Company AI Day for Multi-Industry / Cross-Functional
There's no intentional AI use yet. Maybe one or two people have tried ChatGPT. The team needs language, safety, and first wins.
Pick a date 1-2 weeks out. Send the Pre-Work Survey to your team 3 days before. Print the handouts. Have someone read the Facilitator Script aloud through the day. Capture every workflow built in the Workflow Capture Worksheet. Hand the Leadership Recap to your CEO at the end. You're done.
Your Day at a Glance
Full one-day shape, calibrated for Stage 1 (Foundations + safe experimentation). Build labs are tuned to: Each person builds one prompt that saves them 30 minutes this week.
Why AI matters now, what we'll build today, ground rules.
Activity: CEO or lead reads the kickoff script. Q&A. Set the tone.
Use the Facilitator Script handout.
What AI is good at, where it fails, prompting patterns, safety.
Activity: Live demo: same prompt, two different framings, observe the difference.
Identify repetitive, high-friction work across the team.
Activity: Use the Use-Case Checklist. Score by impact, ease, risk. Pick 3-5 to build.
Use the Use-Case Checklist handout.
Each team builds their first workflow asset. (Foundations + safe experimentation)
Activity: Each person builds one prompt that saves them 30 minutes this week.
Use the Workflow Capture Worksheet.
Capture progress for the leadership recap. Demo a few wins.
Refine, document, and share. Build a second workflow if Lab 1 was quick. (Foundations + safe experimentation)
Activity: Each person builds one prompt that saves them 30 minutes this week.
Each group shares: problem, solution, time saved, what's next.
Activity: 5 min per group. Capture for leadership recap.
Synthesize quick wins, automation candidates, training gaps, governance needs.
Use the Leadership Recap Template.
Each person commits to one AI habit for the next 30 days.
Activity: Round-robin. Capture in the 30-Day Roadmap handout.
Use the 30-Day Roadmap handout.
The Curriculum
AI Day isn't just "try ChatGPT and see what happens." This is the actual curriculum Chelsea teaches — 10 modules covering everything from LLM fundamentals to agent design. Use it as your facilitator's reference. Pull modules into Foundations, Build Lab, or Advanced segments based on your team's stage.
1AI & LLM FundamentalsWhat AI actually is, what LLMs are, and the words you need to stop nodding through.30 minFoundation
Most teams use AI for months without ever defining what it is. That gap is why people overestimate it (expecting magic) and underestimate it (treating it as autocomplete). A 30-minute foundation closes the gap.
- AI / Machine Learning / LLM
- AI is the umbrella. ML is software that learns patterns from data. LLMs (Large Language Models) are a specific kind of ML trained on massive text to predict the next token. ChatGPT, Claude, Gemini are LLM-powered.
- Token
- The unit an LLM reads and writes. Roughly 1 token = 4 characters or 0.75 words. 'Hello world' is ~2 tokens. Pricing and context windows are measured in tokens.
- Context window
- How much text the model can hold at once (input + output). Modern models: 128K to 1M+ tokens. If your conversation gets too long, the model 'forgets' the beginning.
- Temperature
- Creativity dial (0 to 1). Low = deterministic, predictable, factual. High = varied, creative, weirder. Use low for compliance/legal, high for brainstorming.
- Hallucination
- When the model confidently states something untrue. Always possible. Mitigation: ground the model in real source documents and require citations.
- Prompt
- Your instruction to the model. The single biggest lever on output quality.
- System prompt
- The standing instruction that sets behavior across a whole conversation or project. Where you define role, voice, constraints, and rules.
- RAG (Retrieval-Augmented Generation)
- Letting the model search a private knowledge base before answering. How Claude Projects, Custom GPTs, and most internal AI tools work.
- Fine-tuning
- Training a base model on your specific data so it permanently learns your style. Powerful, expensive, rarely the right first move. RAG usually wins.
- Agent
- An AI that can take actions (search the web, use tools, write files, send emails) — not just answer questions. We cover this in Module 10.
