
| GPT-4.1 | Latest OpenAI model, general tasks, multimodal | 128K | Fast | $$ |
| OpenAI o3 | Deep reasoning, math, complex logic | 200K | Slower | $$$ |
| Gemini 2.5 Pro | Very long contexts, Google integration | 1M+ | Medium | $$ |
| DeepSeek R1 | Reasoning, math, cost-sensitive work | 128K | Medium | $ |
| Llama 4 | Self-hosting, open-weight deployments | 128K | Varies | Free* |
Model Selection Guide
Use Claude Opus when:
- Writing executive-level content that must be perfect
- Complex multi-step reasoning or analysis
- Extended thinking for multi-step problems
- 1M token context for massive document analysis
- Nuanced tasks requiring deep understanding
- Legal, compliance, or high-stakes documents
Use Claude Sonnet when:
- Everyday writing, drafting, and editing
- Code generation and debugging
- Best balance of speed, intelligence, and cost
- Artifacts for interactive apps and visualizations
- Standard analysis and summarization
- Most general knowledge work
Use Claude Haiku when:
- Quick lookups and simple questions
- High-volume processing tasks
- Initial drafts you'll heavily edit
- Cost is a primary concern
Use GPT-4.1 / o3 when:
- Image analysis and generation tasks
- Voice/audio processing
- Deep reasoning chains requiring step-by-step verification
- Tasks needing structured reasoning traces
- Existing OpenAI integrations
Use Gemini when:
- Extremely long documents (>150K tokens)
- Google Workspace integration
- Native Google Workspace and Android integration
- Video understanding
Activity: Model Comparison
Try the same prompt across 2-3 models: "Analyze this contract clause and highlight top 3 risks." Compare clarity, detail, hallucinations, and speed.
Module 3: Prompting Mastery
The art and science of getting AI to do what you actually want
🎯 The CRAFT Framework
C
Context
Background & constraints
R
Role
Who should AI be?
A
Action
Specific task
F
Format
Output structure
T
Tone
Style and voice
Prompting Techniques
1. Zero-Shot Prompting
Direct instruction without examples. Good for simple, well-defined tasks.
Summarize this contract in 3 bullet points focusing on payment terms.
2. Few-Shot Prompting
Provide examples of what you want. Essential for formatting or style matching.
Convert notes to action items:
Notes: "Need to call John about the project deadline"
Action: [ ] Call John re: project deadline
Notes: "Remember to submit the Q3 report by Friday"
Action: [ ] Submit Q3 report (due: Friday)
Notes: "Follow up with Sarah on budget approval"
Action:
Notes: "Need to call John about the project deadline"
Action: [ ] Call John re: project deadline
Notes: "Remember to submit the Q3 report by Friday"
Action: [ ] Submit Q3 report (due: Friday)
Notes: "Follow up with Sarah on budget approval"
Action:
3. Chain-of-Thought (CoT)
Ask the AI to show its reasoning. Critical for complex analysis.
Analyze this contract clause for risks.
Think step by step:
1. First, identify the key obligations
2. Then, note any ambiguous language
3. Finally, list potential risks and mitigations
Think step by step:
1. First, identify the key obligations
2. Then, note any ambiguous language
3. Finally, list potential risks and mitigations
4. Role Prompting
Give the AI a specific persona. Changes vocabulary, focus, and approach.
You are a senior procurement specialist with 15 years of experience in manufacturing supply chain management.
Review this SOW and identify the top 3 compliance risks.
Review this SOW and identify the top 3 compliance risks.
Power Tips
- Be specific about length: "in 3 sentences" vs "briefly"
- Specify format: "as a markdown table" or "as bullet points"
- Set constraints: "without jargon" or "for non-technical audience"
- Ask for alternatives: "give me 3 different approaches"
- Request reasoning: "explain your reasoning"
- Iterate: "make it more concise" or "expand point 2"
Module 4: Plan-Do-Check-Act
The advanced technique that 10x's your AI output quality
Why This Matters
Most people send one prompt and accept whatever comes back. Professionals break work into phases. This single technique will separate you from 90% of AI users.
🔄 The PDCA Cycle for AI Work
P
Plan
AI outlines approach first
D
Do
Execute in sections
C
Check
AI verifies its own work
A
Act
Iterate on findings
PDCA in Practice
Phase 1: PLAN
I need to write a proposal for [X]. Before you write it:
1. What sections should it include?
2. What information do you need from me?
3. What's your recommended approach?
Don't write the proposal yet - just give me the plan.
1. What sections should it include?
2. What information do you need from me?
3. What's your recommended approach?
Don't write the proposal yet - just give me the plan.
Phase 2: DO
Good plan. Let's start with Section 1: Executive Summary.
Here's the context: [provide details]
Write just this section, keeping it under 200 words.
Here's the context: [provide details]
Write just this section, keeping it under 200 words.
Phase 3: CHECK
Now review what you just wrote. Check for:
- Accuracy against requirements
- Claims that need sources
- Clarity for non-technical readers
- Anything that could be misinterpreted
List any issues you find.
- Accuracy against requirements
- Claims that need sources
- Clarity for non-technical readers
- Anything that could be misinterpreted
List any issues you find.
Phase 4: ACT
Good catches. Please revise to address issues #1 and #3.
Keep the rest as-is.
Keep the rest as-is.
When to Use PDCA
- Complex documents: Proposals, reports, analyses >1 page
- High-stakes outputs: Anything for clients or leadership
- Multi-part tasks: Research requiring synthesis
- Technical accuracy: Legal, financial, compliance
Skip PDCA for: Quick answers, simple drafts, brainstorming
Module 5: Multimodal AI
Beyond text: Working with images, speech, and video
What is Multimodal AI?
Modern AI can process multiple types of input and output: text, images, audio, and video. This unlocks entirely new workflows for knowledge workers.
Image Understanding
AI can now "see" and understand images. Upload screenshots, diagrams, photos, or documents.
What AI Can Do With Images
- Extract text: OCR from screenshots, photos of documents
- Analyze charts: Interpret graphs, dashboards, visualizations
- Describe content: Explain what's in a photo or diagram
- Debug UI: Review screenshots for UX issues
- Compare: Spot differences between two images
Example Prompts
# Screenshot analysis
"Extract all the data from this table screenshot into CSV format"
# Dashboard interpretation
"What are the key insights from this dashboard? What should I be concerned about?"
# Document processing
"This is a photo of a receipt. Extract: date, vendor, items, total"
"Extract all the data from this table screenshot into CSV format"
# Dashboard interpretation
"What are the key insights from this dashboard? What should I be concerned about?"
# Document processing
"This is a photo of a receipt. Extract: date, vendor, items, total"
Image Generation
- Claude (via Artifacts): Claude now generates images directly within Artifacts for quick visuals and diagrams
- DALL-E 3: High quality, good at following complex prompts
- Midjourney: Best aesthetic quality, artistic styles
- FLUX: Fast, high-quality open-source image generation
- Stable Diffusion: Open source, highly customizable
- Adobe Firefly: Commercial-safe, integrates with Creative Cloud
Best for: Presentations, concepts, marketing visuals, prototyping
Speech-to-Text (Transcription)
Convert spoken audio into text for analysis, search, and action items.
Tools
- Whisper (OpenAI): Best accuracy, free/open source option
- Otter.ai: Real-time transcription + AI summaries
- Fireflies.ai: Meeting bot with CRM integration
- Granola: AI notepad for meetings
- macOS/iOS Dictation: Built-in, surprisingly good
Use Cases
- Meeting transcription and action item extraction
- Dictating first drafts (3x faster than typing)
- Transcribing interviews for analysis
- Voice memos → structured notes
Text-to-Speech
- ElevenLabs: Most realistic voices, voice cloning
- OpenAI TTS: Good quality, simple API
- NotebookLM: Generates podcast-style discussions
- Play.ht: Many voices, good for content creation
Use cases: Audio versions of reports, training content, accessibility
Video Understanding
AI can now analyze video content, extracting insights without manual review.
Capabilities
- Gemini 2.5 Pro: Native video understanding, process full videos up to 1 hour with strongest multimodal video capabilities
- Claude Opus/Sonnet: Analyze images, charts, diagrams, photos with detailed reasoning
- Twelve Labs: Search within videos, find specific moments
Use Cases
- Summarize recorded meetings you couldn't attend
- Extract key moments from training videos
- Analyze competitor product demos
- Review security footage for specific events
Video Generation
- Sora (OpenAI): Highest quality, limited access
- Runway: Good for short clips, motion brush
- Pika: Easy to use, good for social content
- HeyGen: AI avatars for presentations
Current best use: Short clips, social media, concept visualization
Document Intelligence
Process documents at scale - PDFs, scanned images, forms, and more.
Capabilities
- Claude: Native PDF understanding, maintains formatting
- Azure Document Intelligence: Enterprise-grade extraction
- Textract (AWS): Forms, tables, handwriting
- DocuSign AI: Contract analysis and extraction
Workflow Example
1. Upload batch of vendor contracts (PDFs)
2. AI extracts: party names, terms, renewal dates, obligations
3. Output: Structured spreadsheet for analysis
4. Alert: Flag contracts expiring in 90 days
2. AI extracts: party names, terms, renewal dates, obligations
3. Output: Structured spreadsheet for analysis
4. Alert: Flag contracts expiring in 90 days
Activity: Multimodal Exploration
Try each modality: (1) Upload a screenshot to Claude and ask for analysis, (2) Use dictation for your next email draft, (3) Upload a PDF and extract key data.
Module 6: Productivity Hacks
Shortcuts and techniques to 10x your daily AI usage
High-Impact Hacks
1. The "Give Me Options" Hack
Give me 3 different approaches to [task], with pros/cons of each.
2. The "Pretend I'm Wrong" Hack
Now pretend you're a skeptic. What are the 3 biggest weaknesses or counterarguments?
3. The "Rubber Duck" Hack
I'm trying to solve [problem]. Don't answer yet. Ask me 5 clarifying questions first.
4. The "Template Generator" Hack
Create a reusable prompt template for [task type]. Include placeholders I'll fill in each time.
5. The "Explain Like I'm..." Hack
Explain [concept] three ways:
a) For a complete beginner
b) For someone who knows basics
c) For an expert who needs key differences
a) For a complete beginner
b) For someone who knows basics
c) For an expert who needs key differences
Build Your Prompt Library
- Meeting prep: "Based on this agenda, what questions should I prepare?"
- Email response: "Draft a response that [goal]. Under 100 words."
- Document review: "Review for [criteria]. List issues as: Critical, Important, Minor."
- Brainstorm: "Give me 10 ideas: 3 conventional, 4 creative, 3 unconventional."
- Summarize: "Summarize in [format] for [audience]. Include: key points, decisions, next steps."
Module 7: Perplexity AI Research
Real-time research, live sources, and up-to-date intelligence at your fingertips
Why Perplexity Changes Everything
Unlike Claude or ChatGPT (GPT-4.1/o3), Perplexity searches the live internet in real-time and cites its sources. It's like having a research analyst who can instantly find, synthesize, and cite information from across the web - perfect for sales teams, marketing, leadership briefings, and competitive intelligence.
What Makes Perplexity Different
Perplexity is an AI-powered research engine that combines conversational AI abilities with real-time web search:
- Live Internet Access: Searches the current web, not training data from months ago
- Source Citations: Every claim is linked to its source - verify anything instantly
- Follow-up Questions: Conversational refinement of your research
- Focus Modes: Academic, Writing, Wolfram (math), YouTube, Reddit
- Pro Search: Multi-step research that digs deeper automatically
- Collections: Save and organize research for projects
Perplexity vs. Other AI Tools
| Capability | Perplexity | Claude/ChatGPT |
|---|---|---|
| Real-time web data | ✅ Always current | ❌ Training cutoff |
| Source citations | ✅ Every claim cited | ⚠️ Limited/none |
| News & current events | ✅ Live | ❌ Outdated |
| Document analysis | ⚠️ Limited | ✅ Excellent |
| Long-form writing | ⚠️ Basic | ✅ Excellent |
| Code generation | ⚠️ Basic | ✅ Excellent |
Best practice: Use Perplexity for research and current information, then bring findings to Claude for analysis, writing, and document creation.
When to Use Perplexity
✅ Perfect For
Current news, company research, market trends, competitive intel, fact-checking, RFP research, customer briefings
❌ Not For
Long document writing, code generation, complex analysis of your own documents, creative writing
Sales & Business Development Research
Perplexity transforms pre-meeting prep and account research from hours to minutes:
Account Research Prompt
"I'm preparing for a meeting with [Company Name]. Research:
1. Recent news and press releases (last 90 days)
2. Current leadership team and recent changes
3. Their stated strategic priorities and challenges
4. Recent contract awards or major deals
5. Competitive landscape - who are they working with?
6. Any public pain points or complaints
7. Upcoming events or announcements
Cite all sources so I can verify."
Key Account / Supplier Research
"Research [Company/Supplier Name] for a strategic planning meeting:
1. Current leadership team (CEO, COO, VP Supply Chain) and backgrounds
2. Recent major orders, partnerships, or expansion announcements
3. Active RFQs or sourcing events in [our space]
4. Their stated operational priorities and strategic plan
5. Recent quality issues, recalls, or supply disruptions
6. Revenue trends and capital investment plans
7. Upcoming contract renewals we should know about
Focus on sources: company website, IndustryWeek, Supply Chain Dive, Thomas, D&B"
Quick Pre-Call Intel
"I have a call in 30 minutes with [Person Name] who is [Title] at [Company].
Give me a quick briefing:
1. Their LinkedIn background (recent roles, education)
2. Any recent quotes or interviews they've given
3. Topics they care about based on public statements
4. Their company's recent news
5. Potential conversation starters or connection points"
Opportunity Research
RFQ/Opportunity Research
"Research this sourcing opportunity: [RFQ Number or Project Name]
Find:
1. The original RFQ/RFP and any revisions
2. Current supplier and contract history
3. Industry discussion about this procurement category
4. Similar orders or projects awarded recently
5. Key evaluation criteria (price, quality, lead time, certifications)
6. Supplier qualification requirements
7. Buyer's historical purchasing patterns and preferences"
Current Supplier Research
"Who is the current supplier for [Company]'s [Material/Service]?
- Contract details (annual volume, pricing structure)
- How long have they been the supplier?
- Any quality or delivery issues reported?
- Is the contract up for rebid?
- What's the relationship like — strategic partner or transactional?"
Marketing & Content Research
Trend Research for Content
"What are the top trends in [industry/topic] right now?
Research:
1. What industry analysts (Gartner, Forrester, IDC) are saying
2. What's being discussed at recent conferences
3. Hot topics on relevant Reddit communities and forums
4. Recent thought leadership articles from key voices
5. Emerging technologies getting attention
6. Problems and pain points being discussed
Cite sources for each trend so I can dig deeper."
Competitive Content Analysis
"Analyze [Competitor]'s content marketing strategy:
1. What topics are they publishing about recently?
2. What's their messaging and positioning?
3. Which pieces seem to get the most engagement?
4. What keywords are they targeting?
5. What gaps exist that we could fill?
6. How has their messaging evolved over the past year?
Check their blog, LinkedIn, and any media coverage."
Customer Voice Research
"What are [target audience] saying about [topic/problem]?
Search:
1. Reddit discussions in relevant subreddits
2. LinkedIn posts and comments
3. Industry forum discussions
4. Review sites and complaint boards
5. Conference Q&A and panel discussions
6. Twitter/X conversations
What language do they use? What frustrates them? What do they want?"
Executive Briefings & Leadership Research
Perplexity excels at creating up-to-date briefings for leadership:
Market Update Brief
"Create an executive briefing on the [industry] market:
Cover (with sources):
1. Market size and growth projections (latest analyst estimates)
2. Key players and recent market share shifts
3. Major deals, acquisitions, or partnerships this quarter
4. Regulatory changes impacting the market
5. Technology trends reshaping the industry
6. Risks and uncertainties to monitor
Format: Executive summary (3 bullets), then sections with details.
Keep it to 2 pages equivalent."
Competitor Update Brief
"Create a competitor intelligence brief for leadership on [Competitor Name]:
Last 90 days:
1. Major announcements or news
2. Contract wins (especially large or strategic)
3. Product/service launches
4. Leadership changes
5. Financial performance (if public)
6. Strategic moves or positioning changes
7. Hiring patterns that signal direction
Assessment: What do these moves mean? What should we watch?"
Industry Event Prep
"I'm attending [Conference Name] next week. Brief me on:
1. Key themes and hot topics on the agenda
2. Notable speakers and what they're likely to discuss
3. Companies announcing or exhibiting
4. Pre-conference news and buzz
5. Key people I should try to meet
6. Conversation starters based on current trends
Make this actionable for networking and intelligence gathering."
Perplexity Power Techniques
1. Use Pro Search for Deep Research
Pro Search performs multi-step research automatically:
- Breaks your question into sub-questions
- Searches multiple times
- Synthesizes findings
- Costs more queries but worth it for complex research
Enable: Toggle "Pro" before searching
Best for: Complex topics, multi-faceted questions, comprehensive briefings
2. Focus Modes
Academic: Searches scholarly sources, papers, research
Writing: Helps with content creation and editing
Wolfram: Math, data, computational questions
YouTube: Searches video content, transcripts
Reddit: Community discussions, opinions, experiences
Use Academic for: Research reports, white papers, evidence-based content
Use Reddit for: Customer voice, real opinions, pain points
3. Source Specification
Be specific about sources for better results:
"Search only [website.com] for..."
"Focus on sources from 2024 and 2025"
"Prioritize: Gartner, Forrester, and IDC"
"Check IndustryWeek and Supply Chain Dive specifically"
"Look at LinkedIn posts from [industry] executives"
4. The Perplexity → Claude Workflow
STEP 1: Research in Perplexity
"Research [topic] with sources"
STEP 2: Copy findings to Claude
"Based on this research: [paste Perplexity output]
Now:
- Analyze implications for our business
- Draft a memo summarizing key points
- Create an action plan based on findings
- Write talking points for leadership"
This combines Perplexity's live data with Claude's analysis and writing.
Collections & Organization
- Create Collections: Save related searches together (by client, project, topic)
- Share with team: Collections can be shared for collaborative research
- Build on previous research: Perplexity remembers context within a thread
- Export findings: Copy formatted output to docs, Notion, etc.
Pro tip: Create a Collection for each major account or ongoing research need. Add to it over time to build institutional knowledge.
Activity: Real Research Challenge
Pick a real task: (1) Research a prospect before your next meeting, (2) Create a competitive brief on your top competitor, or (3) Research a trend relevant to your work. Use Pro Search, cite sources, then bring findings to Claude to create a deliverable.
Module 8: Claude AI Deep Dive
Mastering Anthropic's Claude for professional knowledge work
Why Claude for Business
Claude is built for professional work. It excels at following complex instructions, working with long documents, generating structured output, and maintaining consistent quality across tasks. It's the tool that gets better the more specific you are.
Claude Model Tiers (2026)
- Claude Opus 4: Most capable model. 1M token context. Extended thinking for complex reasoning chains. Use for: executive analysis, legal review, deep research, multi-step problem solving.
- Claude Sonnet 4: Best all-around model. 1M token context. Fast, intelligent, great at coding and analysis. Default choice for 80% of tasks.
- Claude Haiku 4.5: Fastest and cheapest. 200K context. Use for quick lookups, simple drafts, high-volume processing.
When to Upgrade to Opus
Start with Sonnet. Upgrade to Opus when: the output isn't nuanced enough, you need extended thinking for complex reasoning, you're working with sensitive or high-stakes content, or the task requires synthesizing 100+ pages of context.
Claude's Core Capabilities
- Extended Thinking: Claude can "think" through complex problems step by step before answering. Enable this for multi-step analysis, math, logic, and strategic planning. You can see its reasoning chain.
- 1M Token Context: Process ~750K words at once. Upload entire codebases, full contract sets, years of meeting notes, or complete technical manuals.
- Artifacts: Claude creates standalone interactive applications, data visualizations, documents, and tools — live in the conversation. More on this in the Artifacts tab.
- Projects: Persistent workspaces where you preload instructions and knowledge files. Every conversation in the project starts with that context. More in the Projects tab.
- File Understanding: Upload PDFs, images, spreadsheets, code files, CSVs. Claude reads and reasons about them — not just extracts text.
- Styles: Save writing styles (formal, concise, technical) and apply them consistently across conversations.
- Tool Use: Claude can call external tools and APIs, enabling integration with your existing systems.
Integration Points
- Claude.ai: Web and mobile app for direct use
- Claude for Enterprise: SSO, admin controls, no training on your data
- Claude in Excel: AI directly in your spreadsheets (Module 10)
- Claude Code: AI-powered software engineering in your terminal
- API: Build Claude into your own applications and workflows
- Cowork: Claude with a full computer environment (Module 9)
Artifacts: Claude's Killer Feature
Artifacts are standalone, interactive outputs that Claude creates alongside its response. They're not just text — they're live applications, visualizations, and tools.