Open Claude.ai or ChatGPT. Ask: 'Explain the difference between temperature 0 and temperature 1 using a haiku for each.' Then run it again. Notice what stays the same, what changes. That's tokens, temperature, and probabilistic output in one minute.
The model sounds confident even when wrong. Confidence is not a quality signal in LLM output.
2The LLM Landscape — Who's WhoClaude vs ChatGPT vs Gemini vs Perplexity vs everyone else. What each is good at. What's free.20 minFoundation
Picking the wrong model wastes time and money. Each model has a personality and a sweet spot. Knowing 'when to use what' saves hours per week.
- Frontier models
- The current top tier: Claude (Anthropic), GPT-4/5 (OpenAI), Gemini (Google). All are general-purpose, all are getting better monthly.
- Open-weights models
- Models you can download and run yourself: Llama (Meta), Mistral, Qwen, DeepSeek. Useful for privacy, cost, and customization.
- Search-augmented LLMs
- LLMs that browse the web for fresh info: Perplexity, Google AI Mode, ChatGPT search.
- Multimodal
- Models that handle text + images + audio + video. All frontier models are multimodal now.
Pick the same task — 'Summarize this 5-page PDF and pull out the action items.' Run it in Claude, ChatGPT, and Gemini. Compare. You'll feel the personality differences in 10 minutes.
3Prompting Patterns That Actually WorkStop typing 'write me an email.' Start writing prompts that produce usable first drafts.30 minFoundation
The single biggest lever on AI output quality is the prompt. Most people never improve past 'one-line request.' This 30 minutes pays back forever.
- Role + Goal + Voice + Constraints + Format
- The five-part recipe. Tell the model who it is, what it's doing, in whose voice, with what constraints, and in what format. Use this once and you'll never go back.
- Few-shot examples
- Showing 2-3 examples of input → output before asking the model to do it. The single most powerful upgrade to any prompt.
- Chain-of-thought
- Asking the model to think step by step. Improves reasoning on complex tasks. 'Walk through your reasoning before giving the answer.'
- Reflection / self-critique
- After the first answer, ask: 'Critique this. What's weak? Rewrite to fix.' Output quality jumps.
- Negative constraints
- What NOT to do. 'Do not use the word leverage. Do not use bullet points.' Often more useful than positive instructions.
Take your worst-quality AI output from this week. Rewrite the prompt with all five parts (role, goal, voice, constraints, format). Re-run. Compare. Save the new prompt to your prompt library.
4Safety, Security, and RiskWhat can go wrong. Where the real risks are. How to set up guardrails without killing the upside.30 minFoundation
Most AI 'risks' people worry about are theatrical. The real risks (data leakage, hallucinated facts in outbound work, bias, IP issues) are concrete and manageable — IF you name them.
- Data privacy
- Anything you paste into a consumer AI tool MAY be used to train future models (varies by provider). Use enterprise/Team plans for any client or sensitive data. Read the data-handling page of any tool you adopt.
- Hallucination in outbound work
- AI making up facts that go to customers, clients, or the public. Mitigation: humans review outbound for at least 90 days. Specific facts (dates, prices, names, citations) get verified.
- Shadow AI
- Employees using AI tools nobody at leadership knows about. Not bad — but it means risks aren't being managed. Surface it gently. Don't punish.
- Bias
- LLMs reflect the bias of their training data. Most visible in: hiring, performance reviews, customer-facing tone. Always have a human in the loop for people-impact decisions.
- Copyright / IP
- Images and text from generative AI live in murky legal territory. Don't use AI-generated assets in places where ownership matters (logos, trademarked content, contracts) without legal review.
- Audit trail
- For regulated work (legal, medical, financial), log what AI was used for, by whom, with what inputs. Most enterprise tools support this.
Draft a one-page AI usage policy for your team. Cover: (1) what data is OK / not OK to paste, (2) which tools are sanctioned, (3) what gets human review before going out, (4) who to ask when unsure. Distribute it. Update it quarterly.
Banning AI outright doesn't reduce risk — it pushes everyone to shadow tools where you have zero visibility. Sanction safe tools instead.