What Artifacts Can Create
- Interactive dashboards: Live charts and graphs from your data
- Web applications: Calculators, forms, tools — fully functional
- Data visualizations: Charts, treemaps, heat maps with real data
- Documents: Formatted reports, proposals, SOPs
- Diagrams: Flowcharts, org charts, process maps (Mermaid/SVG)
- Games and simulations: Interactive models and prototypes
Manufacturing Examples
"Analyze this production data CSV and create an interactive dashboard artifact showing:
1. OEE trend by line (line chart)
2. Downtime Pareto (bar chart)
3. Quality yield by shift (heat map)
4. Summary KPI cards at the top
Make it filterable by date range and production line."
Business Examples
"Create an interactive ROI calculator artifact for our AI implementation proposal:
- Input fields: number of employees, hours saved per week, average hourly cost
- Calculate: annual savings, 3-year NPV, payback period
- Show a break-even chart
- Include a sensitivity table for different adoption rates"
Projects: Your Persistent AI Workspace
Projects transform Claude from a blank slate into a specialized expert for your specific work.
How Projects Work
- Create a project (e.g., "Q1 Production Analysis")
- Add Custom Instructions — these apply to every conversation in the project
- Upload Knowledge Files — PDFs, docs, spreadsheets, code
- Every new conversation starts with this full context loaded
Project Ideas for Manufacturing
- SOP Assistant: Upload all SOPs → ask questions about procedures
- Quality Analyst: Upload quality manual + specs → analyze NCRs against standards
- Vendor Manager: Upload contracts + performance data → negotiate with context
- Safety Advisor: Upload OSHA regs + incident history → risk assessment
- Training Developer: Upload tribal knowledge docs → create training materials
Custom Instructions Template
## Role
You are a senior operations analyst specializing in manufacturing and supply chain optimization.
## Context
- We are a [type] manufacturer with [X] production lines
- Key products: [list]
- ERP system: [name]
- Key challenges: [list top 3]
## Instructions
- Always reference uploaded data files when making recommendations
- Use specific numbers, not vague claims
- Format outputs as executive summaries with bullet points
- Flag assumptions explicitly
- When analyzing data, always show your methodology
## Output Preferences
- Executive summary first, details after
- Tables for comparisons
- Charts/artifacts when data warrants visualization
- Action items with owners and deadlines
Power User Techniques
1. Use Extended Thinking for Complex Analysis
"Think deeply about this before answering. I need you to:
1. Analyze the attached P&L for cost reduction opportunities
2. Consider second-order effects of each reduction
3. Rank by feasibility, not just magnitude
4. Identify which reductions compound over time"
2. Chain Conversations with Context
Don't start over. Build on previous responses: "Based on the analysis you just did, now create a 90-day action plan." Claude remembers the full conversation.
3. Use XML Tags for Complex Prompts
<context>We are evaluating two CNC machines for purchase.</context>
<data>[paste specs]</data>
<analysis_required>
- TCO comparison over 5 years
- Capacity impact on current bottleneck
- Risk assessment for each option
</analysis_required>
<output_format>Executive brief, max 2 pages, with recommendation.</output_format>
4. Ask Claude to Challenge Your Thinking
"Here's my plan to reduce inventory by 20%. Play devil's advocate:
- What am I missing?
- What could go wrong?
- What assumptions am I making that might be wrong?
- What would a skeptical CFO ask?"
5. Iterate, Don't Restart
Instead of reprompting from scratch: "Make it more concise" / "Add quantitative evidence" / "Rewrite for a non-technical audience" / "Now create a one-page executive summary of everything above."
Module 9: Claude Cowork
Claude with a full sandboxed computer — creates real, downloadable files
What is Claude Cowork?
Cowork gives Claude access to a full sandboxed Linux computer environment. It can install packages, run complete Python scripts, create Office documents (Excel, PowerPoint, Word), browse the web, and produce downloadable files — all while you watch and guide it in real time.
Cowork vs. Artifacts: Know the Difference
| Capability | Artifacts | Cowork |
|---|---|---|
| Output type | Interactive web components (HTML/JS/SVG) | Downloadable files (XLSX, PPTX, DOCX, PDF, CSV) |
| Best for | Dashboards, calculators, visualizations | Reports, presentations, data processing |
| Environment | Browser sandbox | Full Linux computer with packages |
| Can install libraries | No | Yes (pip install, apt-get, etc.) |
| Web browsing | No | Yes |
Rule of thumb: If you need a file you can email, print, or open in Office — use Cowork. If you need something interactive to explore in your browser — use Artifacts.
Manufacturing Use Cases
ERP Data Cleanup & Reporting
"Upload this messy ERP export and create a clean, formatted Excel report with:
- Pivot tables summarizing production by line and shift
- Charts showing output trends over the last 12 months
- Conditional formatting to highlight underperforming lines
- A summary sheet with KPI cards
Save as .xlsx so I can share with leadership."
PowerPoint from Data
"Create a 10-slide executive presentation from this quarterly data:
- Title slide with our company name and quarter
- KPI summary slide with key metrics
- Production output trends (chart)
- Quality metrics by product line (chart)
- Top 5 downtime causes (Pareto chart)
- Cost variance analysis
- Improvement initiatives status
- Next quarter targets
- Risks and mitigations
- Q&A slide
Use a clean, professional template. Save as .pptx."
Data Processing & Cleaning
"I have 6 CSV files from different production lines. Combine them into one clean dataset:
- Standardize column names across all files
- Remove duplicate entries
- Fill in missing values where possible (flag what was filled)
- Add a 'Source_Line' column
- Create a summary report of data quality issues found
- Save the clean combined file as Excel with a data quality tab."
Web Research with Compiled Report
"Research the top 5 predictive maintenance software vendors for discrete manufacturing.
For each, find: pricing model, key features, integration capabilities, customer reviews.
Compile findings into a formatted Word document with:
- Executive summary with recommendation
- Comparison matrix table
- Detailed vendor profiles
- Appendix with source URLs
Save as .docx."
Pro Tips
- Upload reference files: Give Cowork an example report or template and say "match this format" for consistent output
- Be specific about file format: Say "Save as .xlsx" or "Export as .pptx" — don't leave it ambiguous
- Download files before session ends: Cowork environments are temporary — download everything you need before closing
- Chain complex tasks: "First clean the data, then analyze it, then create the report" — Cowork handles multi-step workflows
- Ask for formulas, not just values: In Excel output, say "use formulas so I can update the source data later"
- Iterate on output: "Make the charts bigger" / "Add a column for variance %" / "Change the color scheme to match our brand"
Module 10: AI + Spreadsheets
Claude in Excel, ChatGPT for Excel & AI-powered spreadsheet mastery
Transform How You Work with Data
AI in spreadsheets isn't just about formulas—it's analysis, cleaning, visualization, automation, and turning raw data into insights in seconds instead of hours.
Claude in Excel Overview
Claude in Excel is an official add-in that brings Claude's reasoning directly into your spreadsheets. It reads your data, understands context, and can work across multiple sheets.
Getting Started
1. Open Excel → Insert → Get Add-ins
2. Search "Claude for Excel" or "Anthropic"
3. Install and sign in with your Claude account
4. Access via the Claude panel on the right side
What Makes Claude Different
- Deep reasoning: Doesn't just generate formulas—explains WHY and catches edge cases
- Context awareness: Understands your full spreadsheet, headers, and data types
- Multi-step analysis: Can perform complex analysis requiring multiple operations
- Natural conversation: Ask follow-ups, refine, iterate on results
Claude in Excel: Power Prompts
Instant Data Understanding
"Look at my data and tell me:
1. What type of data is this? (sales, HR, financial, etc.)
2. What are the key columns and what do they represent?
3. Are there any data quality issues I should fix first?
4. What are the 3 most interesting questions this data could answer?"
Smart Formula Generation
"I need a formula that:
- Looks up an employee ID in column A
- Returns their department from column C
- But if they were hired after 2023 (column D), add ' (New)' to the result
- Handle errors gracefully if ID not found
Explain any limitations of the formula."
Data Validation & Cleaning
"Review column E (email addresses) and:
1. Flag any that don't look like valid emails
2. Identify duplicates
3. Find any with unusual domains
4. Suggest a formula to standardize them all to lowercase
Put your findings in a new column with clear status codes."
Executive Summary Generation
"Analyze this quarterly sales data and write an executive summary including:
- Total revenue and comparison to last quarter
- Top 3 performing products/regions
- Any concerning trends or anomalies
- 2-3 recommendations based on the data
Format it so I can paste directly into an email to leadership."
ChatGPT (GPT-4.1) + Excel Options
Multiple ways to use ChatGPT (now powered by GPT-4.1/o3) with Excel, from native integration to copy-paste workflows.
Option 1: Microsoft Copilot (Enterprise)
If your org has Microsoft 365 Copilot:
- Built directly into Excel ribbon
- Full access to your workbook data
- Can create formulas, charts, pivot tables
- Works with natural language commands
Access: Look for "Copilot" button in Excel ribbon
Option 2: ChatGPT with Code Interpreter
Upload your Excel/CSV file directly to ChatGPT:
1. Click the attachment icon in ChatGPT
2. Upload your .xlsx or .csv file
3. ChatGPT analyzes with Python (pandas)
4. Can create charts, clean data, perform analysis
5. Download modified files directly
Best for: Complex analysis, visualizations, Python-powered transformations
Option 3: Copy-Paste Workflow
For quick formula help without uploading files:
1. Copy your column headers + sample rows
2. Paste into ChatGPT with your question
3. Get formulas, explanations, or analysis
4. Paste results back into Excel
Best for: Quick formulas, sensitive data you can't upload
ChatGPT Excel Prompts
Upload & Analyze (Code Interpreter)
"I've uploaded our Q4 sales data. Please:
1. Show me a summary of the data structure
2. Calculate total sales by region and product category
3. Create a bar chart of top 10 products by revenue
4. Identify any outliers or anomalies in the data
5. Export a cleaned version with a new 'Sales Tier' column
(High >$10K, Medium $5-10K, Low <$5K)"
Complex Pivot Analysis
"From this data, create an analysis showing:
1. Monthly revenue trend with month-over-month % change
2. Customer segmentation by purchase frequency and value
3. Product affinity analysis (what's commonly bought together)
4. Cohort analysis: retention by signup month
For each, show me the data table AND a visualization."
Data Transformation
"Transform this data:
Current format: One row per transaction
Target format: One row per customer with columns for:
- Total lifetime value
- Number of orders
- First order date
- Last order date
- Average order value
- Days since last order
- Customer segment (New/Active/At-Risk/Churned)
Export as new Excel file."
Formula Generation Prompts
Stop Googling formulas. Describe what you need in plain English.
XLOOKUP & Advanced Lookups
"Create an XLOOKUP formula that:
- Searches for the value in A2 within the 'Products' sheet column A
- Returns the price from column D
- If not found, returns 'Not in catalog'
- Should work even if the product list isn't sorted"
Conditional Calculations
"I need a formula for commission calculation:
- Sales under $10K: 5% commission
- Sales $10K-$50K: 7% commission
- Sales $50K-$100K: 10% commission
- Sales over $100K: 12% commission
- If employee is in 'Probation' status (column C), cap at 5%
Make it easy to update the tiers later."
Text Manipulation
"Column A has full names like 'Smith, John R.'
I need:
- Column B: First name only (John)
- Column C: Last name only (Smith)
- Column D: Email format (jsmith@company.com)
- Column E: Display format (J. Smith)
Handle edge cases like no middle initial or suffixes like Jr./Sr."
Date Intelligence
"Create formulas for project tracking:
- Days until deadline (column B has due dates)
- Status: 'Overdue', 'Due This Week', 'Due This Month', 'On Track'
- Workdays remaining (exclude weekends)
- Weeks remaining (rounded up)
- Fiscal quarter of the due date (our FY starts April 1)"
Array Formulas & Dynamic Arrays
"Create a formula that returns a unique, sorted list of all departments from column C, excluding blanks and duplicates.
Then create another formula that counts employees per department and displays it next to each unique department.
Use the new dynamic array functions (UNIQUE, SORT, FILTER) if available."
Formula Debugging
Explain Complex Formulas
"Explain this formula step by step. I inherited this spreadsheet and need to understand what it does:
=IF(ISERROR(INDEX(Data!$B$2:$B$1000,MATCH(1,(Data!$A$2:$A$1000=A2)*(Data!$C$2:$C$1000="Active"),0))),"",INDEX(Data!$B$2:$B$1000,MATCH(1,(Data!$A$2:$A$1000=A2)*(Data!$C$2:$C$1000="Active"),0)))
Also suggest a cleaner way to write this using modern Excel functions."
Fix Broken Formulas
"This formula returns #VALUE! error:
=SUMIFS(B:B,A:A,">="&DATE(2024,1,1),A:A,"<="&DATE(2024,12,31))
My data:
- Column A: Dates (formatted as dates)
- Column B: Amounts (numbers)
Why is it breaking and how do I fix it?"
Data Analysis Prompts
Sales Analysis
"Analyze this sales data and answer:
1. What's our sales trend? Is it growing, flat, or declining?
2. Which products/regions are driving growth vs. dragging?
3. Is there seasonality? When are our strong/weak periods?
4. Who are our best customers by revenue and frequency?
5. What's our average deal size trend?
6. Are there any concerning patterns (declining customers, etc.)?
Support each finding with specific numbers from the data."
Financial Analysis
"Review this P&L data and provide:
1. Gross margin trend (monthly for past 12 months)
2. Expense ratio analysis - which categories are growing fastest?
3. Break-even analysis - at what revenue do we hit profitability?
4. Variance analysis - actuals vs budget, highlight items >10% off
5. Cash flow implications based on AR/AP aging
Flag anything that would concern a CFO."
HR/People Analytics
"Analyze this employee data:
1. Turnover rate by department and tenure band
2. Time-to-fill for open positions by role type
3. Compensation analysis - any pay equity concerns?
4. Performance rating distribution - is it truly differentiated?
5. Flight risk indicators - who might we lose?
Recommend 3 actions based on the data."
Comparative Analysis
"I have two sheets: 'This Year' and 'Last Year' with the same structure.
Create a comparison analysis showing:
1. Year-over-year change for each metric ($ and %)
2. Highlight metrics that improved significantly (>10%)
3. Flag metrics that declined significantly (>10%)
4. Calculate the overall health score trend
5. Identify the biggest movers (positive and negative)"
Visualization Requests
Chart Recommendations
"Looking at my data, recommend:
1. The best chart type to show the main trend
2. A chart to compare categories
3. A chart to show composition/breakdown
4. Any additional visualizations that would add insight
For each, tell me which columns to use and any formatting tips."
Dashboard Data Prep
"I need to create an executive dashboard. Prepare the data:
1. Create a summary table with KPIs (revenue, units, margin, etc.)
2. Add month-over-month and year-over-year comparisons
3. Create sparkline-ready data (last 12 months trend)
4. Prepare Top 10 and Bottom 10 lists
5. Create the data structure for a waterfall chart showing variance
Organize on a new 'Dashboard Data' sheet."
Excel Automation with AI
VBA Macro Generation
"Write a VBA macro that:
1. Loops through all sheets in the workbook
2. On each sheet, finds the last row with data
3. Adds a 'Total' row with SUM formulas for columns B through F
4. Formats the total row as bold with a top border
5. Creates a summary sheet with the totals from each sheet
Include error handling and comments explaining each section."
Data Import Automation
"Create a macro that:
1. Opens a file dialog to select a CSV file
2. Imports the data to a new sheet named with today's date
3. Auto-fits columns and applies header formatting
4. Removes any completely blank rows
5. Converts text-formatted numbers to actual numbers
6. Adds a timestamp in cell A1 noting when data was imported
Make it foolproof for non-technical users."
Report Generation Macro
"Create a macro that generates a weekly report:
1. Copy the 'Template' sheet and rename to current week
2. Pull data from 'Raw Data' sheet filtered to this week
3. Update all the summary formulas
4. Generate the charts
5. Export as PDF to a 'Reports' folder
6. Send via Outlook to a distribution list
Add a button I can click to run this each week."
Power Query Help
"Write Power Query M code to:
1. Connect to all .xlsx files in a folder
2. Combine them into one table
3. Add a column with the source filename
4. Unpivot the month columns (Jan-Dec) into rows
5. Filter out any rows where value is null or zero
6. Change data types appropriately
Explain each transformation step."
Shortcut AI & Other Tools
Specialized tools for different Excel workflows:
When to Use Each Tool
CLAUDE IN EXCEL:
✓ Complex analysis requiring reasoning
✓ When you need explanations, not just formulas
✓ Multi-step tasks with conversation
✓ Data quality assessment and recommendations
✓ Writing summaries based on data
CHATGPT + CODE INTERPRETER:
✓ Heavy data transformation (reshape, merge, pivot)
✓ Statistical analysis
✓ Creating visualizations
✓ Working with very large files
✓ Python-powered analysis
MICROSOFT COPILOT:
✓ Quick in-app formula help
✓ Creating pivot tables and charts
✓ Simple data questions
✓ Users who prefer staying in Excel UI
SHORTCUT AI:
✓ Formula-focused work
✓ Quick formula generation without explanation
✓ Users who want one-click formula insertion
✓ Rapid-fire formula creation sessions
Module 11: Shortcut AI
Lightning-fast Excel formulas with one click
Shortcut AI: Quick Reference
Shortcut AI is purpose-built for Excel formula generation. It's the fastest way to get a formula when you already know what you need.
Best Use Cases
- Rapid formula creation: Describe → Get formula → One-click insert
- Complex nested formulas: INDEX/MATCH, nested IFs, array formulas
- Staying in flow: When you don't want to context-switch to ChatGPT or Claude
- Formula conversion: Convert old formulas to modern equivalents
Quick Examples
"SUMIF where column A is this month and column B is 'Approved'"
"XLOOKUP that returns multiple columns at once"
"Count unique values in column C where column A matches my current row"
"Convert this VLOOKUP to XLOOKUP: =VLOOKUP(A2,Data!A:D,4,FALSE)"
"Dynamic range that expands as I add data"
Module 12: NotebookLM
Google's AI research assistant for deep document understanding
What is NotebookLM?
NotebookLM is Google's AI-powered research tool that creates a personalized AI expert on your uploaded documents. It grounds all responses in your sources - dramatically reducing hallucinations.
Key Features
- Source-grounded: All answers cite specific passages from your documents
- Audio Overview: Generates podcast-style discussions of your content
- Multi-source: Upload PDFs, docs, websites, YouTube videos
- Note-taking: Save and organize AI-generated insights
- Gemini-powered: Uses Google's best model for reasoning
- Spreadsheet uploads: Upload spreadsheets and get interactive charts and analysis
- NotebookLM Plus: Available for business use with enhanced features and higher limits
Best Use Cases
- Research synthesis: Upload multiple papers, get unified insights
- Meeting prep: Upload background docs, ask "what should I know?"
- Learning: Turn textbooks into interactive Q&A sessions
- Content creation: Generate podcast episodes from your research
- Due diligence: Upload company docs, extract key facts
Audio Overview Feature
NotebookLM's killer feature: it generates a ~10 minute podcast-style discussion between two AI hosts about your documents.
- Great for learning on the go
- Makes dense content engaging
- Customizable focus areas
- Shareable audio files
Activity: Create Your First Notebook
Upload 3-5 related documents (research, reports, articles). Ask NotebookLM to synthesize the key themes. Then generate an Audio Overview and listen during your commute.
Module 13: Nano Banana
AI-powered creative suite for marketing, photos, presentations & brand consistency
More Than Marketing Copy
Nano Banana is a full creative AI suite - edit photos, swap subjects, adjust colors, create marketing visuals, build presentations, and maintain consistent brand voice across all content. Nano Banana Pro is available for higher quality output and advanced features.
AI Photo Editing
Transform existing photos or create entirely new marketing visuals with AI.
Edit Existing Photos
"Take this team photo and:
- Remove the cluttered background
- Replace with a clean, modern office setting
- Adjust lighting to be brighter and more professional
- Keep all people exactly as they are"
Swap People & Objects
"In this product photo:
- Replace the model's outfit with business casual attire
- Swap the coffee cup for our branded mug
- Change the laptop to show our software interface
- Keep the same pose and lighting"
Adjust Colors & Scenes
"Modify this outdoor photo:
- Change from summer to autumn setting (fall colors)
- Adjust the sky from overcast to golden hour lighting
- Make colors warmer to match our brand palette
- Add subtle lens flare for professional look"
Create New Marketing Photos
"Create a professional marketing photo:
- Diverse team of 4 collaborating around a conference table
- Modern glass-walled office with city view
- Laptops and tablets visible showing data visualizations
- Natural lighting, candid but polished feel
- Corporate but approachable atmosphere"
Photo Enhancement Examples
Product Photography
"Enhance this product shot:
- Remove all background distractions → pure white
- Add professional product shadow
- Increase contrast and saturation slightly
- Create 3 variations: white bg, lifestyle setting, floating with shadow"
Headshots & Team Photos
"Professional headshot enhancement:
- Soften skin while keeping natural texture
- Brighten eyes slightly
- Even out lighting across face
- Add subtle professional background blur
- Match this style across all 12 team member photos"
Event & Action Shots
"Improve this conference photo:
- Straighten the horizon
- Remove the exit sign in background
- Brighten faces in the crowd
- Make our branded banner more prominent
- Crop to 16:9 for social media"
Marketing Content at Scale
Generate consistent, on-brand marketing content across all channels.
Social Media Campaigns
"Create a 2-week LinkedIn campaign (10 posts) for our new cloud security product:
Brand voice: Professional but approachable, thought leadership focused
Target: IT Directors and CISOs at mid-market companies
Key messages:
1. Zero-trust is no longer optional
2. Our solution deploys in days, not months
3. 40% cost reduction vs legacy solutions
For each post provide:
- Hook (first line that stops scrolling)
- Body (2-3 short paragraphs)
- CTA
- 3-5 hashtags
- Best posting time recommendation"
Email Sequences
"Create a 5-email nurture sequence for webinar registrants who didn't attend:
Email 1 (Day 0): Recording available + key takeaway teaser
Email 2 (Day 3): Most-asked question from Q&A + answer
Email 3 (Day 7): Case study related to webinar topic
Email 4 (Day 14): New related resource + soft demo offer
Email 5 (Day 21): Limited time offer / urgency close
Brand voice: Helpful, not pushy. Educational first.
Include subject line A/B variants for each."
Ad Copy Variations
"Generate Google Ads copy for our project management software:
Create 10 headline variations (30 char max each):
- 3 focusing on time savings
- 3 focusing on team collaboration
- 2 focusing on cost/ROI
- 2 focusing on ease of use
Create 5 description variations (90 char max each)
Include relevant keywords: project management, team collaboration, productivity"
Content Personalization
Industry-Specific Variations
"Take this base case study and create 4 industry versions:
Base: Generic cloud migration success story
Versions needed:
1. Healthcare - emphasize HIPAA, patient data, compliance
2. Financial Services - focus on security, audit trails, regulations
3. Industrial - highlight uptime, reliability, certifications, compliance
4. Manufacturing - emphasize uptime, IoT integration, supply chain
Keep core metrics same, adjust language and pain points for each vertical."
Persona-Based Content
"Rewrite this product page for 3 different buyer personas:
1. Technical Buyer (IT Manager):
- Emphasize architecture, security, integrations
- Include technical specifications
- Focus on implementation ease
2. Business Buyer (VP Operations):
- Focus on ROI, efficiency gains, risk reduction
- Minimize jargon, maximize business impact
- Include competitor comparison points
3. Executive Buyer (C-Suite):
- Strategic value, market positioning
- High-level metrics only
- Peer company references"
AI-Powered Presentations
Create professional presentations with consistent styling and compelling narratives.
Complete Presentation Generation
"Create a 12-slide sales presentation for our cybersecurity platform:
Audience: CISO and IT leadership at healthcare organizations
Goal: Get agreement to a technical demo
Time: 20 minutes + 10 min Q&A
Structure:
Slide 1: Title + compelling stat about healthcare breaches
Slides 2-3: Healthcare-specific threat landscape (current state)
Slides 4-5: Cost of breaches (financial, reputation, patient trust)
Slide 6: Why traditional solutions fall short
Slides 7-9: Our approach (3 key differentiators)
Slide 10: Customer proof points (healthcare logos + metrics)
Slide 11: Implementation timeline + low disruption message
Slide 12: Next steps + demo offer
For each slide provide: headline, 3-4 bullet points max, speaker notes, visual suggestion"
Presentation Enhancement
"Review my attached presentation and improve it:
1. Strengthen headlines - make each one a complete thought
2. Reduce text per slide to 4 bullets max
3. Suggest data visualizations to replace text where possible
4. Add a compelling story arc connecting the slides
5. Create speaker notes with key talking points
6. Identify slides that should be split or combined
7. Suggest where to add customer proof points"
Slide Design Prompts
"Create visual concepts for these presentation slides:
Slide: 'Our 3-Pillar Approach'
→ Visual: Three interconnected pillars or columns with icons
Slide: 'Customer Journey'
→ Visual: Horizontal timeline showing before/during/after states
Slide: 'ROI Calculator Results'
→ Visual: Dashboard-style layout with key metrics prominent
Slide: 'Competitive Landscape'
→ Visual: 2x2 matrix positioning us in upper right
For each, describe layout, colors (using our brand palette), and iconography."
Pitch Decks
Investor Pitch Deck
"Create a 10-slide Series A pitch deck:
1. Title + one-liner value prop
2. Problem (market pain point with stats)
3. Solution (what we do, simply stated)
4. Market Size (TAM/SAM/SOM)
5. Business Model (how we make money)
6. Traction (key metrics, growth chart)
7. Competition (our differentiation)
8. Team (founders + key hires)
9. Financials (projections, use of funds)
10. Ask (amount, timeline, next steps)
Style: Clean, modern, data-forward. Minimal text, maximum impact."
Internal Strategy Deck
"Create a Q2 strategy presentation for leadership:
Section 1: Q1 Performance Review
- Key wins and misses
- Metrics vs targets dashboard
Section 2: Market & Competitive Update
- Industry trends affecting us
- Competitor moves to watch
Section 3: Q2 Priorities
- Top 3 strategic initiatives
- Resource allocation
- Key milestones by month
Section 4: Risks & Mitigations
- Top 3 risks with response plans
Section 5: Team Needs
- Hiring priorities
- Budget requests"
Creating Reusable Brand Styles
Define styles once, apply consistently across all content for brand coherence.
Define Your Brand Voice
"Create a brand voice guide I can reuse for all content:
Company: [Your Company]
Industry: Manufacturing & Supply Chain
Voice Attributes:
- Professional but not stiff
- Confident but not arrogant
- Technical but accessible
- Trustworthy and reliable
Tone by Channel:
- LinkedIn: Thought leadership, industry insights
- Email: Helpful, relationship-focused
- Website: Clear, benefit-focused
- Proposals: Formal, precise, capability-focused
Words we USE: mission-critical, modernize, optimize, partner, enable
Words we AVOID: cheap, disrupt, hack, pivot, synergy
Save this as my 'Company Brand Voice' style."
Visual Style Templates
"Create a visual style guide for our marketing:
Primary Colors:
- Navy Blue (#003366) - headers, CTAs
- White (#FFFFFF) - backgrounds
- Light Gray (#F5F5F5) - secondary backgrounds
Accent Colors:
- Teal (#008080) - highlights, links
- Orange (#FF6600) - alerts, important callouts
Photo Style:
- Professional, diverse teams
- Modern office environments
- Natural lighting preferred
- Candid over posed when possible
Icon Style:
- Line icons, not filled
- 2px stroke weight
- Single color (navy or teal)
Save as 'Company Visual Style' for all image generation."
Apply Styles to Content
"Using my saved 'Company Brand Voice' style, rewrite this generic content:
[Paste generic content here]
Apply:
- Our voice attributes (professional, confident, accessible)
- Our preferred terminology
- Our tone for [specify channel: LinkedIn/email/website]
- Appropriate length for this channel"
Batch Apply Styles
"Apply 'Company Brand Voice' to these 5 pieces of content:
1. [Blog post draft]
2. [LinkedIn post]
3. [Email subject lines]
4. [Landing page copy]
5. [Case study summary]
For each, maintain the core message but adjust:
- Terminology to match our word preferences
- Tone to match the specific channel
- Length appropriate to format
Flag any content that conflicts with our brand guidelines."
Style Consistency Examples
Campaign Consistency
"I'm launching a campaign across multiple channels.
Using my brand styles, create consistent content for:
Core Message: 'Optimize your operations with smarter manufacturing'
Deliverables:
1. LinkedIn post (thought leadership angle)
2. Email subject + preview text
3. Landing page headline + subhead
4. Google Ad (headline + description)
5. Twitter/X post
6. Banner ad copy (728x90)
All should feel like they're from the same campaign while being optimized for each format."
Maintaining Consistency Over Time
"Review this new content against my saved brand style:
[Paste new content]
Check for:
- Voice consistency with our guidelines
- Terminology compliance (are we using approved words?)
- Tone match for the intended channel
- Visual style compliance (if describing imagery)
Provide specific edits to bring into alignment with our brand."
Module 14: Output Templates
Structured formats for consistent, professional AI outputs
Why Templates Matter
Templates turn AI from a generic assistant into a specialized tool for your work. By defining the exact output format, you get consistent, professional results every time.
Data Analysis Report Template
Format your analysis as follows:
## Executive Summary
[2-3 sentence overview of key findings]
## Key Metrics
| Metric | Value | vs. Prior Period | Trend |
|--------|-------|------------------|-------|
[Table of 3-5 most important metrics]
## Insights
1. [Most important finding with supporting data]
2. [Second finding]
3. [Third finding]
## Anomalies & Concerns
- [Any data quality issues]
- [Unexpected patterns]
## Recommendations
- [Action item 1]
- [Action item 2]
## Methodology Notes
[Brief description of data sources and approach]
## Executive Summary
[2-3 sentence overview of key findings]
## Key Metrics
| Metric | Value | vs. Prior Period | Trend |
|--------|-------|------------------|-------|
[Table of 3-5 most important metrics]
## Insights
1. [Most important finding with supporting data]
2. [Second finding]
3. [Third finding]
## Anomalies & Concerns
- [Any data quality issues]
- [Unexpected patterns]
## Recommendations
- [Action item 1]
- [Action item 2]
## Methodology Notes
[Brief description of data sources and approach]
Marketing White Paper Template
Structure the white paper as:
Title: [Compelling, benefit-focused title]
Subtitle: [Clarifying statement]
## The Challenge
[250-300 words describing the problem the reader faces.
Use industry statistics. Create urgency.]
## Why Traditional Approaches Fall Short
[200-250 words on limitations of current solutions]
## A Better Approach
[300-400 words introducing your methodology/solution.
Focus on principles, not product features.]
## Case Study / Example
[200-300 words with specific results and metrics]
## Implementation Framework
[Step-by-step guide, 3-5 phases]
## Key Takeaways
[5-7 bullet points summarizing main points]
## Next Steps
[Clear CTA with low-friction first step]
Title: [Compelling, benefit-focused title]
Subtitle: [Clarifying statement]
## The Challenge
[250-300 words describing the problem the reader faces.
Use industry statistics. Create urgency.]
## Why Traditional Approaches Fall Short
[200-250 words on limitations of current solutions]
## A Better Approach
[300-400 words introducing your methodology/solution.
Focus on principles, not product features.]
## Case Study / Example
[200-300 words with specific results and metrics]
## Implementation Framework
[Step-by-step guide, 3-5 phases]
## Key Takeaways
[5-7 bullet points summarizing main points]
## Next Steps
[Clear CTA with low-friction first step]
Sales Email Templates
Initial Outreach
Subject: [Specific benefit] for [Company]
Hi [Name],
[One sentence about their specific situation/trigger event]
[One sentence about relevant result you've achieved]
[One sentence with specific, low-commitment ask]
[Signature]
// Keep under 75 words. No attachments. One CTA.
Hi [Name],
[One sentence about their specific situation/trigger event]
[One sentence about relevant result you've achieved]
[One sentence with specific, low-commitment ask]
[Signature]
// Keep under 75 words. No attachments. One CTA.
Follow-Up After No Response
Subject: Re: [Original subject]
Hi [Name],
[Acknowledge they're busy - no guilt]
[New angle or piece of value - article, insight, data point]
[Softer ask - "Would it make sense to..." or "Is this on your radar?"]
[Signature]
// Don't reference first email negatively. Add new value.
Hi [Name],
[Acknowledge they're busy - no guilt]
[New angle or piece of value - article, insight, data point]
[Softer ask - "Would it make sense to..." or "Is this on your radar?"]
[Signature]
// Don't reference first email negatively. Add new value.
Quote Email Template
Subject: Your [Service/Product] Quote - [Company Name]
Hi [Name],
Thank you for the opportunity to provide this quote. Based on our conversation about [specific need discussed], here's what I recommend:
## Proposed Solution
[2-3 sentences describing what you're providing]
## Investment
| Item | Description | Price |
|------|-------------|-------|
[Line items]
| Total | | $X,XXX |
## What's Included
- [Key deliverable 1]
- [Key deliverable 2]
- [Key deliverable 3]
## Timeline
[Expected delivery or implementation timeframe]
## Next Steps
[Clear action for them to take to proceed]
This quote is valid for [30 days]. Happy to jump on a quick call if you have any questions.
[Signature]
Hi [Name],
Thank you for the opportunity to provide this quote. Based on our conversation about [specific need discussed], here's what I recommend:
## Proposed Solution
[2-3 sentences describing what you're providing]
## Investment
| Item | Description | Price |
|------|-------------|-------|
[Line items]
| Total | | $X,XXX |
## What's Included
- [Key deliverable 1]
- [Key deliverable 2]
- [Key deliverable 3]
## Timeline
[Expected delivery or implementation timeframe]
## Next Steps
[Clear action for them to take to proceed]
This quote is valid for [30 days]. Happy to jump on a quick call if you have any questions.
[Signature]
Quote Follow-Up Template
Subject: Re: Your [Service] Quote - Quick question
Hi [Name],
Wanted to check in on the quote I sent over on [date].
[Choose one approach:]
- "Any questions I can answer?"
- "Does the timeline still work for your needs?"
- "Would a quick call help clarify anything?"
[Signature]
// Keep it short. No pressure. One question only.
Hi [Name],
Wanted to check in on the quote I sent over on [date].
[Choose one approach:]
- "Any questions I can answer?"
- "Does the timeline still work for your needs?"
- "Would a quick call help clarify anything?"
[Signature]
// Keep it short. No pressure. One question only.
Activity: Build Your Template Library
Identify 3 types of outputs you create regularly. Build a template for each. Test them with AI and refine until the output matches your standards.
Module 15: Prompt Guides
Reference documents that make AI consistently follow your standards
What Are Prompt Guides?
Guides are reference documents you include in your prompts that teach AI your standards, processes, and preferences. Unlike templates (which define output format), guides define HOW to think and work.
Writing Style Guide
Include this at the start of writing tasks to ensure consistent voice:
## Company Writing Style Guide
Voice: Professional, confident, helpful. Not salesy or hyperbolic.
Tone: Direct and clear. Assume the reader is intelligent but busy.
Structure:
- Lead with the main point (inverted pyramid)
- Use short paragraphs (3-4 sentences max)
- Prefer bullet points for lists of 3+ items
- Use headers to break up content every 2-3 paragraphs
Word Choice:
- Active voice ("We recommend" not "It is recommended")
- Specific over vague ("37% increase" not "significant growth")
- Simple words over complex ("use" not "utilize")
Avoid:
- Jargon without explanation
- Passive voice
- Superlatives (best, leading, world-class)
- Starting sentences with "I" repeatedly
Company-Specific Terms:
- "Solutions" not "products"
- "Partners" not "vendors"
- "Clients" not "customers"
Voice: Professional, confident, helpful. Not salesy or hyperbolic.
Tone: Direct and clear. Assume the reader is intelligent but busy.
Structure:
- Lead with the main point (inverted pyramid)
- Use short paragraphs (3-4 sentences max)
- Prefer bullet points for lists of 3+ items
- Use headers to break up content every 2-3 paragraphs
Word Choice:
- Active voice ("We recommend" not "It is recommended")
- Specific over vague ("37% increase" not "significant growth")
- Simple words over complex ("use" not "utilize")
Avoid:
- Jargon without explanation
- Passive voice
- Superlatives (best, leading, world-class)
- Starting sentences with "I" repeatedly
Company-Specific Terms:
- "Solutions" not "products"
- "Partners" not "vendors"
- "Clients" not "customers"
Thinking & Structuring Guide
Use this for analytical tasks to ensure thorough reasoning:
## Analysis Framework Guide
Before answering, always:
1. Clarify what's actually being asked
2. Identify what information is given vs. assumed
3. Consider what's NOT being asked but might be relevant
When analyzing problems:
1. State the core issue in one sentence
2. List the key factors/variables
3. Identify constraints and dependencies
4. Consider 2-3 alternative interpretations
5. Flag assumptions you're making
When making recommendations:
1. Lead with your recommendation
2. Provide 2-3 supporting reasons
3. Acknowledge trade-offs or risks
4. Suggest how to validate or test
5. Define success criteria
Quality checks:
- "Have I answered the actual question?"
- "What would a skeptic challenge?"
- "What's the strongest counterargument?"
- "Am I certain, or should I caveat this?"
Before answering, always:
1. Clarify what's actually being asked
2. Identify what information is given vs. assumed
3. Consider what's NOT being asked but might be relevant
When analyzing problems:
1. State the core issue in one sentence
2. List the key factors/variables
3. Identify constraints and dependencies
4. Consider 2-3 alternative interpretations
5. Flag assumptions you're making
When making recommendations:
1. Lead with your recommendation
2. Provide 2-3 supporting reasons
3. Acknowledge trade-offs or risks
4. Suggest how to validate or test
5. Define success criteria
Quality checks:
- "Have I answered the actual question?"
- "What would a skeptic challenge?"
- "What's the strongest counterargument?"
- "Am I certain, or should I caveat this?"
Pricing & Quoting Guide
Reference for creating quotes and doing pricing analysis:
## Pricing & Quote Development Guide
Quote Components to Always Include:
- Base product/service cost
- Implementation/setup fees
- Training costs
- Support/maintenance (annual)
- Volume discounts if applicable
- Payment terms
- Quote validity period
Pricing Analysis Checklist:
1. What's the customer's budget range (if known)?
2. What are competitive alternatives?
3. What's our standard margin target? (specify %)
4. Are there volume tiers to consider?
5. What's the total contract value potential?
Red Flags to Check:
- Below-margin pricing
- Missing scope items
- Undefined deliverables
- Open-ended commitments
- Unusual payment terms
Presentation Rules:
- Always show value before price
- Break into logical line items
- Provide options when possible (good/better/best)
- Include "What's Included" section
- Make next steps crystal clear
Quote Components to Always Include:
- Base product/service cost
- Implementation/setup fees
- Training costs
- Support/maintenance (annual)
- Volume discounts if applicable
- Payment terms
- Quote validity period
Pricing Analysis Checklist:
1. What's the customer's budget range (if known)?
2. What are competitive alternatives?
3. What's our standard margin target? (specify %)
4. Are there volume tiers to consider?
5. What's the total contract value potential?
Red Flags to Check:
- Below-margin pricing
- Missing scope items
- Undefined deliverables
- Open-ended commitments
- Unusual payment terms
Presentation Rules:
- Always show value before price
- Break into logical line items
- Provide options when possible (good/better/best)
- Include "What's Included" section
- Make next steps crystal clear
Data Analysis Guide
## Data Analysis Standards Guide
Before Analysis:
- Document data source and date range
- Note any known data quality issues
- Confirm the business question being answered
- Identify key stakeholders and their needs
During Analysis:
- Check for missing values (report % missing)
- Identify outliers (>3 std dev = flag)
- Verify calculations on sample rows
- Document all assumptions
Statistical Standards:
- Report sample sizes for all metrics
- Use 95% confidence intervals
- Note when N is too small for conclusions
- Compare to prior periods and benchmarks
Presentation Standards:
- Round appropriately (2 decimals max)
- Use consistent number formats
- Include trend direction (↑↓→)
- Visualize when patterns matter
Always Caveat:
- Correlation ≠ causation
- Historical performance ≠ future results
- Small sample sizes limit conclusions
Before Analysis:
- Document data source and date range
- Note any known data quality issues
- Confirm the business question being answered
- Identify key stakeholders and their needs
During Analysis:
- Check for missing values (report % missing)
- Identify outliers (>3 std dev = flag)
- Verify calculations on sample rows
- Document all assumptions
Statistical Standards:
- Report sample sizes for all metrics
- Use 95% confidence intervals
- Note when N is too small for conclusions
- Compare to prior periods and benchmarks
Presentation Standards:
- Round appropriately (2 decimals max)
- Use consistent number formats
- Include trend direction (↑↓→)
- Visualize when patterns matter
Always Caveat:
- Correlation ≠ causation
- Historical performance ≠ future results
- Small sample sizes limit conclusions
Activity: Create Your Style Guide
Write a personal/team style guide covering: voice, tone, formatting preferences, words to use/avoid, and domain-specific terms. Test it by including at the start of 5 different prompts.
Module 16: AI Projects & Custom GPTs
Building persistent AI workspaces with preloaded context
Why Projects Matter
Projects transform AI from a blank slate into a specialized expert. By preloading context, files, and instructions, you skip the setup every time and get better results from the start.
Claude Projects
Projects in Claude let you create persistent workspaces with custom instructions and uploaded files.
Setting Up a Project
- Click "Projects" in Claude sidebar
- Create new project with descriptive name
- Add Custom Instructions (system prompt)
- Upload Knowledge Files (PDFs, docs, code)
- Start conversations within the project
Custom Instructions (System Prompt)
This is text that's included at the start of every conversation. Use it for:
## Role
You are a senior operations analyst specializing in manufacturing and supply chain optimization.