5No-Code AI Tools — Build Without an EngineerConnect AI to the rest of your stack — automations, dashboards, decks, sites — without writing code.45 minIntermediate
The leap from 'I use ChatGPT' to 'AI runs a workflow for me' usually happens through no-code tools. This is where time savings compound.
- Automation platforms
- Tools that connect apps and let an AI step run inside the chain. Trigger → AI does the thing → Output goes somewhere.
- Trigger / Action
- Trigger = the event that starts it (new email, new lead, new file). Action = what happens after (AI summarizes, then posts to Slack).
- Webhook
- A URL that listens for events. Your AI gets called when something happens elsewhere. The glue between systems.
Pick one repetitive workflow your team does (e.g., new contact form submission → enrich → notify). In Zapier or Make, build a 3-step zap with an AI step in the middle (summarize, classify, draft). Ship it.
6AI in Coding — Even If You Don't CodeCursor, Claude Code, GitHub Copilot, Codex. Who they're for, what they replace, what they enable.30 minIntermediate
Coding tools are now the leading edge of agentic AI. Even non-engineers benefit from understanding what's possible — because it changes what to ask your engineers (or contractors) for.
- Coding assistant
- AI in your code editor. Suggests completions, explains code, writes tests, refactors. Multiplies developer productivity.
- Agentic coding
- AI that does the work end-to-end — reads files, plans, edits, runs tests, iterates. Cursor's agent mode and Claude Code are examples.
- MCP (Model Context Protocol)
- Open protocol that lets AI tools securely connect to your databases, APIs, and services. Why Claude Code can talk to your GitHub, Slack, Linear, etc.
Even if you don't code: open v0.dev and prompt 'Build me a one-page landing for [my business] with email signup.' Watch a real component get built. You now know what's possible to ask for.
7AI Coworkers — Daily-Work ToolsBeyond the chat box: tools that act like teammates for research, decks, browsing, and execution.30 minIntermediate
The leap from 'AI as a thing I open' to 'AI as a coworker' happens when you adopt tools that live alongside your work — research, browsing, presentations, autonomous tasks.
- AI coworker
- An AI tool that acts on your behalf, in context, where you actually work — not just inside a separate chat window.
- Autonomous agent
- AI that takes a goal and does the multi-step work itself, checking in only when it needs to.
Pick the next thing on your to-do list. Try to do it with an AI coworker instead of by hand: research → Perplexity, deck → Gamma, multi-step task → Manus, recurring workflow → Claude Project. Note where it saved time and where it didn't.
8Project Folders — Organize Your AI Like Your TeamStop starting from scratch every conversation. Set up Projects, knowledge files, and folder structure once.20 minIntermediate
The 10x leverage in AI is repeatable context: same instructions, same knowledge, same voice every time. Without project structure, you re-explain yourself constantly. With it, the AI 'knows' your business.
- Project / Custom GPT / Workspace
- A persistent container with a system prompt + knowledge files + chat history. Claude calls it Projects. ChatGPT calls them Custom GPTs. Same idea.
- Knowledge files
- Documents you upload once that the AI can reference in every conversation. Style guides, SOPs, FAQs, examples.
- System prompt (project-level)
- The standing instructions for the project: who the AI is, how it should behave, what to never do.
- Conversation hygiene
- Start a fresh conversation when the topic changes. Long conversations dilute focus and burn through context.
Pick one role on your team (e.g., 'sales SDR'). Create a Claude Project named after it. Upload: voice guidelines, ICP doc, top 10 objection responses, 5 example emails. Set a system prompt. Save the link. Now anyone on that role has a teammate.
People build a Project, never tune the system prompt, then complain it's not better than ChatGPT. The system prompt is the brain. Tune it.
9Free-Tier Experimentation GuideWhat's free, what's worth paying for, and the order to try things.15 minFoundation
You can run a serious AI adoption program for $0/month per person to start. Knowing the free landscape removes 'we can't afford to experiment' as an excuse.
- Free tier patterns
- Most tools offer: unlimited basic use with caps, OR free trial with all features. Read which is which before adopting.
- When to upgrade
- Upgrade when free is genuinely blocking you, not when you 'might want more.' Most people overspend by upgrading too early.