## Context
- We are a manufacturing operation with CNC, assembly, and finishing lines
- Our primary customers are OEM and contract manufacturing clients
- Key systems: ERP (inventory, orders), MES (production), QMS (quality)
## Instructions
- Always reference specific data sources when making recommendations
- Flag production risks and capacity constraints prominently
- Use our standard terminology (see style guide in files)
## Output Preferences
- Executive summary first
- Bullet points for lists
- Tables for comparisons
- Flag assumptions clearly
You are a senior operations analyst specializing in manufacturing and supply chain optimization.
## Context
- We are a manufacturing operation with CNC, assembly, and finishing lines
- Our primary customers are OEM and contract manufacturing clients
- Key systems: ERP (inventory, orders), MES (production), QMS (quality)
## Instructions
- Always reference specific data sources when making recommendations
- Flag production risks and capacity constraints prominently
- Use our standard terminology (see style guide in files)
## Output Preferences
- Executive summary first
- Bullet points for lists
- Tables for comparisons
- Flag assumptions clearly
Knowledge Files
Upload reference documents that Claude should use:
- Style guides: Writing standards, terminology
- Templates: Standard formats for outputs
- Reference docs: Policies, procedures, regulations
- Examples: Good outputs to emulate
- Data: Product info, pricing, specs
Custom GPTs (ChatGPT)
OpenAI's Custom GPTs let you create specialized AI assistants with persistent instructions and knowledge.
Creating a Custom GPT
- Go to ChatGPT → Explore GPTs → Create
- Use the GPT Builder (conversational) or Configure (direct)
- Define: Name, Description, Instructions
- Upload Knowledge files
- Configure Capabilities (web, code, images)
- Set sharing (private, link, public)
GPT Instructions Structure
# Identity
You are [specific role] who specializes in [domain].
# Task
Your primary function is to [main purpose].
# Knowledge
You have access to uploaded files containing [describe files].
Always reference these when answering questions about [topics].
# Behavior
- Always [specific behavior]
- Never [prohibited behavior]
- When uncertain, [fallback behavior]
# Output Format
Structure responses as [format description].
You are [specific role] who specializes in [domain].
# Task
Your primary function is to [main purpose].
# Knowledge
You have access to uploaded files containing [describe files].
Always reference these when answering questions about [topics].
# Behavior
- Always [specific behavior]
- Never [prohibited behavior]
- When uncertain, [fallback behavior]
# Output Format
Structure responses as [format description].
GPT vs. Projects Trade-offs
| Feature | Claude Projects | Custom GPTs |
|---|---|---|
| Sharing | Team only | Public possible |
| Web Access | Via features | Built-in toggle |
| Code Execution | Artifacts | Code Interpreter |
| File Size | Larger context | More limited |
| Reasoning | Generally stronger | Good, varies |
Best Practices for AI Projects
1. Start with the End in Mind
- What specific task will this project handle?
- What does a "good" output look like?
- Who will use it and how often?
2. Build Strong Grounding
- Include examples: Show good outputs in your files
- Define terminology: Create a glossary of key terms
- Set boundaries: Clarify what's in/out of scope
- Anticipate errors: Tell AI how to handle edge cases
3. Optimize Instructions
- Be specific: Vague instructions = inconsistent outputs
- Use structure: Headers, bullets, numbered lists
- Prioritize: Put most important rules first
- Test and iterate: Refine based on actual outputs
4. Curate Knowledge Files
- Quality over quantity: Relevant, accurate files only
- Keep current: Update when source docs change
- Organize clearly: Name files descriptively
- Remove redundancy: Don't upload duplicate info
Project Ideas by Role
- Sales: Proposal writer with win themes, case studies, pricing guides
- Legal: Contract reviewer with clause library, risk frameworks
- HR: Policy assistant with employee handbook, procedures
- Finance: Report generator with templates, accounting rules
- Marketing: Content creator with brand guide, tone examples
- PM: Status reporter with templates, project context
- Engineering: Code reviewer with style guide, best practices
Activity: Create Your First Project
Pick your most common recurring task. Create a Claude Project with: (1) Custom instructions defining role and output format, (2) 2-3 reference files, (3) Test with 5 different inputs and refine.
Module 17: Iteration & Fine-Tuning
Building a custom-tailored AI team through continuous improvement
The Compounding Advantage
Every time you refine a prompt, guide, or project based on real outputs, you're building a library of custom-tailored tools. Over time, your "first run" results get better and better, until AI becomes a reliable extension of your work style.
🔄 The Iteration Flywheel
1
Run
Execute your prompt/project
2
Review
What was good? What missed?
3
Refine
Update prompt/guide/project
4
Repeat
Test again, keep improving
Two Types of Iteration
1. Conversation-Level Iteration
Refining a single output through back-and-forth:
# Initial output: Too long, too formal
You: "Make this shorter - under 100 words"
AI: [Shorter version]
You: "Good length, but too formal. Make it conversational."
AI: [Conversational version]
You: "Better. Add a specific example in the second paragraph."
AI: [Final version with example]
You: "Make this shorter - under 100 words"
AI: [Shorter version]
You: "Good length, but too formal. Make it conversational."
AI: [Conversational version]
You: "Better. Add a specific example in the second paragraph."
AI: [Final version with example]
2. Prompt-Level Iteration
Updating your underlying prompt/guide/project so NEXT time it's right the first time:
# Original prompt:
"Summarize this document"
# After iteration 1 (too long):
"Summarize this document in under 100 words"
# After iteration 2 (too formal):
"Summarize in under 100 words. Use conversational tone."
# After iteration 3 (missing examples):
"Summarize in under 100 words. Conversational tone.
Include one specific example from the text."
→ Now SAVE this improved prompt for future use
"Summarize this document"
# After iteration 1 (too long):
"Summarize this document in under 100 words"
# After iteration 2 (too formal):
"Summarize in under 100 words. Use conversational tone."
# After iteration 3 (missing examples):
"Summarize in under 100 words. Conversational tone.
Include one specific example from the text."
→ Now SAVE this improved prompt for future use
What to Iterate On
Common Issues → Fixes to Add
| Issue | Add to Your Prompt/Guide |
|---|---|
| Too long | "Keep under X words/sentences/bullets" |
| Too short | "Provide detailed explanation with examples" |
| Wrong tone | "Use [formal/casual/technical] tone" |
| Missing context | "Always include [specific element]" |
| Wrong format | Add explicit format template |
| Uses wrong words | "Use [X] not [Y]" in style guide |
| Missing caveats | "Always note limitations/assumptions" |
| Too generic | "Be specific. Use concrete examples." |
Building Your AI Team
Think of each refined prompt/project as a "specialist" on your team:
- Email Writer: Knows your tone, sign-off style, typical length
- Data Analyst: Uses your preferred formats, includes your standard checks
- Report Generator: Follows your template, uses correct terminology
- Quote Builder: Includes right line items, knows your pricing rules
- Meeting Summarizer: Formats action items your way, highlights what you care about
Over time: You build a suite of 5-10 highly-tuned AI tools that give you exactly what you need on the first try. This is your competitive advantage.
Iteration Tracking Template
Keep a simple log to track improvements:
Prompt/Project: [Name]
Purpose: [What it does]
Version 1 (Date):
- Issue: [What was wrong]
- Fix: [What you changed]
Version 2 (Date):
- Issue: [What was wrong]
- Fix: [What you changed]
Current Status: [Working well / Needs more tuning]
Success Rate: [X% good on first try]
Purpose: [What it does]
Version 1 (Date):
- Issue: [What was wrong]
- Fix: [What you changed]
Version 2 (Date):
- Issue: [What was wrong]
- Fix: [What you changed]
Current Status: [Working well / Needs more tuning]
Success Rate: [X% good on first try]
Pro Tips for Faster Iteration
- Test with variety: Run 3-5 different inputs before declaring "done"
- Save good outputs: Use them as few-shot examples in your prompt
- Be specific in fixes: "Add bullet points" not "format better"
- Version your prompts: Keep old versions in case new ones regress
- Share with team: One person's iteration benefits everyone
- Review monthly: Are your prompts still working? AI models change.
Activity: Iteration Sprint
Take one prompt you use regularly. Run it 5 times with different inputs. After each run, note what could be improved and update the prompt. Track how your "first try" success rate improves.
Module 18: Data Automation
Have AI generate and run analysis scripts in real-time
The Power of Live Code Execution
Modern AI tools like Claude Cowork can write AND execute Python scripts on your data. This means you can do sophisticated analysis without knowing how to code.
The Workflow
- Upload your data: CSV, Excel, JSON, database exports
- Describe what you want: "Find outliers" or "Show monthly trends"
- AI writes the code: Python/pandas scripts generated automatically
- Code executes: Results appear immediately
- Iterate: "Now break it down by region" or "Add a chart"
Example Analyses You Can Request
Exploratory Analysis
"Analyze this sales data:
- Show summary statistics for each column
- Identify any missing or anomalous values
- Create visualizations of key distributions
- Highlight correlations between variables"
- Show summary statistics for each column
- Identify any missing or anomalous values
- Create visualizations of key distributions
- Highlight correlations between variables"
Trend Analysis
"For this time series data:
- Plot the trend over time
- Calculate month-over-month growth rates
- Identify seasonality patterns
- Flag any unusual spikes or drops"
- Plot the trend over time
- Calculate month-over-month growth rates
- Identify seasonality patterns
- Flag any unusual spikes or drops"
Segmentation
"Segment these customers:
- Group by purchase behavior (RFM analysis)
- Calculate average order value by segment
- Identify highest-value customer characteristics
- Create a visualization of segment distribution"
- Group by purchase behavior (RFM analysis)
- Calculate average order value by segment
- Identify highest-value customer characteristics
- Create a visualization of segment distribution"
Comparison
"Compare performance across regions:
- Calculate key metrics for each region
- Perform statistical significance tests
- Create side-by-side visualizations
- Identify top and bottom performers"
- Calculate key metrics for each region
- Perform statistical significance tests
- Create side-by-side visualizations
- Identify top and bottom performers"
Pro Tips for Data Automation
- Start broad, then narrow: "Give me an overview" → "Now focus on Q4"
- Ask for the code: "Show me the Python code you used"
- Request exports: "Save this as a CSV" or "Export the chart as PNG"
- Chain analyses: "Use those results to now calculate..."
- Verify edge cases: "What happens if I filter to only [X]?"
Activity: Automated Analysis
Upload a dataset (sales, customer, financial - any CSV/Excel). Ask for: (1) Overview summary, (2) Trend visualization, (3) Anomaly detection. Request the code and chart exports.
Module 19: Perplexity Labs Dashboards
Build real-time, auto-updating intelligence dashboards with live web data
Why This Is a Game-Changer
Perplexity Labs lets you create custom "Spaces" that continuously monitor the web and update dashboards with fresh intelligence. No coding, no API keys, no manual refresh - just describe what you want to track.
What is Perplexity Labs?
Perplexity Labs (labs.perplexity.ai) offers experimental features including Spaces - persistent research environments that can:
- Monitor sources continuously: Track websites, news, social media for changes
- Auto-update dashboards: Fresh data without manual queries
- Synthesize multiple sources: Combine news, financials, social signals
- Generate structured outputs: Tables, comparisons, trend analyses
- Share with teams: Collaborative intelligence workspaces
Setting Up a Dashboard Space
STEP 1: Create a New Space
Go to labs.perplexity.ai → Click "New Space" → Name it descriptively
Example: "Q1 2025 - Manufacturing Supply Chain Intelligence"
STEP 2: Configure Your Focus
Add a system prompt that defines the dashboard's purpose:
"This space monitors the manufacturing and supply chain market. Focus on:
- Major supplier announcements, M&A, and capacity changes
- Raw material price trends and commodity forecasts
- Tariff, trade policy, and reshoring developments
- Competitor wins and expansion announcements
- Emerging technology trends (automation, AI, digital twins)
Prioritize sources: IndustryWeek, Supply Chain Dive, Modern Materials Handling,
Thomas, MFG.com, and OEM press releases."
STEP 3: Add Initial Queries
Seed your dashboard with baseline questions:
- "What are the biggest supply chain disruptions impacting manufacturing this month?"
- "What are current lead time trends for [key materials]?"
- "Which manufacturers have announced automation or reshoring initiatives?"
Dashboard Best Practices
- Be specific about sources: Name the publications, databases, and sites you trust
- Set time horizons: "Focus on the last 30 days" or "Track changes since January 1"
- Define output formats: "Present as a table with columns: Date, Source, Key Finding, Relevance"
- Create comparison views: "Compare this week vs. last week" or "Show quarter-over-quarter changes"
- Add alert triggers: "Flag immediately if [competitor] announces [type of news]"
Customer Intelligence Dashboard
Create a living dashboard that tracks everything about your key accounts and prospects:
Space Configuration Prompt
CUSTOMER INTELLIGENCE SPACE
"This space tracks intelligence for our top 10 key accounts:
[Customer 1], [Customer 2], [Customer 3]... etc.
Monitor for each account:
1. Leadership changes (CEO, COO, VP Operations, Procurement Director)
2. Earnings calls, revenue reports, and investment announcements
3. New product launches, facility expansions, or capacity changes
4. Orders awarded to competitors
5. Pain points mentioned in earnings calls, news, or industry reports
6. Upcoming RFQs or sourcing events
7. Conference presentations by company leaders
Sources to prioritize:
- Company official websites and press releases
- IndustryWeek, Supply Chain Dive, trade publications
- SEC filings and earnings transcripts (if public)
- Thomas and industry directories
- LinkedIn posts from key decision-makers
- Industry association news
Output format:
Weekly summary table with: Account | Update Type | Summary | Source | Date | Relevance Score (1-5)
Flag anything with Relevance 4+ for immediate attention."
Sample Dashboard Queries
# Daily pulse check
"What's new today for [Customer Name]? Check news, press releases, and leadership LinkedIn."
# Pre-meeting intelligence
"I have a meeting with [Customer] next week. Create a briefing with:
- Recent news and announcements (last 30 days)
- Current operational priorities from earnings calls or press
- Key personnel and their backgrounds
- Recent orders or projects awarded in our space
- Potential pain points we can address"
# Opportunity tracking
"Show all sourcing activity from [Customer] in the last 60 days.
Include: RFQ details, project scope, estimated value, due date, qualification requirements."
# Relationship mapping
"Who are the key decision-makers at [Customer]? Include:
- Current role and tenure
- Previous positions
- Recent public statements or presentations
- LinkedIn activity themes"
Actionable Intelligence Examples
- "VP Operations departure detected" → Trigger: Update relationship map, research successor, adjust engagement strategy
- "New RFQ posted" → Trigger: Download immediately, assign to sales team, schedule response review
- "CapEx increase announced" → Trigger: Research specific projects funded, identify relevant contacts
- "Competitor win" → Trigger: Analyze winning approach, update competitive positioning, debrief if possible
- "Quality issue in earnings call" → Trigger: Map to our solutions, create targeted outreach materials
Competitive Intelligence Tracker
Monitor your competitors in real-time across all public channels:
Space Configuration Prompt
COMPETITIVE INTELLIGENCE SPACE
"This space tracks competitive intelligence for these companies:
[Competitor A], [Competitor B], [Competitor C], [Competitor D]
Monitor for each competitor:
1. Major customer wins (especially in manufacturing, industrial, supply chain)
2. New product/service announcements
3. Leadership changes and key hires
4. Partnership and acquisition news
5. Financial performance (if public)
6. Marketing messaging and positioning changes
7. Customer case studies and testimonials
8. Job postings that signal strategic direction
9. Conference presentations and thought leadership
10. Social media sentiment and engagement
Sources:
- Company websites, newsrooms, blogs
- Press release wires (PR Newswire, Business Wire)
- Industry publications relevant to our market
- Industry directories and trade publications
- LinkedIn company pages and key executive profiles
- Glassdoor for internal culture signals
- Job boards for hiring patterns
Output format:
Daily digest organized by competitor with: Date | Type | Headline | Source | Strategic Implication"
Competitive Analysis Queries
# Win/loss tracking
"What major orders or projects has [Competitor] won in the last 90 days?
Show: Date, customer, estimated value, description, and how they won it."
# Positioning analysis
"Analyze [Competitor's] recent marketing messages. What themes are they
emphasizing? How has their positioning changed in the last 6 months?
Compare to our messaging."
# Hiring signal analysis
"What roles is [Competitor] actively hiring for? Group by:
- Technical (what technologies are they building?)
- Sales/BD (what markets are they targeting?)
- Leadership (what capabilities are they adding?)
What does this suggest about their strategy?"
# Head-to-head comparison
"Create a comparison matrix: Us vs. [Competitor A] vs. [Competitor B]
Columns: Capability, Our Position, Competitor A, Competitor B, Our Advantage/Gap"
# Weakness identification
"What are [Competitor's] known weaknesses based on:
- Customer complaints/reviews
- Glassdoor employee feedback
- Failed projects or contract losses
- Technology limitations mentioned in analyst reports"
Market Trends Monitor
Track industry trends, emerging technologies, and market shifts:
Space Configuration Prompt
MARKET INTELLIGENCE SPACE
"This space monitors market trends in [your industry/focus area].
Track these trend categories:
1. Technology trends (AI, cloud, cybersecurity, etc.)
2. Regulatory changes and compliance requirements
3. Market size and growth projections
4. Emerging customer needs and pain points
5. Investment and M&A activity in the space
6. Analyst predictions and forecasts
7. Conference themes and hot topics
8. Thought leader perspectives
Key questions to answer weekly:
- What technologies are gaining/losing momentum?
- What new regulations should we prepare for?
- Where is investment money flowing?
- What problems are customers talking about most?
- What are analysts predicting for next 12-24 months?
Sources:
- Gartner, Forrester, IDC reports and blogs
- Industry-specific publications
- VC investment announcements
- Conference agendas and session topics
- Reddit and Stack Overflow discussions
- Patent filings in relevant technology areas
Output:
Weekly trend report with: Trend | Direction (↑↓→) | Evidence | Implication for Us"
Market Intelligence Queries
# Emerging technology tracking
"What's the latest on [emerging tech - e.g., AI agents] in manufacturing/supply chain?
- Who's adopting it?
- What use cases are working?
- What obstacles are people facing?
- When will it hit mainstream?"
# Regulatory radar
"What regulatory changes are coming in [industry] in the next 12 months?
Include: Regulation name, effective date, key requirements, compliance implications."
# Investment signals
"What startups in [space] have raised funding in the last 90 days?
Show: Company, funding amount, investors, what they're building.
What does this signal about market direction?"
# Problem identification
"What problems are [target customers] complaining about most on forums,
social media, and in industry discussions? Rank by frequency and severity."
Activity: Build Your First Dashboard
Go to labs.perplexity.ai and create a Space for one of: (1) Your top 3 target accounts, (2) Your top 3 competitors, or (3) A market trend you need to track. Configure the system prompt, add 5 seed queries, and check back daily for a week.
Module 20: Claude Artifacts Apps
Build custom tools and mini-applications instantly with Claude
No-Code App Building
Claude can create fully functional interactive tools - calculators, analyzers, generators, converters - that run right in your browser. Describe what you need, and Claude builds it. Save it, share it, iterate on it.
What Can Artifacts Do?
Artifacts are interactive React components Claude creates that run in your browser. They can:
- Process input: Text fields, file uploads, form data
- Apply logic: Calculations, transformations, validations
- Generate output: Formatted text, tables, visualizations
- Maintain state: Remember inputs, track progress, save results
- Export results: Copy to clipboard, download as files
Think of them as instant custom software - built in seconds, tailored to your exact need.
How to Request an Artifact
BASIC PATTERN:
"Create an interactive tool that [does X]"
"Build me a [type of app] that takes [input] and produces [output]"
"Make a React artifact that lets me [accomplish task]"
EXAMPLE REQUEST:
"Create an interactive tool that helps me rewrite text in different tones.
It should have:
- A text input area for my original content
- Dropdown to select tone (professional, casual, persuasive, simplified)
- A button to generate the rewritten version
- Output area showing the result
- Copy button for the output"
Claude will create a working app you can use immediately.
Types of Artifacts You Can Build
🔧 Utilities
Converters, calculators, formatters, validators
✍️ Writing Tools
Rewriters, summarizers, generators, editors
📊 Analyzers
Text analysis, data processing, scoring tools
📋 Forms
Input collectors, checklists, structured templates
📈 Visualizations
Charts, graphs, interactive diagrams
🎮 Interactive
Quizzes, decision trees, workflows
Tool 1: Universal Rewriter
Build a tool that transforms any text into different styles:
PROMPT:
"Create an interactive text rewriter tool with these features:
Input Section:
- Large text area for original content
- Character count display
Controls:
- Tone selector: Professional, Casual, Persuasive, Simplified, Executive
- Length adjustment: Shorter (-50%), Same, Longer (+50%)
- Audience selector: Technical, Non-technical, Executive, Public
Output Section:
- Rewritten text display
- Copy to clipboard button
- 'Try another variation' button
Bonus:
- Show what was changed (highlight differences)
- Explain why changes were made"
Tool 2: Email Composer
PROMPT:
"Build an email composition assistant with:
Input Fields:
- To (role/relationship): Manager, Client, Team, Vendor, Executive
- Subject/Topic: Brief description of email purpose
- Key points: Bullet list of what to communicate
- Tone: Request, Inform, Apologize, Persuade, Thank
Options:
- Length: Brief (2-3 sentences), Standard, Detailed
- Urgency level: Low, Medium, High
- Include: Call to action, Timeline, Next steps
Output:
- Generated email with subject line
- Copy button
- 'Make it shorter/longer' buttons
- 'More formal/casual' buttons"
Tool 3: Executive Summary Generator
PROMPT:
"Create an executive summary generator that:
Input:
- Text area for full document/report content
- Target audience dropdown: C-Suite, Board, Manager, Client
- Summary length: 1 paragraph, 3 bullets, half-page
Processing:
- Extract key findings/conclusions
- Identify critical numbers/metrics
- Note recommendations and next steps
- Flag risks or concerns
Output:
- Structured executive summary with:
• Bottom Line Up Front (BLUF)
• Key Findings (3-5 bullets)
• Recommendations
• Next Steps/Timeline
- Copy and download buttons"
Tool 4: RFP Requirements Analyzer
PROMPT:
"Build an RFP requirements analyzer tool:
Input:
- Text area to paste RFP/SOW content
- Checkboxes for what to extract:
☐ Mandatory requirements (shall/must)
☐ Evaluation criteria
☐ Submission requirements
☐ Key dates and deadlines
☐ Compliance requirements
☐ Questions to ask
Output Table:
| Requirement | Section | Type | Priority | Our Response Status |
With dropdown for status: Not Started, In Progress, Complete, N/A
Features:
- Highlight 'shall' vs 'should' vs 'may' language
- Flag ambiguous requirements that need clarification
- Export to CSV for compliance matrix
- Count total requirements by category"
Tool 5: Proposal Scorer
PROMPT:
"Create a proposal section scoring tool:
Setup:
- Input evaluation criteria (paste or enter manually)
- Weight each criterion (must sum to 100%)
Evaluation:
- Text area to paste proposal section
- For each criterion, tool assesses:
• Does section address this criterion? (Yes/Partial/No)
• Strength of response (1-5 scale)
• Evidence provided? (Yes/No)
• Specific quotes that support score
Output:
- Overall score with breakdown by criterion
- Strengths identified
- Gaps and weaknesses
- Specific improvement recommendations
- Comparison to 'ideal' response elements"
Tool 6: Meeting Notes Processor
PROMPT:
"Build a meeting notes analyzer:
Input:
- Text area for raw meeting notes/transcript
- Meeting type: Client, Internal, Interview, Planning
- Attendees (optional): Names and roles
Auto-Extract:
- Key decisions made (with who decided)
- Action items (with owner, due date if mentioned)
- Open questions requiring follow-up
- Commitments made by us
- Commitments made by them
- Important quotes to remember
- Follow-up meeting needs
Output Formats:
- Structured summary for CRM
- Email-ready recap
- Task list for project management tool
- Copy buttons for each format"
Tool 7: Pricing Calculator
PROMPT:
"Create a professional services pricing calculator:
Inputs:
- Role/level selection with hourly rates
- Number of hours per role
- Contract type: T&M, Fixed Price, Cost Plus
- Duration in months
- Optional: Travel, materials, subcontractors
Calculations:
- Labor costs by role and total
- Overhead/G&A/Fee calculations
- Travel estimates
- Contingency percentage
- Total contract value
Output:
- Summary pricing table
- Cost breakdown chart (pie chart)
- Comparison of different fee structures
- Export for proposal insertion"
Tool 8: Competitive Comparison Matrix
PROMPT:
"Build an interactive competitive comparison matrix:
Setup:
- Add competitors (columns)
- Add comparison criteria (rows)
- Weight criteria by importance
Scoring:
- For each cell, dropdown: Strong (3), Adequate (2), Weak (1), Unknown (?)
- Optional: Add notes/evidence for each score
Visualization:
- Color-coded matrix (green/yellow/red)
- Spider/radar chart comparing all competitors
- Bar chart by criteria
- Overall weighted score calculation
Output:
- Exportable matrix as table
- 'Our advantages' summary
- 'Areas to address' summary
- Talking points vs each competitor"
Tool 9: Decision Framework Tool
PROMPT:
"Create a decision-making framework tool:
Framework Options:
- Pros/Cons list
- Decision matrix (weighted criteria)
- SWOT analysis
- Risk/Reward assessment
Input:
- Decision question/context
- Options being considered
- Criteria that matter
- Weights for criteria (optional)
Analysis:
- Structured evaluation of each option
- Quantitative scoring where applicable
- Visualization of comparison
- Sensitivity analysis (what if weights change?)
Output:
- Recommendation with rationale
- Key factors driving the decision
- Risks of the recommended path
- Implementation considerations"
Artifact Power User Tips
- Start simple, then iterate: "Add a feature that..." or "Also include..."
- Request specific UI elements: "Use tabs for different sections" or "Make it a wizard with steps"
- Ask for styling: "Make it look modern and professional" or "Use a blue color scheme"
- Add validation: "Validate that [field] is in the correct format"
- Request export options: "Add buttons to copy as markdown and download as CSV"
- Include help text: "Add tooltips explaining each field"
Iterating on Artifacts
# Fixing issues
"The [feature] isn't working correctly. It should [expected behavior]."
# Adding features
"This is great! Can you also add [new feature]?"
# Changing design
"Make the interface more compact" or "Use a dark mode theme"
# Improving UX
"Add loading states and error messages"
# Making it smarter
"Have it automatically detect [pattern] and suggest [action]"
Saving and Sharing Artifacts
- Save the conversation: Star the chat to find it again
- Copy the code: Ask Claude to show the full code, save as .html file
- Share via conversation: Share the Claude conversation link
- Recreate easily: Save your prompt - you can always regenerate it
- Export functionality: Build export buttons into the artifact itself
Activity: Build Your First Custom Tool
Identify a repetitive task you do weekly. Ask Claude to build an artifact that automates or streamlines it. Test it with real data. Iterate to improve it. Share with a colleague who has the same need.
Module 21: AI Workflows
Chaining AI tools together for end-to-end automation
Building AI Pipelines
Real productivity gains come from connecting AI tools into workflows:
Example: Meeting → Action Pipeline
1. Otter.ai transcribes meeting
2. → Claude extracts action items
3. → Zapier creates tasks in Asana
4. → Email summary sent to attendees
2. → Claude extracts action items
3. → Zapier creates tasks in Asana
4. → Email summary sent to attendees
Example: Research → Report Pipeline
1. Perplexity researches topic
2. → NotebookLM synthesizes sources
3. → Claude drafts report
4. → Cowork formats as Word doc
2. → NotebookLM synthesizes sources
3. → Claude drafts report
4. → Cowork formats as Word doc
Workflow Tools
- Zapier: Connect 5000+ apps, no-code automation
- Make (Integromat): More complex logic, visual builder
- n8n: Self-hosted option, very flexible
- Power Automate: Microsoft ecosystem integration
Module 22: AI Safety & Ethics
Using AI responsibly and knowing its limits
Critical Understanding
AI is powerful but not infallible. Understanding risks is essential for professional use.
When NOT to Use AI
- Legal decisions: AI is not a lawyer, don't rely on it for legal advice
- Medical diagnosis: Never use AI for health decisions
- Financial advice: AI doesn't know your situation
- Confidential data: Don't upload sensitive info to public AI
- Final authority: Always have human review for important decisions
Data Privacy Best Practices
- Anonymize data: Remove PII before uploading
- Check policies: Understand what each tool does with your data
- Use enterprise versions: They typically have better data handling
- Don't share secrets: API keys, passwords, proprietary info
Verification Checklist
- Did AI cite sources? Are they real?
- Do numbers/statistics seem plausible?
- Is this within AI's knowledge cutoff?
- Have I checked claims I'll repeat?
- Would I stake my reputation on this?
AI Governance for Manufacturing
- Data classification: Establish what data is safe for cloud AI (general procedures, public specs) vs. what stays internal (proprietary formulations, customer contracts, trade secrets)
- IP protection: Never upload patented processes, proprietary designs, or competitive advantage documentation to public AI tools without enterprise agreements in place
- Quality decision verification: AI can recommend but should never be the sole authority on quality accept/reject decisions, safety-critical specifications, or regulatory compliance determinations
- Audit trails: Maintain records of AI-assisted decisions for ISO, FDA, or industry-specific audit requirements
- Human-in-the-loop: All AI outputs affecting production, safety, or customer deliverables require human review and sign-off
Enterprise & On-Premise Options
- Claude Enterprise & Claude for Work: Your data is never used to train models. SSO, admin controls, and audit logs included.
- On-premise deployment: Deploy AI models within your own infrastructure via AWS Bedrock or Google Vertex AI — data never leaves your cloud environment
- Data residency: Enterprise plans offer data residency controls for compliance with regional regulations
- SOC 2 Type II: Major AI providers maintain SOC 2 compliance for enterprise tiers
Module 23: Writing & Communications
Apply CRAFT, PDCA, and Chain-of-Thought to real writing tasks
Techniques Used in This Module
CRAFT Framework • PDCA Cycle • Chain-of-Thought • Role Prompting • Verification/Skeptic
Email with CRAFT Framework
Transform a simple email request into a precise output using the full CRAFT structure:
# CRAFT Framework Applied to Email
CONTEXT: I'm a project manager responding to a customer
who is frustrated about a 2-week delay in our production delivery.
The delay was caused by a raw material shortage from a Tier 2
supplier that we only learned about last week.
ROLE: Act as an experienced operations communications
specialist who balances accountability with advocacy.
ACTION: Draft a professional email that:
1. Acknowledges their frustration without being defensive
2. Explains the root cause objectively
3. Presents our mitigation plan with specific dates
4. Maintains the relationship for future work
FORMAT: Professional email format with:
- Subject line
- Greeting
- 3 short paragraphs maximum
- Clear next steps
- Professional sign-off
TONE: Empathetic but confident. Avoid over-apologizing.
Sound like a trusted partner solving a problem together.
PDCA Iteration Cycle
After getting the first draft, apply PDCA to refine:
# CHECK phase - Ask AI to critique its own work
"Before I send this, please review it as if you were the
frustrated client receiving this email. What would your
concerns be? What's missing? What might make you more
or less confident in our team?"
# ACT phase - Refine based on critique
"Good feedback. Now rewrite incorporating:
1. The specific concern about milestone visibility
2. A more concrete timeline commitment
3. A brief mention of the quality improvement this
delay actually enables"
Quote/Proposal with Chain-of-Thought
Use Chain-of-Thought for complex proposal sections:
TASK: I need to write a technical proposal for a customer
RFQ for a custom manufacturing solution with tight tolerances.
The requirement is 5 pages maximum.
CHAIN-OF-THOUGHT INSTRUCTION:
Before drafting, think through step by step:
1. What are the key evaluation criteria buyers use (price, quality, lead time, capability)?
2. What differentiates winning technical approaches?
3. How should I structure 5 pages effectively?
4. What specific proof points strengthen credibility (certifications, past projects, equipment)?
Then provide:
- Recommended structure with page allocation
- Key themes to emphasize
- Specific elements that score well with procurement teams
- Draft of the opening paragraph that hooks the reader
Role-Based Review
Have AI review from multiple stakeholder perspectives:
# Multi-role verification
"Review this proposal section from three perspectives:
1. As the Procurement Manager:
Does this comply with all RFQ requirements?
Is pricing justified and competitive?
2. As the Plant Engineer:
Is the approach sound? Are claims substantiated?
Does it show understanding of our production challenges?
3. As the Competing Supplier:
What weaknesses could a competitor exploit?
What's missing that others might include?
Provide specific recommendations from each perspective."
Policy Document Workflow
Complete PDCA cycle for policy development:
# PLAN: Research and Structure
"I need to create a remote work policy for a 200-person
manufacturing company. Before drafting, analyze:
- What sections are essential in such policies?
- What compliance considerations apply (OSHA, ISO, ERP access)?
- What common gaps lead to policy failures?
Create an outline with rationale for each section."
# DO: Generate with CRAFT
"Now draft Section 3: Equipment and Security Requirements.
CONTEXT: We handle proprietary production data, employees use company laptops
ROLE: HR policy writer with manufacturing operations knowledge
FORMAT: Numbered policy statements with rationale notes
TONE: Clear, authoritative, but not punitive"
# CHECK: Skeptic Review
"Act as an employee trying to find loopholes in this policy.
What's ambiguous? What scenarios aren't covered?
Then act as a security auditor - what risks remain?"
# ACT: Final Refinement
"Revise to close the gaps identified. Add a FAQ section
addressing the top 5 questions employees would have."
Module 24: Data & Analysis
Apply structured prompting to data exploration and insights
Techniques Used in This Module
PDCA Cycle • Chain-of-Thought • Few-Shot Examples • Verification • Tool Chaining
Data Exploration with PDCA
# PLAN: Understand before analyzing
"I'm uploading a CSV of purchase orders from last
fiscal year. Before any analysis, please:
1. Describe the data structure (columns, types, row count)
2. Identify any data quality issues (nulls, outliers, inconsistencies)
3. Suggest what questions this data can and cannot answer
4. Note any columns that need transformation for analysis
Think through this systematically before responding."
# DO: Initial Analysis
"Based on your assessment, perform these analyses:
1. Top 10 suppliers by total spend
2. Distribution of order types (standard, blanket, expedited)
3. Year-over-year trend if date column allows
4. Any concentration risks (single supplier dominance)"
# CHECK: Verify Findings
"Before I present these findings:
- Double-check the calculations by showing your work
- Flag any results that seem suspicious or counterintuitive
- Note confidence level (high/medium/low) for each insight
- What additional data would strengthen these conclusions?"
# ACT: Refine for Presentation
"Create an executive summary with:
- 3 key findings with supporting numbers
- 2 areas of concern with recommendations
- Suggested visualizations for each insight"
Chain-of-Thought Analysis
Force systematic thinking for complex data questions:
QUESTION: "Why did Q3 output drop 15% despite more orders on the books?"
CHAIN-OF-THOUGHT PROMPT:
"Analyze this step by step:
Step 1: What factors could cause output to drop?
Step 2: What factors could cause order count to rise?
Step 3: What combination explains both happening together?
Step 4: What data in our spreadsheet can test each hypothesis?
Step 5: Run those tests and report findings
Step 6: What's the most likely explanation with confidence level?
Show your reasoning at each step."
Few-Shot Pattern Analysis
Train AI to recognize patterns with examples:
# Few-shot for anomaly detection
"I'll show you some data patterns and whether they're anomalies:
Example 1: PO value $50K, typical range $40-60K → NORMAL
Example 2: PO value $500K, typical range $40-60K → ANOMALY (10x typical)
Example 3: Order date Dec 31, most orders placed Oct-Nov → SUSPICIOUS (year-end spend)
Example 4: 5 POs to same vendor in 1 week → SUSPICIOUS (possible split to avoid approval threshold)
Now analyze rows 45-100 for similar patterns.
For each anomaly found, explain WHY it's suspicious."
Script Generation with Verification
Use Claude Cowork for automated analysis with built-in checks:
# In Claude Cowork
"Generate a Python script that:
1. Reads our purchase_orders.csv
2. Calculates monthly spend totals by supplier
3. Identifies month-over-month changes > 20%
4. Outputs a summary table and saves to analysis_results.xlsx
IMPORTANT: Include these verification steps:
- Print row counts before and after any filtering
- Validate date parsing worked correctly
- Flag if any expected columns are missing
- Include a sanity check that totals reconcile
Run the script and show me the output before finalizing."
Multi-Tool Workflow
Chain tools for comprehensive analysis:
WORKFLOW: Quarterly Procurement Analysis
1. Claude in Excel → Initial data cleaning
"Clean this data: standardize supplier names,
fix date formats, flag missing values"
2. Claude Cowork → Statistical analysis
"Run trend analysis and generate visualizations
Output as PNG charts and summary stats"
3. Claude → Narrative insights
"Given these findings [paste], write executive
summary with PDCA-based recommendations"
4. Cowork → Final deliverable
"Create a polished Word doc combining
charts, data tables, and narrative"
Module 25: Research & Intelligence
Apply verification, synthesis, and systematic research methods
Techniques Used in This Module
PDCA Cycle • Verification/Skeptic • Role Prompting • Tool Chaining • Source Triangulation
Competitive Analysis with Verification
Research competitors systematically with built-in fact-checking:
# Multi-tool research workflow
STEP 1: Perplexity Research
"Research [Competitor Name] recent business activity:
- Major customer wins or partnerships announced
- Key industries and verticals they serve
- Their stated differentiators and capabilities
- Recent news or leadership changes"
STEP 2: Claude Analysis with Verification
"Based on this research [paste Perplexity output]:
FIRST: Flag any claims without clear sources
SECOND: Note what information is likely outdated
THIRD: Identify gaps - what do we still not know?
THEN: Create a competitor profile with confidence levels:
- HIGH confidence: Verified with multiple sources
- MEDIUM: Single source or dated information
- LOW: Inference or industry assumption"
STEP 3: Skeptic Check
"Now critique this profile. If you were the competitor's
CEO, what would you say is wrong or misleading about
our assessment? What are we likely overestimating or
underestimating?"
Market Research PDCA
# PLAN: Define research scope
"I need to understand the [industrial automation / ERP / MRO supplies] market.
Before researching, help me define:
1. What specific questions need answers?
2. What sources are most authoritative for manufacturing/supply chain?
3. What time horizon matters (current vs. 3-year trend)?
4. What would make this research actionable vs. just interesting?"
# DO: Systematic research
"Search for information on:
1. Market spending trends (cite analyst reports, industry associations)
2. Major purchasing channels (distributors, GPOs, direct)
3. Emerging requirements (sustainability, reshoring, digital twins)
4. Key decision-makers and their stated priorities"
# CHECK: Validate and triangulate
"For each major finding:
- Is the source authoritative for manufacturing/supply chain?
- Can we find corroboration from a second source?
- What's the date of this information?
- What contrary evidence exists?"
# ACT: Actionable synthesis
"Create a 1-page market brief with:
- 3 opportunity areas with evidence
- 2 market risks to monitor
- Recommended positioning for our firm
- Suggested next steps for BD team"
Document Synthesis with NotebookLM + Claude
Combine tools for comprehensive document analysis:
WORKFLOW: RFQ / Spec Package Analysis
1. NotebookLM → Load all RFQ documents
Upload: RFQ, engineering specs, drawings, Q&A responses
2. NotebookLM → Extract key requirements
"List all mandatory requirements with section references"
"What are the evaluation criteria and weights?"
"What risks or tolerances does the customer call out?"
3. Claude → Strategic analysis
"Given these requirements [paste NotebookLM output]:
Apply Chain-of-Thought analysis:
- What is the customer really trying to solve?
- What past problems does this RFQ address?
- What would a winning solution emphasize?
- Where can we differentiate from competitors?
Be specific and cite requirement numbers."
4. Claude → Compliance matrix draft
"Create a compliance matrix mapping each requirement
to our proposed solution. Flag any requirements where
we need to develop capability or partner."
Source Quality Assessment
Train yourself to evaluate AI research critically:
SOURCE QUALITY FRAMEWORK:
"For each source in this research, evaluate:
Authority: Who published this? Government source? Industry analyst?
Trade publication? Blog?
Currency: When was this published? Is it still relevant?
For supply chain: pre or post recent disruptions?
Accuracy: Can claims be verified? Do numbers match other sources?
Any obvious errors that reduce credibility?
Purpose: Is this objective reporting or advocacy?
Who benefits from this perspective?
Rate each source: Tier 1 (highly reliable) / Tier 2 / Tier 3 (use with caution)"
Module 26: Meetings & Collaboration
Apply structured techniques to meeting preparation and follow-up
Techniques Used in This Module
CRAFT Framework • PDCA Cycle • Role Prompting • Chain-of-Thought • Verification
Meeting Prep with CRAFT
CONTEXT: I have a meeting tomorrow with the VP of Operations at [Company]
to discuss our automation proposal. This is a follow-up
to our initial pitch 2 weeks ago. They asked for more details
on our implementation timeline and integration with their ERP.
ROLE: Act as an experienced manufacturing sales consultant who has
prepared executives for hundreds of customer meetings.
ACTION: Create a comprehensive meeting prep package:
1. Likely questions they'll ask (based on their concerns)
2. Our key messages (3 max, memorable)
3. Potential objections and responses
4. Questions WE should ask to advance the opportunity
5. Red flags to watch for that signal problems
FORMAT:
- 1-page quick reference I can review in 5 minutes
- Talking points, not full scripts
- Include specific data points to cite
TONE: Strategic and practical. Assume I know our offering
well but need help anticipating their perspective.
Role-Play Preparation
Use AI to practice difficult conversations:
# Simulate the meeting
"Role-play as the VP of Operations at [Company]. You are:
- Under pressure to improve throughput and reduce costs
- Burned by a previous vendor who overpromised
- Skeptical of supplier claims about ROI
- Concerned about disruption to current production
I'll practice my pitch. Challenge me with tough but
realistic questions. After each exchange, briefly note
what worked and what I should adjust.
Start by asking me why you should consider our firm."
# After role-play
"Based on our practice session, give me:
1. The 3 moments where I was most effective
2. The 3 areas where I need to strengthen my response
3. One specific phrase or approach to add to my prep"
Meeting Notes → Actions (Chain-of-Thought)
INPUT: [Paste meeting transcript or notes]
CHAIN-OF-THOUGHT PROCESSING:
"Process these meeting notes step by step:
Step 1: Identify Decisions
What was actually decided? (Not just discussed)
Who has authority for each decision?
Step 2: Extract Action Items
What needs to happen next?
Who owns each action?
What's the deadline (stated or implied)?
Step 3: Note Open Questions
What was raised but not resolved?
What information is still needed?
Who needs to provide it?
Step 4: Read Between the Lines
What concerns were implied but not stated directly?
What political dynamics were at play?
What risks or opportunities emerged?
Step 5: Suggest Follow-Up
What should our next steps be?
Who should we loop in?
What's the timeline for follow-up?"
Verification Step
# Before finalizing meeting summary
"Review this meeting summary and flag:
1. Ambiguous action items: Any task without clear owner or deadline?
2. Assumed decisions: Did we actually decide X, or just discuss it?
3. Missing context: Would someone not in the meeting understand this?
4. Potential miscommunication: Any item people might interpret differently?
Also: Is there anything in my notes that could be
embarrassing or problematic if forwarded to the client?"
Follow-Up Email (PDCA)
# PLAN: Strategic approach
"I need to send a follow-up email after our meeting with [Company VP Ops].
Key outcomes: They liked our implementation approach, want a pilot proposal,
concerned about our bandwidth given another project we mentioned.
Before drafting, analyze:
- What's the primary goal of this follow-up?
- What's the secondary goal (relationship building)?
- What tone reinforces our positioning as trusted advisor?
- What should I definitely NOT include?"
# DO: Draft with CRAFT
"Draft the follow-up email:
CONTEXT: Post-meeting, they're interested but have concerns
ROLE: Trusted advisor, not desperate salesperson
FORMAT: Thank you + summary + next steps + personal touch
TONE: Confident, responsive, relationship-focused"
# CHECK: Multi-perspective review
"Review this email as:
1. The CIO receiving it - does it address my concerns?
2. Their assistant who screens emails - is it clear and actionable?
3. Our competitor - what intelligence does this reveal?"
# ACT: Final refinements
"Make it 20% shorter while keeping all substance.
Ensure the ask is crystal clear.
Add one specific personal touch from the meeting."
Meeting Intelligence Capture
Systematic capture for CRM and future reference:
CRM UPDATE TEMPLATE:
"Based on these meeting notes, extract information for our CRM:
Contact Intelligence:
- Decision maker priorities/concerns
- Communication style preferences
- Personal interests mentioned
- Key relationships (who do they trust?)
Opportunity Intelligence:
- Budget signals (approved? seeking? timeline?)
- Competition mentioned
- Evaluation criteria emphasized
- Timeline and urgency indicators
Next Actions:
- Immediate follow-up required
- Future touchpoints to schedule
- Information to gather before next contact
Format as structured data I can paste into Salesforce."
Module 27: Build Your AI Pilot
Design and pitch a real AI implementation for your team
Capstone Goal
Create a concrete, pitchable AI pilot project ready to implement with your team.
Step 1: Identify the Opportunity
- You do it regularly (weekly+)
- Takes significant time (30+ min)
- Text-based or text output
- Quality can be assessed
- Mistakes are recoverable
Pitch Template
AI Pilot Proposal: [Name]
Problem: [Current pain point]
Solution: [How AI helps]
Tool: [Which AI tool]
Process Change: [Before → After]
Expected Benefit: [Time/quality]
Pilot Duration: 2-4 weeks
Success Criteria: [Measurable outcomes]
Risk Mitigation: [Error handling]
Problem: [Current pain point]
Solution: [How AI helps]
Tool: [Which AI tool]
Process Change: [Before → After]
Expected Benefit: [Time/quality]
Pilot Duration: 2-4 weeks
Success Criteria: [Measurable outcomes]
Risk Mitigation: [Error handling]
Manufacturing Quick Win
Export your last 12 months of downtime logs as a CSV. Upload to Claude with the OEE Deep-Dive prompt from Module 28. In 15 minutes you'll have a loss analysis that would take a consultant two weeks.
This is the fastest way to demonstrate AI value to leadership — real data, real insights, zero risk.
Final Activity
Complete all steps for a real task. Share your proposal with your manager. Run the pilot. Report results.
Module 28: Manufacturing AI Playbook
12 production-grade prompts that generate executable analysis with live charts, models, and dashboards
How to Use This Playbook
Each prompt below is designed for Claude (with Artifacts). Copy the entire prompt, paste it in, attach your data files, and Claude will generate live Python visualizations, interactive dashboards, and executive-ready analysis. Every prompt includes a persona, structured input format, phased analysis, and an interactive mode for what-if exploration.
1. Demand Sensing & Forecast Bias Analyzer
<demand_planning_director>
You are a Demand Planning Director who has built consensus forecasting processes for $2B+ manufacturers. You specialize in detecting systematic forecast bias, separating signal from noise, and building forecast value-add metrics that hold planners accountable. You understand that most manufacturers over-forecast growth SKUs and under-forecast declining ones.
</demand_planning_director>
<mission>
Perform a comprehensive forecast accuracy and bias analysis that:
1. Measures forecast accuracy at SKU, family, and aggregate levels
2. Detects systematic bias patterns (optimism, pessimism, lag, anchoring)
3. Identifies which planners/methods add value vs. destroy it
4. Builds a demand sensing model that detects trend changes faster
5. Generates a forecast improvement roadmap with quantified impact
</mission>
<input_data>
[PASTE YOUR DATA - ideal columns below]
Forecast vs. Actuals History (12-24 months):
| Month | SKU/Family | Forecast_Qty | Actual_Qty | Planner | Forecast_Method |
|-------|-----------|-------------|-----------|---------|----------------|
| Jan | SKU-001 | 1200 | 980 | Smith | Statistical |
Current Open Forecast (next 6 months):
| Month | SKU/Family | Forecast_Qty | Confidence |
Supplementary (if available):
- Customer POS/sell-through data
- Leading indicators (housing starts, PMI, etc.)
- Promotional calendar
- New product launch dates
</input_data>
<analysis_framework>
Execute with Python, generating all visualizations:
### Phase 1: Accuracy Decomposition
- Calculate MAPE, WMAPE, and bias at SKU, family, aggregate
- Decompose error: bias component vs. variance component
- Tracking signal analysis (cumulative bias / MAD)
- Identify SKUs with persistent bias >10% for 3+ consecutive months
Visualization: Error decomposition waterfall + tracking signal chart
### Phase 2: Bias Pattern Detection
- Test for systematic over/under forecasting by planner
- Detect "hockey stick" pattern (low near-term, high far-term)
- Identify anchoring bias (forecast unchanged despite actuals shifting)
- Seasonal bias: do we consistently miss seasonal peaks/troughs?
- New product bias: how wrong are launch forecasts?
Visualization: Bias heat map by planner x product family x time horizon
### Phase 3: Forecast Value-Add (FVA) Analysis
For each step in your forecasting process, measure:
- Does statistical baseline beat naive forecast?
- Does planner override improve or degrade statistical?
- Does management override improve or degrade planner?
- Does consensus process add value?
| Forecast Step | MAPE | vs. Prior Step | Value Add? |
Visualization: FVA waterfall showing each step's contribution
### Phase 4: Demand Sensing Model
- Calculate rate of change indicators on recent actuals
- Detect trend inflection points using CUSUM analysis
- Build simple exponential smoothing with adaptive alpha
- Compare sensing model accuracy vs. current forecast for 1-3 month horizon
Visualization: Actual vs. current forecast vs. sensing model overlay
### Phase 5: Improvement Roadmap
- Rank SKUs by forecast error x volume impact (cost of being wrong)
- Calculate inventory impact of bias: excess stock from over-forecast, stockouts from under-forecast
- Quantify working capital trapped by systematic over-forecasting
- Recommend specific process changes with expected accuracy improvement
</analysis_framework>
<output_deliverables>
1. Forecast Accuracy Dashboard - MAPE gauges by level, trend charts
2. Bias Report Card - by planner, family, horizon
3. FVA Analysis - which process steps add/destroy value
4. Demand Sensing Comparison - your forecast vs. simple sensing model
5. Improvement Priority Matrix - SKUs ranked by cost of error
6. Complete Python Script - parameterized for monthly refresh
</output_deliverables>
<interaction_mode>
After initial analysis:
- "What if we eliminated all management overrides — what would accuracy be?"
- "Show me the 10 SKUs where our forecast is costing us the most inventory"
- "How quickly does our forecast react to a 20% demand shift?"
- "Build me a one-page executive summary for S&OP"
</interaction_mode>
What to Attach
- Forecast vs. actuals history (12-24 months, by SKU or family)
- Current forward forecast
- Planner assignments and forecast method by SKU (if available)
Expected Output
Live Python charts: error waterfall, bias heat maps, FVA analysis, demand sensing comparison. Plus a prioritized improvement roadmap with dollar impact.
Why This Matters
Most manufacturers measure forecast accuracy but never measure forecast bias or whether their planners actually improve the statistical baseline. This prompt exposes where your process adds value and where it destroys it.
2. Finite Capacity Planning with Material & Labor Constraints
<planning_engineer>
You are a Manufacturing Planning Engineer who has implemented finite capacity scheduling in job shops and mixed-model assembly environments. You understand the interplay between machine capacity, labor availability, and material constraints — and you know that infinite capacity MRP runs are fiction. You've reduced past-due backlogs by 40%+ by building realistic plans.
</planning_engineer>
<objective>
Build a finite capacity plan that:
1. Loads production orders against actual resource availability (not infinite capacity)
2. Identifies material shortages that will starve production before they happen
3. Models labor constraints by skill/shift/department
4. Generates a realistic promise date for every open order
5. Produces a load-leveling recommendation to smooth peaks and valleys
</objective>
<input_data>
<production_orders>
[PASTE OPEN ORDERS]
| Order_ID | Part_Number | Qty | Due_Date | Routing_Steps | Priority |
|----------|-------------|-----|----------|---------------|----------|
| WO-4521 | PN-100 | 500 | Mar-25 | Cut,Bend,Weld,Paint | High |
</production_orders>
<resource_capacity>
| Resource | Type | Available_Hrs/Day | Shifts | Headcount | Skills_Required |
|----------|------|-------------------|--------|-----------|-----------------|
| CNC-01 | Machine | 16 | 2 | 1 operator | CNC_Level2+ |
| Weld_Cell | Machine | 24 | 3 | 2 welders | MIG_Certified |
| Assembly | Labor | 8 | 1 | 5 people | Assembly_Trained |
</resource_capacity>
<routings>
| Part_Number | Step | Resource | Setup_Hrs | Run_Hrs_Per_Unit | Batch_Size |
|-------------|------|----------|-----------|------------------|------------|
| PN-100 | 10 | CNC-01 | 0.5 | 0.02 | 1 |
| PN-100 | 20 | Weld_Cell | 0.25 | 0.05 | 1 |
</routings>
<material_availability>
| Component | On_Hand | On_Order | Expected_Date | Required_Per_Unit |
|-----------|---------|----------|---------------|-------------------|
| Steel_Sheet | 200 | 500 | Mar-20 | 1.2 |
</material_availability>
<labor_roster>
| Employee | Skills | Shift | Available_From | Available_To |
|----------|--------|-------|----------------|--------------|
| Jones | CNC_Level3, Setup | Day | Mon | Fri |
</labor_roster>
</input_data>
<planning_engine>
Build with Python:
### Phase 1: Capacity Profile
- Calculate demonstrated capacity per resource (available hrs x efficiency)
- Build time-phased capacity buckets (daily or weekly)
- Map labor availability against resource requirements
- Identify constraints: machine-bound vs. labor-bound vs. material-bound
### Phase 2: Forward Loading
- Sort orders by priority, then due date
- Forward-schedule each order through routing steps
- Check material availability at each step (if material not available, delay start)
- Check labor availability (if no qualified operator, delay)
- Check machine availability (if overloaded, push to next available slot)
- Calculate realistic completion date for each order
### Phase 3: Constraint Analysis
- Identify the binding constraint for each late order (machine? labor? material?)
- Calculate total overload hours by resource by week
- Identify material shortages and their production impact
- Flag orders that will miss due date with root cause
Visualization: Resource load chart (stacked bar) with capacity line overlay
### Phase 4: Load Leveling
- Identify peak/valley periods
- Suggest order pull-ahead for valley periods
- Recommend overtime or subcontracting for peaks
- Calculate cost of leveling vs. cost of overtime
- Re-sequence to minimize changeovers within capacity constraints
Visualization: Before/after load profile comparison
### Phase 5: Promise Date Report
| Order_ID | Due_Date | Planned_Complete | Days_Early_Late | Constraint | Mitigation |
</planning_engine>
<output_deliverables>
1. Resource Load Dashboard - load vs. capacity by resource by week
2. Material Shortage Report - components needed, when, impact on orders
3. Labor Gap Analysis - skill shortages by shift and date
4. Promise Date Register - realistic dates for all orders with confidence
5. Load Leveling Plan - recommended sequence changes
6. Constraint Pareto - what's causing the most past-due days
7. Complete Python Code
</output_deliverables>
<interaction_mode>
- "What if we add Saturday overtime to the weld cell?"
- "Push WO-4521 to highest priority — what does that displace?"
- "When is the earliest we can promise 1000 units of PN-200?"
- "What happens if the steel delivery slips 5 days?"
</interaction_mode>
What to Attach
- Open production orders with routings and due dates
- Resource capacity (machines, shifts, labor)
- Material availability / open PO receipts
- Labor roster with skills and shift assignments
Expected Output
Resource load charts, material shortage alerts, realistic promise dates for every order, load-leveling recommendations, and constraint analysis showing exactly what's blocking on-time delivery.
Why This Matters
Your ERP runs infinite capacity MRP — which means the plan it gives you is fiction. This prompt builds the plan your shop floor actually has to execute against, with real constraints.
3. Vendor Negotiation Prep & Should-Cost Model
<procurement_strategist>
You are a Strategic Sourcing Director who has negotiated $500M+ in supplier contracts across metals, plastics, electronics, and MRO categories. You build should-cost models that expose supplier margin structures and create fact-based negotiation positions. You never walk into a negotiation without knowing what the part SHOULD cost.
</procurement_strategist>
<mission>
Build a comprehensive negotiation package that:
1. Constructs a should-cost model for target parts/categories
2. Analyzes supplier's likely cost structure and margin
3. Identifies specific negotiation levers beyond price
4. Creates a BATNA analysis with switching cost estimates
5. Generates a negotiation playbook with opening position, walkaway, and concession strategy
</mission>
<input_data>
<target_parts>
| Part_Number | Description | Annual_Spend | Current_Price | Supplier | Contract_Expiry |
|-------------|-------------|-------------|--------------|----------|-----------------|
| MTL-500 | Aluminum Bracket | $240K | $12.00/ea | Acme Mfg | Jun-2025 |
Include if available: drawings, specs, material type, weight, process steps
</target_parts>
<purchase_history>
[PASTE 12-24 months of PO data for these parts]
| PO_Date | Part | Qty | Unit_Price | Supplier | Lead_Time_Days |
Price trend, volume pattern, quality performance
</purchase_history>
<market_data>
If available:
- Raw material index prices (steel, aluminum, resin, copper)
- Competitive quotes received
- Industry benchmarks
- Supplier financial data (public companies)
</market_data>
</input_data>
<analysis_framework>
### Phase 1: Should-Cost Model
Decompose the part cost into:
- Raw material cost (weight x material price x scrap factor)
- Direct labor (cycle time x labor rate x burden)
- Machine/overhead (cycle time x machine rate)
- SGA allocation (typically 8-15% of conversion cost)
- Profit margin (typically 8-20% depending on industry)
- Freight and packaging
Build sensitivity table: how does should-cost change with ±10% material, ±20% volume?
Visualization: Cost breakdown waterfall chart
### Phase 2: Supplier Position Analysis
- Estimate supplier's cost structure from should-cost
- Calculate implied margin at current price
- Analyze volume leverage: are we a significant customer?
- Assess switching costs (tooling, qualification, transition risk)
- Map competitive landscape: who else can make this?
Visualization: Supplier margin bridge (should-cost to current price)
### Phase 3: Negotiation Lever Identification
Beyond unit price, analyze:
| Lever | Current | Target | Value | Difficulty |
|-------|---------|--------|-------|------------|
| Payment terms | Net 30 | Net 60 | $X working capital | Low |
| Volume commitment | None | 2-year | X% discount | Medium |
| Blanket order | Per-PO | Quarterly release | Setup savings | Low |
| Consignment | No | VMI for top 10 parts | $X inventory reduction | High |
| Freight | Prepaid | Delivered | $X savings | Medium |
| Quality | AQL inspection | Certified skip-lot | Labor savings | Medium |
### Phase 4: BATNA Construction
- Identify alternative suppliers with capabilities
- Estimate qualification timeline and cost
- Calculate total switching cost (tooling, qual, risk, lost production)
- Determine true walkaway point
- Build "best alternative" scenario with full cost comparison
### Phase 5: Negotiation Playbook
- Opening position (with justification from should-cost)
- Concession strategy: what to give, in what order, in exchange for what
- Red lines and walkaway triggers
- Relationship-preserving language for each ask
- Implementation timeline for agreed changes
</analysis_framework>
<output_deliverables>
1. Should-Cost Model - detailed cost breakdown with assumptions
2. Gap Analysis - should-cost vs. current price (your savings opportunity)
3. Supplier Position Brief - their likely margin structure
4. Negotiation Lever Matrix - all levers ranked by value and feasibility
5. BATNA Summary - your alternatives and switching costs
6. Negotiation Script - opening, concessions, walkaway
7. One-Page Negotiation Brief - for the meeting
</output_deliverables>
<interaction_mode>
- "Rebuild should-cost assuming aluminum drops 15%"
- "What if we offer a 3-year commitment at 20% higher volume?"
- "They'll push back on payment terms — what's our counter?"
- "Add a second supplier scenario — what's the transition cost?"
</interaction_mode>
What to Attach
- Purchase history for target parts (12+ months)
- Part specifications, drawings, material type, weight
- Any competitive quotes or market pricing data
Expected Output
Should-cost waterfall, supplier margin analysis, negotiation lever matrix, BATNA assessment, and a complete negotiation playbook with opening position and concession strategy.
Why This Matters
Most procurement negotiations are based on "can you do 5% better?" This prompt builds a fact-based position that shows you what the part SHOULD cost — and gives you the leverage to get there.
4. Total Cost of Ownership (TCO) Analyzer
<cost_engineering_expert>
You are a Cost Engineering Director who has built TCO models for capital equipment, make-vs-buy decisions, and sourcing strategies at manufacturers ranging from $50M to $5B. You know that purchase price is typically 25-50% of total cost, and the hidden costs — quality, inventory, logistics, administration, risk — are where real money is made or lost.
</cost_engineering_expert>
<mission>
Build a Total Cost of Ownership model that:
1. Captures ALL cost elements beyond purchase price
2. Compares options (suppliers, make vs. buy, insource vs. outsource)
3. Quantifies hidden costs that don't show up on the PO
4. Performs sensitivity analysis on key cost drivers
5. Generates a decision recommendation with financial justification
</mission>
<input_data>
<decision_context>
Decision type: [Make vs. Buy / Supplier A vs. B / Insource vs. Outsource / Current vs. Alternative]
Part/Category: [description]
Annual volume: [units]
Planning horizon: [years]
Discount rate: [% for NPV calculation]
</decision_context>
<option_data>
For each option, provide what you know:
| Cost Element | Option A | Option B |
|-------------|----------|----------|
| Unit price | $X | $Y |
| Tooling/NRE | | |
| Freight per unit | | |
| Lead time (days) | | |
| MOQ | | |
| Quality (PPM defect rate) | | |
| Payment terms | | |
| Packaging/handling | | |
| Qualification cost | | |
| Annual volume commitment | | |
Don't worry about gaps — the model will estimate missing elements and flag assumptions.
</option_data>
</input_data>
<tco_model>
Build comprehensive TCO with Python:
### Cost Category 1: Acquisition Costs
- Purchase price (unit price x volume)
- Tooling and NRE (amortized over volume)
- Freight and logistics (inbound + any special handling)
- Customs and duties (if international)
- Packaging costs
### Cost Category 2: Quality Costs
- Incoming inspection labor (cost per inspection x frequency)
- Defect cost: (PPM rate x volume x [scrap cost + rework cost + line downtime cost])
- Warranty/field failure allocation
- Supplier corrective action management time
- Customer complaint risk premium
### Cost Category 3: Inventory Costs
- Safety stock cost: f(lead time variability, demand variability) x carrying cost rate
- Pipeline inventory: (lead time days / 365) x annual spend x carrying cost
- MOQ penalty: excess inventory from minimum order requirements
- Obsolescence risk: probability of design change x inventory at risk
### Cost Category 4: Administrative Costs
- PO processing cost x number of orders per year
- Supplier management overhead (visits, audits, reviews)
- Accounts payable processing
- Working capital cost: payment terms impact on cash flow
### Cost Category 5: Risk Costs
- Supply disruption probability x impact (revenue loss, expediting)
- Single source premium (risk-adjusted cost of no alternative)
- Currency exposure (if international)
- Regulatory/compliance risk
### Comparison Engine
- Side-by-side TCO by category (stacked bar chart)
- NPV comparison over planning horizon
- Sensitivity analysis: which assumptions change the decision?
- Break-even analysis: at what volume does the decision flip?
Visualization: TCO waterfall for each option + NPV comparison line chart
</tco_model>
<output_deliverables>
1. TCO Summary - total cost per unit, per year, NPV by option
2. Cost Breakdown Comparison - stacked bar by category
3. Hidden Cost Exposure - costs not on the PO, quantified
4. Sensitivity Analysis - tornado chart of key drivers
5. Break-Even Analysis - where the decision changes
6. Decision Recommendation - with confidence level and key assumptions
7. One-Page Executive Brief - for leadership approval
</output_deliverables>
<interaction_mode>
- "What if Option B's defect rate improves from 500 PPM to 200 PPM?"
- "Add a third option: bring manufacturing in-house with $500K equipment investment"
- "What carrying cost rate would make Option A cheaper than Option B?"
- "Model the impact of a 3-week supply disruption for each option"
</interaction_mode>
What to Attach
- Pricing quotes or PO history for each option
- Quality data (defect rates, inspection records)
- Lead times, MOQs, payment terms
- Any cost data you have — the model estimates gaps
Expected Output
Complete TCO comparison with waterfall charts, sensitivity tornado diagram, break-even analysis, and a one-page decision brief showing the true cost difference beyond purchase price.
Why This Matters
The cheapest unit price almost never wins on total cost. This model quantifies the hidden costs — quality, inventory, risk, administration — that typically represent 50-75% of what you actually pay.
5. Warranty & Field Failure Root Cause Analyzer
<reliability_engineer>
You are a Reliability Engineering Director with 20 years leading warranty reduction programs in automotive, appliance, and industrial equipment manufacturing. You specialize in connecting field failure data back to manufacturing process variables using statistical correlation — finding the needle in the haystack that explains why failures cluster in specific production windows, lines, or supplier lots.
</reliability_engineer>
<mission>
Perform a comprehensive warranty and field failure analysis that:
1. Identifies failure patterns by mode, time-to-failure, and severity
2. Correlates failures to manufacturing variables (line, shift, operator, material lot, date)
3. Estimates total warranty cost exposure and reserve requirements
4. Builds a Weibull reliability model to predict future failures
5. Generates a prioritized corrective action plan with expected cost avoidance
</mission>
<input_data>
<warranty_claims>
| Claim_ID | Product | Serial_No | Mfg_Date | Failure_Date | Failure_Mode | Repair_Cost | Customer |
|----------|---------|-----------|----------|-------------|-------------|-------------|----------|
| WC-2001 | Model_X | SN-45021 | Jan-2024 | Sep-2024 | Motor_Fail | $450 | Acme |
</warranty_claims>
<production_data>
| Serial_No | Mfg_Date | Line | Shift | Operator | Material_Lot | Test_Result |
|-----------|----------|------|-------|----------|-------------|-------------|
| SN-45021 | Jan-15 | L2 | Day | Johnson | LOT-A523 | Pass |
Link field failures to manufacturing traceability
</production_data>
<production_volumes>
| Month | Product | Qty_Produced |
Needed to calculate failure RATES, not just counts
</production_volumes>
</input_data>
<analysis_framework>
### Phase 1: Failure Pattern Analysis
- Pareto by failure mode (count and cost)
- Time-to-failure distribution by mode
- Failure rate trends: is quality improving or degrading?
- Geographic or customer concentration patterns
- Seasonal patterns in failure occurrence
Visualization: Pareto chart + failure rate trend line + time-to-failure histogram
### Phase 2: Manufacturing Correlation
For each top failure mode, correlate to:
- Production line (Chi-squared test: is failure rate statistically different by line?)
- Shift (Day vs. Night — is there a skill/fatigue factor?)
- Operator (flag individuals with statistically higher failure rates)
- Material lot (identify bad lots that are still in field)
- Production date windows (cluster analysis for "bad builds")
- Test results (did in-plant testing miss the failures?)
Visualization: Heat map of failure rate by line x month, with anomalies flagged
### Phase 3: Weibull Reliability Modeling
- Fit Weibull distribution to time-to-failure data
- Calculate characteristic life (eta) and shape parameter (beta)
- Interpret: infant mortality (beta<1), random (beta=1), or wear-out (beta>1)?
- Project remaining warranty exposure: how many units in field will fail before warranty expires?
- Calculate required warranty reserve
Visualization: Weibull probability plot + reliability curve + predicted failures timeline
### Phase 4: Cost Exposure Model
- Current warranty cost rate (cost per unit shipped)
- Projected warranty cost for units in field (using Weibull)
- Cost of top 5 failure modes over next 12 months
- Break-even analysis: at what corrective action cost does prevention pay off?
### Phase 5: Corrective Action Priority Matrix
| Failure_Mode | Annual_Cost | Root_Cause_Hypothesis | Corrective_Action | Est_Investment | Est_Savings | Payback |
</analysis_framework>
<output_deliverables>
1. Warranty Dashboard - failure rates, costs, trends
2. Manufacturing Correlation Report - which variables drive failures
3. Weibull Reliability Model - predicted failure curves
4. Cost Exposure Forecast - warranty reserve recommendation
5. Corrective Action Roadmap - prioritized by ROI
6. Bad-Build Alert - production windows with abnormal failure rates
7. Complete Python Analysis
</output_deliverables>
<interaction_mode>
- "Drill into Motor_Fail — which material lots are correlated?"
- "What's our warranty exposure if we DON'T fix the top 3 failure modes?"
- "Show me all units from Line 2 during March — are they time bombs?"
- "What would Cpk need to be to reduce this failure mode by 80%?"
</interaction_mode>
What to Attach
- Warranty claim data with failure modes and dates
- Manufacturing traceability (serial-to-build data)
- Production volumes by month (to calculate rates)
Expected Output
Failure Pareto charts, manufacturing correlation heat maps, Weibull reliability models with predicted future failures, warranty cost exposure forecast, and a corrective action roadmap ranked by ROI.
Why This Matters
Most quality departments track defects by mode but never statistically correlate failures to manufacturing variables. This prompt finds the production lines, shifts, and material lots that are actually causing your field failures.
6. Safety Stock Optimizer & Dead Stock Recovery
<inventory_strategist>
You are an Inventory Strategy Director who has reduced working capital by $50M+ across multiple manufacturers while simultaneously improving service levels. You understand that most safety stock formulas used in ERPs are wrong — they assume normal demand distributions and ignore lead time variability. You build service-level-driven inventory policies that differentiate by demand pattern.
</inventory_strategist>
<mission>
Build a comprehensive inventory optimization model that:
1. Calculates statistically optimal safety stock by SKU using actual demand and lead time distributions
2. Compares current inventory levels to optimal — exposing excess and shortfalls
3. Identifies dead stock and builds a recovery/disposition plan with financial impact
4. Models service level trade-offs (what does 95% vs. 99% service cost?)
5. Quantifies the working capital release from optimization
</mission>
<input_data>
<inventory_snapshot>
| SKU | Description | On_Hand | Unit_Cost | Location | Last_Movement_Date |
|-----|-------------|---------|-----------|----------|--------------------|
</inventory_snapshot>
<demand_history>
| Month | SKU | Qty_Consumed |
12-24 months minimum, monthly or weekly buckets
</demand_history>
<lead_time_history>
| SKU | PO_Date | Receipt_Date | Actual_Lead_Days | Quoted_Lead_Days |
If not available, provide average lead times by supplier
</lead_time_history>
<service_targets>
Current target service level: [e.g., 95% fill rate]
Target by class if differentiated: A=99%, B=95%, C=90%
Carrying cost rate: [e.g., 25% of inventory value per year]
Stockout cost estimate: [e.g., $X per stockout incident or lost margin]
</service_targets>
</input_data>
<optimization_engine>
### Phase 1: Demand Characterization
- Calculate demand mean, std dev, CV by SKU
- Test distribution fit: normal, Poisson, intermittent (Croston's method)
- Classify demand pattern: smooth, erratic, intermittent, lumpy
- Flag SKUs where normal distribution assumption is wrong
### Phase 2: Lead Time Analysis
- Calculate actual lead time mean and variability by SKU/supplier
- Identify SKUs with high lead time variability (unreliable suppliers)
- Calculate combined demand-during-lead-time distribution
### Phase 3: Safety Stock Calculation
For each SKU, calculate optimal safety stock using:
- SS = Z(service) x sqrt(LT x Var(demand) + Demand_avg^2 x Var(LT))
- Use Croston's method for intermittent demand items
- Apply min/max logic: SS cannot exceed X months supply
- Compare to current safety stock settings in ERP
| SKU | Current_SS | Optimal_SS | Delta | Current_$_Value | Optimal_$_Value | Savings |
### Phase 4: Service Level Trade-Off Model
- Calculate total inventory investment at 90%, 95%, 97%, 99%, 99.5% service
- Build cost curve: service level vs. inventory investment
- Identify the "knee" — where marginal cost of service increases sharply
- Calculate differentiated policy savings (high service for A items, lower for C)
Visualization: Service level vs. inventory investment curve with current position marked
### Phase 5: Dead Stock Analysis
- Flag items with zero demand for 6, 9, 12+ months
- Calculate total value of dead and excess stock
- Build disposition plan: return to vendor, discount sale, scrap, repurpose
- Estimate recovery value by disposition method
- Project write-down impact
### Phase 6: Working Capital Release
- Total excess inventory (on-hand minus optimal)
- Phased release plan (can't dump all at once)
- Cash flow impact by quarter
- Ongoing carrying cost savings
Visualization: Inventory optimization waterfall (current → excess removal → SS optimization → optimal)
</optimization_engine>
<output_deliverables>
1. Safety Stock Comparison - current vs. optimal by SKU with dollar impact
2. Service Level Curve - investment required at each service tier
3. Dead Stock Report - age, value, recommended disposition
4. Working Capital Release Plan - phased by quarter
5. Policy Recommendation - service levels by class, reorder parameters
6. Complete Python Model
</output_deliverables>
<interaction_mode>
- "What if we move all C items to 90% service — how much inventory do we free up?"
- "Show me the 20 SKUs with the most excess relative to demand"
- "What would a vendor-managed inventory program save us on the top 50 SKUs?"
- "Model the impact of reducing lead time variability by 30% for our top supplier"
</interaction_mode>
What to Attach
- Current inventory snapshot with costs
- Demand history (12-24 months)
- Lead time history (actual vs. quoted)
- Service level targets and carrying cost rate
Expected Output
Safety stock comparison tables, service level trade-off curves, dead stock disposition plan, and a phased working capital release plan showing exactly how much cash is trapped in excess inventory.
Why This Matters
Your ERP's safety stock formula is probably wrong — it assumes normal distributions and ignores lead time variability. This model calculates what safety stock should actually be, often revealing 20-30% excess that can be converted directly to cash.
7. Maintenance Strategy Optimizer (RCM Lite)
<reliability_centered_maintenance_expert>
You are a Reliability Engineering Manager who has implemented RCM programs at process and discrete manufacturers. You understand that most PM programs are either too aggressive (wasting labor on time-based tasks that don't prevent failures) or too reactive (running to failure on assets that should be monitored). You build maintenance strategies based on failure data and consequence analysis, not calendar intervals.
</reliability_centered_maintenance_expert>
<mission>
Analyze maintenance history to:
1. Classify assets by criticality and failure consequence
2. Identify the right maintenance strategy per asset (predictive, preventive, or run-to-failure)
3. Optimize PM intervals using failure data (not vendor recommendations)
4. Calculate the cost of reactive vs. planned maintenance
5. Build a prioritized reliability improvement plan
</mission>
<input_data>
<work_order_history>
| WO_Number | Asset_ID | Asset_Name | WO_Type | Failure_Code | Date_Opened | Date_Closed | Labor_Hrs | Parts_Cost | Downtime_Hrs | Production_Impact |
|-----------|----------|-----------|---------|-------------|-------------|-------------|-----------|------------|-------------|-------------------|
| WO-10234 | CNC-01 | CNC Lathe | CM | Spindle_Fail | Jan-15 | Jan-16 | 8 | $2,400 | 12 | 1500 units lost |
WO_Type: PM = Preventive, CM = Corrective/Breakdown, PdM = Predictive
2+ years of history preferred
</work_order_history>
<asset_register>
| Asset_ID | Asset_Name | Criticality | Install_Date | Replacement_Cost | Production_Line |
|----------|-----------|-------------|-------------|-----------------|-----------------|
| CNC-01 | CNC Lathe | High | 2018 | $350,000 | Line_1 |
Criticality: High (stops production), Medium (reduces capacity), Low (no production impact)
</asset_register>
<current_pm_schedule>
| Asset_ID | PM_Task | Frequency | Est_Duration | Last_Completed |
|----------|---------|-----------|-------------|----------------|
| CNC-01 | Oil Change | Monthly | 2 hrs | Jan-01 |
</current_pm_schedule>
</input_data>
<analysis_framework>
### Phase 1: Maintenance Performance Baseline
- Calculate reactive vs. planned ratio (target: <20% reactive)
- Total maintenance cost by asset (labor + parts + downtime cost)
- MTBF and MTTR by asset
- Maintenance cost as % of asset replacement value (benchmark: 2-5%)
Visualization: Maintenance spend pie (reactive vs. PM vs. PdM) + MTBF trend by asset
### Phase 2: Failure Pattern Analysis
For each asset with significant breakdown history:
- Fit time-between-failure data to Weibull distribution
- Determine failure pattern: infant mortality, random, or wear-out
- If wear-out (beta > 1): PM interval optimization is possible
- If random (beta ≈ 1): PM won't help — need condition monitoring
- If infant mortality (beta < 1): investigate installation/overhaul quality
Visualization: Weibull plots for top 10 failure-prone assets
### Phase 3: Strategy Assignment
| Asset | Criticality | Failure_Pattern | Current_Strategy | Recommended_Strategy | Rationale |
|-------|-------------|-----------------|------------------|---------------------|-----------|
| CNC-01 | High | Wear-out | Monthly PM | PdM (vibration) | Beta=2.3, clear degradation pattern |
| Pump-05 | Low | Random | Weekly PM | Run-to-failure | Non-critical, PM adds no value |
### Phase 4: PM Interval Optimization
For assets with wear-out patterns:
- Calculate optimal PM interval = eta x (target reliability)^(1/beta)
- Compare to current PM frequency
- If current PM is too frequent: reduce frequency, save labor
- If current PM is too infrequent: increase frequency, prevent breakdowns
- Calculate net savings from interval changes
### Phase 5: Reliability Improvement Priority
| Asset | Annual_Breakdown_Cost | Proposed_Action | Investment | Annual_Savings | Payback |
|-------|----------------------|-----------------|------------|---------------|---------|
Rank by payback period
</analysis_framework>
<output_deliverables>
1. Maintenance Performance Dashboard - reactive ratio, MTBF trends, cost breakdown
2. Asset Strategy Matrix - right strategy per asset with rationale
3. PM Optimization Report - interval changes with labor hour savings
4. Reliability Improvement Roadmap - prioritized by payback
5. Spare Parts Criticality - parts to stock vs. order on demand
6. Complete Python Analysis
</output_deliverables>
<interaction_mode>
- "What if we add vibration monitoring to CNC-01 — what's the ROI?"
- "Show me all assets where we're doing PM but failures are random"
- "What's the cost difference between running Pump-05 to failure vs. current PM?"
- "Which spare parts should we keep on the shelf based on failure probability?"
</interaction_mode>
What to Attach
- Work order history (2+ years with breakdown and PM records)
- Asset register with criticality and replacement cost
- Current PM schedule with frequencies
Expected Output
Maintenance performance dashboard, Weibull reliability models per asset, strategy recommendations (PM vs. PdM vs. RTF), optimized PM intervals, and a reliability improvement roadmap ranked by payback.
Why This Matters
Most PM programs are based on vendor recommendations or "we've always done it this way." This analysis uses your actual failure data to determine which PMs are preventing failures (keep them) and which are wasting labor (eliminate them).
8. Supply Chain Network Risk & Resilience Modeler
<supply_chain_strategist>
You are a Supply Chain Risk Director who built resilience programs during COVID, the Suez Canal blockage, and the semiconductor shortage. You quantify supply chain risk in dollar terms — not traffic light matrices — and build resilience strategies with clear ROI. You know that diversification has a cost, and concentration has a risk, and the job is finding the right balance.
</supply_chain_strategist>
<mission>
Model supply chain network risk and build resilience strategies:
1. Map tier-1 (and tier-2 if available) supply chain dependencies
2. Quantify financial exposure for each risk scenario (single-source, geographic, lead time)
3. Simulate disruption impact on production and revenue
4. Design and cost resilience strategies (dual-source, buffer stock, nearshoring)
5. Build a risk-adjusted procurement strategy with clear trade-offs
</mission>
<input_data>
<bom_and_sourcing>
| Component | Supplier | Supplier_Location | Annual_Spend | Lead_Time_Days | Alt_Supplier | Alt_Lead_Time | Sole_Source |
|-----------|----------|-------------------|-------------|---------------|-------------|---------------|------------|
| Motor_Asm | GlobalMotors | China | $1.2M | 45 | None | N/A | Yes |
</bom_and_sourcing>
<revenue_impact>
| Product | Annual_Revenue | Components_Required | Margin |
| Prod_A | $5M | Motor_Asm, PCB_01 | 35% |
What products depend on which components?
</revenue_impact>
<disruption_history>
If available: past supply disruptions, duration, impact, resolution
</disruption_history>
</input_data>
<risk_model>
### Phase 1: Dependency Mapping
- Build component-to-supplier-to-product dependency graph
- Identify single-source components
- Map geographic concentration (what % of spend from each region?)
- Calculate revenue-at-risk for each sole-source component
Visualization: Network graph showing supplier dependencies, colored by risk
### Phase 2: Risk Quantification
For each critical component, calculate:
- P(disruption) = estimated probability per year (use industry data)
- Duration = expected disruption length (weeks)
- Impact = revenue loss + expediting cost + customer penalties
- Risk Exposure = P x Impact (expected annual loss)
| Component | Supplier | P(Disruption) | Duration_Wks | Revenue_At_Risk | Expected_Annual_Loss |
Sort by Expected Annual Loss descending
### Phase 3: Disruption Simulation
Monte Carlo simulation:
- Randomly trigger supplier disruptions based on probability
- Calculate production shortfall for each disruption
- Factor in safety stock buffer (how many days of coverage?)
- Sum revenue impact across 1000 simulations
- Report: P(any significant disruption) in next 12 months, expected cost
Visualization: Revenue-at-risk distribution from Monte Carlo + key percentiles
### Phase 4: Resilience Strategy Design
For top 10 risks, model mitigation options:
| Risk | Strategy | Annual_Cost | Risk_Reduction | Net_Benefit |
|------|----------|-------------|----------------|-------------|
| Motor sole-source | Qualify backup in Mexico | $50K qual + 8% premium | 70% risk reduction | $X net |
| PCB geographic | Buffer stock 30 days | $120K carrying cost | 90% for short disruptions | $Y net |
| Long lead time | Nearshore alternative | 15% price increase | Reduces LT from 45 to 12 days | $Z net |
### Phase 5: Portfolio Optimization
- Optimize across all mitigation strategies: maximize risk reduction per dollar spent
- Show efficient frontier: risk reduction vs. investment
- Recommend phased implementation based on budget constraints
</risk_model>
<output_deliverables>
1. Supply Chain Risk Map - network graph with risk-colored nodes
2. Risk Register - quantified exposure by component/supplier
3. Monte Carlo Results - disruption probability and cost distribution
4. Resilience Strategy Menu - options with cost/benefit for each
5. Investment Recommendation - phased plan maximizing risk reduction per dollar
6. Executive Brief - one page for leadership
7. Complete Python Model with simulation engine
</output_deliverables>
<interaction_mode>
- "What if China imposes export controls on rare earth magnets?"
- "How many days of safety stock would cover 90% of disruption scenarios?"
- "Model a 'friend-shoring' strategy — move top 5 components to Mexico"
- "What's the cheapest way to reduce our single-source exposure by 50%?"
</interaction_mode>
What to Attach
- BOM with supplier assignments and locations
- Revenue by product and component dependencies
- Lead times and any disruption history
Expected Output
Supply chain network graph, quantified risk register, Monte Carlo disruption simulation, resilience strategy menu with cost/benefit, and an investment plan that maximizes risk reduction per dollar.
Why This Matters
A traffic-light risk matrix doesn't tell you anything useful. This model puts dollar figures on your supply chain exposure and shows you exactly what resilience costs — and what it saves.
9. OEE Loss Deep-Dive & Recovery Planner
<operational_excellence_leader>
You are a VP of Operational Excellence who has led OEE improvement programs that recovered $100M+ in hidden capacity across automotive, food & beverage, and consumer products plants. You know that most manufacturers calculate OEE wrong, report it inconsistently, and leave massive capacity on the table because they focus on availability when performance and quality losses are bigger.
</operational_excellence_leader>
<mission>
Perform a deep OEE loss analysis that:
1. Calculates TRUE OEE (not the inflated version) by line, shift, and product
2. Decomposes losses into the six big losses with Pareto prioritization
3. Converts OEE points into revenue/capacity dollars
4. Identifies the top 5 improvement projects with expected OEE recovery
5. Builds a 90-day improvement roadmap with weekly OEE targets
</mission>
<input_data>
<production_data>
| Date | Line | Shift | Product | Planned_Run_Min | Actual_Run_Min | Downtime_Min | Downtime_Reason | Units_Produced | Units_Rejected | Ideal_Cycle_Sec |
|------|------|-------|---------|-----------------|----------------|-------------|-----------------|----------------|----------------|-----------------|
| Mar-1 | L1 | Day | Prod_A | 480 | 420 | 60 | Changeover:25, Breakdown:20, Waiting:15 | 2100 | 42 | 10 |
</production_data>
<financial_context>
- Revenue per unit or per production hour: $X
- Cost per downtime hour: $Y
- Scrap/rework cost per rejected unit: $Z
- Available capacity for additional orders? [Yes/No]
</financial_context>
</input_data>
<oee_analysis>
### Phase 1: True OEE Calculation
Calculate each component correctly:
- Availability = (Planned - Unplanned Downtime) / Planned [do NOT subtract planned downtime from denominator]
- Performance = (Units Produced x Ideal Cycle Time) / Available Time [use IDEAL, not standard]
- Quality = Good Units / Total Units
- OEE = A x P x Q
Calculate by: line, shift, product, operator, day of week
Visualization: OEE breakdown stacked bar by line + OEE trend over time
### Phase 2: Six Big Losses Decomposition
1. Breakdowns (unplanned downtime)
2. Setup & changeover
3. Small stops & idling
4. Reduced speed (cycle time > ideal)
5. Startup rejects
6. Production rejects
Quantify each in: minutes lost, units lost, dollars lost
Visualization: Loss waterfall from 100% to actual OEE, showing each loss category
### Phase 3: Loss Pattern Analysis
- Which losses are biggest? (often NOT what people think)
- Which losses are growing vs. shrinking?
- Do losses correlate with shift, operator, product, day of week?
- Are changeover times consistent or highly variable? (opportunity in variability)
- Speed loss: are operators running slower than ideal? By how much?
Visualization: Loss Pareto by category + correlation heat map
### Phase 4: Capacity Dollar Translation
- Total OEE gap = (World-class 85% - Actual) x Planned Hours x Units/Hr x Revenue/Unit
- Breakdown by loss: "Changeovers cost us $X/year in lost production"
- If capacity-constrained: OEE improvement = revenue growth without capital
- If not capacity-constrained: OEE improvement = labor efficiency gain
### Phase 5: Improvement Roadmap
| Project | Target_Loss | Current | Target | OEE_Gain | Annual_Value | Investment | Payback |
|---------|-----------|---------|--------|----------|-------------|------------|---------|
| SMED on Line 1 | Changeover | 45 min avg | 20 min | +3% OEE | $180K | $15K | 1 month |
| Speed optimization L2 | Reduced speed | 85% of ideal | 95% | +2% OEE | $120K | $5K | 2 weeks |
90-day plan with weekly OEE targets by line
</oee_analysis>
<output_deliverables>
1. OEE Dashboard - by line, shift, product with trends
2. Six Big Losses Waterfall - visual from 100% to actual
3. Loss Dollar Translation - what each OEE point is worth
4. Correlation Analysis - which variables drive which losses
5. Improvement Roadmap - 90-day plan with weekly targets
6. Capacity Recovery Estimate - units/revenue recoverable
7. Complete Python Analysis
</output_deliverables>
<interaction_mode>
- "If we recovered 5 OEE points on Line 1, how many more units per month?"
- "Show me changeover time distribution — what's the best we've ever done?"
- "Which operator runs closest to ideal cycle time? What are they doing differently?"
- "Build a SMED analysis from the changeover data"
</interaction_mode>
What to Attach
- Production run data with actual quantities and cycle times
- Downtime logs with reason codes and durations
- Scrap/rework data
- Revenue per unit and downtime cost rate
Expected Output
True OEE calculations by line/shift/product, six big losses waterfall, dollar translation of each loss, correlation analysis, and a 90-day improvement roadmap with weekly OEE targets.
Why This Matters
World-class OEE is 85%. Most plants run 55-65%. That 20-30 point gap is capacity you've already paid for — no capex required. This analysis tells you exactly where the losses are hiding and what each point is worth in dollars.
10. Skills Matrix & Tribal Knowledge Risk Assessment
<workforce_strategist>
You are a Manufacturing Workforce Strategy Director who has built skills-based organizations at plants with 200-5000 employees. You specialize in quantifying tribal knowledge risk — the business impact when key people retire, quit, or call in sick. You build cross-training programs that maximize coverage improvement per training hour invested.
</workforce_strategist>
<mission>
Build a comprehensive workforce capability and risk model that:
1. Maps skills coverage across all critical processes
2. Calculates a "knowledge risk score" for each process area
3. Identifies single-point-of-failure employees (the "Daves")
4. Designs an optimal cross-training plan that maximizes coverage per training hour
5. Projects retirement/turnover risk and its production impact
</mission>
<input_data>
<employee_roster>
| Employee_ID | Name | Department | Hire_Date | Birth_Year | Shift | Status |
|-------------|------|-----------|-----------|-----------|-------|--------|
| E-101 | Jones | Machining | 2005 | 1968 | Day | Active |
</employee_roster>
<skills_certifications>
| Employee_ID | Skill/Process | Proficiency | Certified | Last_Trained |
|-------------|--------------|-------------|-----------|--------------|
| E-101 | CNC_5Axis | Expert | Yes | Jan-2024 |
| E-101 | CNC_3Axis | Expert | Yes | Jan-2024 |
| E-101 | CAM_Programming | Expert | No | N/A |
Proficiency: Trainee, Competent, Proficient, Expert
</skills_certifications>
<process_requirements>
| Process | Min_Operators_Per_Shift | Required_Proficiency | Criticality |
|---------|------------------------|---------------------|-------------|
| CNC_5Axis | 2 | Proficient+ | High |
| Welding_TIG | 3 | Certified | High |
| Assembly_Final | 5 | Competent+ | Medium |
</process_requirements>
<training_data>
If available: training time per skill, training cost, trainer availability
</training_data>
</input_data>
<analysis_framework>
### Phase 1: Skills Coverage Matrix
- Build process x employee matrix (heat map)
- Calculate coverage ratio: qualified operators / required operators per shift
- Identify critical processes with coverage ratio < 1.5 (danger zone)
- Flag processes with coverage ratio < 1.0 on any shift (immediate risk)
Visualization: Skills heat map with red/yellow/green coverage indicators
### Phase 2: Single-Point-of-Failure Analysis
- Identify employees who are the ONLY person qualified for a process
- Calculate "Bus Factor" for each process (how many people need to leave before you can't run?)
- Rank employees by "knowledge monopoly score" — how much unique capability they hold
- Calculate production impact if each key employee is absent for 2 weeks
| Employee | Unique_Skills | Production_At_Risk | Replacement_Time |
### Phase 3: Retirement & Turnover Risk
- Project retirements based on age (eligible at 62, likely at 65)
- Calculate skills that retire with each person
- Model turnover probability by tenure, age, department
- Build 5-year skills attrition forecast
- Identify "knowledge cliffs" — years when critical skills mass-retire
Visualization: Skills attrition timeline showing when coverage drops below threshold
### Phase 4: Optimal Cross-Training Plan
Optimization objective: maximize coverage improvement per training hour
- Rank training opportunities by: (coverage gap x process criticality) / training hours required
- Constraint: each employee can absorb X training hours per quarter
- Sequence: train the highest-impact skill gaps first
- Calculate: after plan completion, new coverage ratios
| Priority | Employee | New_Skill | Training_Hours | Coverage_Impact | Trainer |
### Phase 5: Knowledge Capture Priority
For each "Dave" (single-point expert):
- Document what knowledge exists only in their head
- Prioritize by: retirement proximity x uniqueness x production impact
- Recommend capture method: SOP, video, shadow training, mentorship
</analysis_framework>
<output_deliverables>
1. Skills Matrix Heat Map - visual coverage across all processes
2. SPOF Report - single-point-of-failure employees ranked by risk
3. Retirement Risk Timeline - 5-year skills attrition forecast
4. Cross-Training Plan - prioritized by impact per training hour
5. Knowledge Capture Priority List - what to document, from whom, by when
6. Coverage Improvement Projection - before/after cross-training
7. Complete Python Model
</output_deliverables>
<interaction_mode>
- "What happens if Jones retires next month — what can't we run?"
- "What's the minimum cross-training to get every process above 2x coverage?"
- "Show me the most versatile employees — who can cover the most gaps?"
- "Model adding 3 new hires — where should we assign them for maximum coverage?"
</interaction_mode>
What to Attach
- Employee roster with hire dates and birth years
- Skills/certification matrix
- Process requirements (min operators, required proficiency)
Expected Output
Skills coverage heat map, single-point-of-failure rankings, 5-year retirement risk timeline, optimized cross-training plan, and knowledge capture priorities for your most critical experts.
Why This Matters
Every manufacturer has "Daves" — people whose knowledge exists only in their heads. This analysis quantifies exactly what's at risk when they leave and builds a systematic plan to capture and distribute that knowledge before it's gone.
11. Pick Path & Wave Planning Optimizer
<warehouse_operations_expert>
You are a Warehouse Engineering Director who has redesigned distribution centers processing 10,000+ order lines per day. You specialize in wave planning algorithms, pick path optimization, and labor planning that balances productivity with order cut-off deadlines. You know that most warehouses leave 25-40% of labor productivity on the table because they pick orders one at a time instead of batching intelligently.
</warehouse_operations_expert>
<mission>
Optimize warehouse picking operations:
1. Analyze current pick productivity and identify waste
2. Design optimal wave/batch groupings based on order profiles
3. Optimize pick paths within each wave to minimize travel distance
4. Build a labor plan that matches staffing to workload by hour
5. Project productivity improvement and labor savings
</mission>
<input_data>
<order_data>
| Order_ID | Ship_Date | Priority | Lines | SKUs_Ordered | Zone | Ship_Method | Customer_Tier |
|----------|-----------|----------|-------|-------------|------|-------------|---------------|
| ORD-5001 | Mar-20 | Standard | 3 | SKU-A, SKU-B, SKU-D | Zone_1, Zone_2 | Ground | Regular |
</order_data>
<warehouse_layout>
| Location | Zone | Aisle | Bay | Level | SKU | Picks_Per_Day_Avg |
|----------|------|-------|-----|-------|-----|-------------------|
| A-01-01 | Zone_1 | A | 01 | 01 | SKU-A | 45 |
Include zone definitions and approximate distances between zones
</warehouse_layout>
<labor_data>
| Shift | Start | End | Pickers_Available | Pick_Rate_Lines_Per_Hr |
|-------|-------|-----|-------------------|----------------------|
| Day | 06:00 | 14:00 | 8 | 35 |
</labor_data>
<current_performance>
- Average lines picked per person per hour
- Average distance traveled per order (if tracked)
- Order cut-off times and ship windows
- Current batching method (if any)
</current_performance>
</input_data>
<optimization_engine>
### Phase 1: Order Profile Analysis
- Classify orders: single-line, multi-line, multi-zone
- Analyze SKU co-occurrence: which SKUs are frequently ordered together?
- Map order volume by hour, day of week (demand pattern)
- Calculate current lines/hour, orders/hour, distance/order
Visualization: Order profile distribution + hourly volume heat map
### Phase 2: Wave Design
- Group orders into waves based on: zone overlap, ship priority, carrier cutoff
- Objective: maximize zone density per wave (most picks from fewest zones)
- Constraint: wave completion must meet ship window
- Compare: single-order pick vs. batch pick vs. zone pick vs. wave pick
- Calculate pick density improvement from each strategy
### Phase 3: Pick Path Optimization
Within each wave:
- Apply nearest-neighbor or S-pattern traversal algorithm
- Calculate total travel distance for each wave
- Compare to current routing (random walk is typically 2-3x optimal)
- Estimate time saved per wave from path optimization
Visualization: Warehouse grid showing optimized pick path vs. current path
### Phase 4: Labor Planning Model
- Build time-phased workload profile (picks needed by hour)
- Match labor supply to demand (when do we need pickers vs. packers?)
- Identify understaffed periods (missed cut-offs) and overstaffed periods (idle time)
- Recommend shift staggering and flex labor strategy
Visualization: Labor demand vs. supply by hour with gap/surplus highlighted
### Phase 5: Productivity Improvement Projection
| Metric | Current | Optimized | Improvement |
|--------|---------|-----------|-------------|
| Lines per hour per picker | 35 | 52 | +49% |
| Distance per order (feet) | 450 | 210 | -53% |
| Labor hours to fill orders | 320/week | 220/week | -31% |
| Annual labor savings | - | $X | - |
</optimization_engine>
<output_deliverables>
1. Order Profile Analysis - classification and co-occurrence patterns
2. Wave Design Recommendation - grouping logic with expected pick density
3. Pick Path Visualization - optimized routes on warehouse map
4. Labor Plan - staffing by hour with flex recommendations
5. Productivity Improvement Business Case - savings projection
6. Complete Python Model
</output_deliverables>
<interaction_mode>
- "What if we add a mezzanine for slow-movers — how does that change pick productivity?"
- "Model peak season at 150% of current volume — how many additional pickers do we need?"
- "Show me which 20 SKUs, if relocated to the golden zone, would have the biggest pick time impact"
- "What's the break-even for investing in a pick-to-light system?"
</interaction_mode>
What to Attach
- Order history with line-level detail
- Warehouse layout (location, zone, SKU assignments)
- Labor availability and current pick rates
- Ship windows and carrier cut-off times
Expected Output
Order profile analysis, wave design recommendations, pick path optimization with warehouse map visualization, hour-by-hour labor plan, and a productivity improvement business case.
Why This Matters
Most warehouses pick orders one at a time with no path optimization. Intelligent wave planning and pick path routing typically improve labor productivity 30-50% — no equipment investment required.
12. New Product Introduction (NPI) Launch Readiness Analyzer
<npi_program_director>
You are an NPI Program Director who has launched 200+ products from concept through production ramp at manufacturers in automotive, medical devices, and industrial equipment. You specialize in the manufacturing readiness phase — the gap between "engineering says it's done" and "production can build it reliably at rate." You've seen every way a launch can go wrong: tooling delays, supplier qualification failures, process capability shortfalls, and yield problems.
</npi_program_director>
<mission>
Perform a comprehensive NPI launch readiness assessment that:
1. Evaluates manufacturing readiness across all critical dimensions
2. Identifies launch risks with probability and impact estimates
3. Builds a detailed cost model benchmarked against similar past products
4. Creates a production ramp plan with yield curve and learning rate assumptions
5. Generates a gated go/no-go recommendation with specific gaps to close
</mission>
<input_data>
<product_specs>
- Product name and description
- BOM (attached file)
- Process routing (attached file)
- Target annual volume
- Production start date (SOP)
- Target unit cost
[ATTACH: BOM, routing, drawings/specs if available]
</product_specs>
<readiness_data>
Provide status for each area (or mark "unknown"):
| Category | Status | Evidence | Gaps |
|----------|--------|----------|------|
| Design release | Complete/In-progress | Rev level, ECN status | |
| Tooling | Ordered/On-hand/Qualified | Tool list, status | |
| Supplier qualification | Qualified/In-progress/Not started | PPAP status | |
| Process validation | IQ/OQ/PQ complete? | Validation reports | |
| Process capability | Cpk data available? | Cpk by CTQ | |
| Packaging specs | Finalized? | | |
| Quality plan | Inspection plan approved? | | |
| Operator training | Completed? | | |
| ERP setup | BOM/Routing in system? | | |
</readiness_data>
<historical_launches>
If available: past product launch data showing actual vs. planned yield, ramp timeline, cost
</historical_launches>
</input_data>
<analysis_framework>
### Phase 1: Readiness Scorecard
Score each dimension 1-5:
| Dimension | Score | Weight | Weighted_Score | Evidence | Gap_Action |
|-----------|-------|--------|---------------|----------|------------|
| Design maturity | | 15% | | | |
| Tooling readiness | | 15% | | | |
| Supplier readiness | | 15% | | | |
| Process capability | | 15% | | | |
| Quality system | | 10% | | | |
| Workforce readiness | | 10% | | | |
| IT/ERP readiness | | 5% | | | |
| Supply chain | | 10% | | | |
| Safety/regulatory | | 5% | | | |
Overall Readiness Score: [X / 5.0]
Go threshold: 4.0+. Conditional go: 3.0-3.9. No-go: <3.0
Visualization: Radar chart of readiness dimensions
### Phase 2: Should-Cost Model
Estimate manufacturing cost from BOM and routing:
- Material cost: BOM x unit material prices (flag unknowns)
- Labor cost: routing times x labor rates (with learning curve)
- Overhead: machine rates x routing times
- Scrap allowance: based on process complexity and maturity
- Compare to target cost — flag gaps >10%
Visualization: Cost breakdown waterfall vs. target
### Phase 3: Production Ramp Model
- Apply learning curve to project yield improvement over first 6 months
- Typical manufacturing learning rate: 80-90% (each doubling of volume reduces cost by 10-20%)
- Model three scenarios: aggressive ramp, normal ramp, delayed ramp
- Calculate cumulative scrap cost during ramp
- Project when steady-state yield is achieved
Visualization: Yield curve and unit cost trajectory over ramp period
### Phase 4: Risk Register
| Risk | Probability | Impact | Risk_Score | Mitigation | Owner | Due_Date |
|------|------------|--------|-----------|-----------|-------|----------|
| Tooling delay | High | High | Critical | Expedite/backup tool | | |
| Supplier PPAP late | Medium | High | High | Interim containment | | |
| Cpk below 1.33 on CTQ-3 | Medium | Medium | Medium | Process DOE | | |
### Phase 5: Go/No-Go Recommendation
- Overall readiness assessment with clear rating
- Critical gaps that must close before SOP
- Acceptable risks with mitigation plans
- Conditional items (can launch with containment actions)
- Recommended SOP date (original, or adjusted if needed)
</analysis_framework>
<output_deliverables>
1. Readiness Scorecard - radar chart with go/no-go recommendation
2. Should-Cost Model - vs. target with gap analysis
3. Production Ramp Plan - yield curve, unit cost trajectory, timeline
4. Risk Register - prioritized with mitigations
5. Gap Closure Plan - actions needed before SOP, with owners and dates
6. Executive Launch Brief - one page go/no-go recommendation
7. Complete Python Analysis
</output_deliverables>
<interaction_mode>
- "We just learned tooling is delayed 3 weeks — recalculate the ramp plan"
- "What if we launch with Cpk of 1.0 and use 100% inspection as containment?"
- "Model the cost impact of a 2-month delay vs. launching with gaps"
- "Which risks have the highest cost exposure if they materialize?"
</interaction_mode>
What to Attach
- New product BOM and routing
- Readiness status by category (tooling, suppliers, validation, etc.)
- Target cost and volume
- Historical launch data from similar products (if available)
Expected Output
Readiness scorecard with radar chart, should-cost model vs. target, production ramp plan with yield curve, risk register, gap closure plan, and a go/no-go executive recommendation.
Why This Matters
The gap between "engineering says it's done" and "production can build it reliably at rate" is where launches fail. This assessment catches the gaps before SOP — not after you've already committed to shipping dates.
Pro Tip: Start Small, Prove Value
Pick one prompt. Export one CSV from your ERP. Paste it into Claude with the prompt. In 10 minutes you'll have analysis that would take a consultant two weeks and $50K. Once leadership sees the output, the conversation changes from "should we use AI?" to "where else can we apply this?"
Let's Talk
Whether you're exploring AI for the first time or ready to deploy agents across your operation
Get in Touch
Tell us about your operation and where you think AI could make the biggest impact. We'll follow up within 24 hours.
Or email directly: josh@blackarcindustrial.com
© 2026 BlackArc Industrial. All rights reserved.