Make a 'try this month' list of 3 free tools you haven't used. One slot per category: chat, automation, coworker. Block 30 minutes per tool to try it on real work. Decide go / no-go after each.
10Advanced — When You Actually Need an AgentAgents vs chatbots. Why most teams should NOT build agents yet. When and how it's worth it.30 minAdvanced
'Agent' is the most overused word in AI right now. Most use cases don't need an agent — they need a well-tuned Project. But when an agent IS the right tool, it changes what a small team can accomplish.
- Chatbot vs agent
- Chatbot answers your question. Agent takes a goal and figures out the steps itself, taking real actions across systems.
- When you need an agent
- Multi-step tasks across multiple systems, where the steps can vary based on what's found. If a workflow is the same every time, you don't need an agent — you need automation (Module 5).
- Agent platforms
- Frameworks/products that let you build or run agents: Claude Agent SDK, OpenAI Agents, AutoGen, CrewAI, LangGraph. Plus consumer agents like Manus and Devin.
- Tool use / function calling
- Letting the model call APIs, search, edit files. The mechanical foundation of every agent.
- Evals
- How you measure whether an agent is doing the right thing reliably. Production agents without evals will silently break.
Don't build an agent yet. Instead, write down 3 workflows you THINK need an agent. For each, ask: 'Is this multi-step? Multi-system? Variable? If yes to all three, agent. If not, you need a Project (Module 8) or automation (Module 5).' Most rows will not be agents. That's the lesson.
Building an agent before you've maxed out Projects + automations is almost always a mistake. Agents are powerful but expensive (in time and tokens) and brittle without evals.
AI Opportunities in Universal — Cross-Functional
Built for hybrid businesses, multi-line operators, and any team that doesn't fit one industry box. The same patterns that move the needle in every operation: meeting notes, SOPs, customer email, internal knowledge, proposals, status reports, hiring, content, voice memos, hard conversations, and pricing. Pick this pack if your business spans multiple verticals, you wear several hats, or you want the broadest cross-functional starter kit. AI drafts. Humans send.
Turn raw call notes or transcripts into owners, deadlines, follow-ups, and open questions — the meeting actually moves work after.
Capture how a process is done today and convert it into a clean, repeatable SOP a new hire could follow on day one.
Generate first drafts of standard replies in your voice from sample emails — quick edits before sending, never auto-send.
Build a private GPT or Claude Project that answers staff questions from your handbook, policies, processes, and resources.
Generate first-draft proposals or quotes from a structured intake — your structure, your voice, flagged assumptions.
Turn activity logs or a 5-minute voice memo into a clean weekly status update with owners, blockers, and next-week focus.
Cold outreach variants, social posts, presentation outlines, and document explainers — operator-grade drafts, not generic AI slop.
Your Prompt Pack
16 prompts pre-tested for universal — cross-functional. Each one is a starting point — edit the bracketed placeholders for your situation.
Meeting notes to action planAnyone
You will receive meeting notes or a transcript. Produce: (1) one-sentence summary, (2) decisions made, (3) action items as Owner — Action — Due, (4) follow-up topics, (5) open questions. Be conservative — if an owner or due date isn't specified, mark it as 'unassigned.' Do NOT invent owners or dates. Notes: [PASTE]
SOP from interviewOperations
You are documenting a process as an SOP. Use only the information provided. Structure: trigger, owner, prerequisites, numbered steps with sub-bullets, decision points, tools, expected timing, handoffs, quality check. Output as Markdown. Flag anything that needs more detail to be reproducible. Process interview notes: [PASTE]
Email draft (in voice)Anyone
Draft an email response in my voice. My voice: [3-5 SAMPLE EMAILS or VOICE NOTES]. Goal of this email: [GOAL]. Recipient: [WHO]. Constraints: [CONSTRAINTS]. Length: [LENGTH]. Do NOT add 'I hope this finds you well' or other generic openers I never use. Incoming email or context: [PASTE]
Customer FAQ answerCustomer service
You answer customer questions using only the company knowledge base provided. If the answer isn't in the knowledge base, say so and recommend escalation. Tone: helpful, direct, no fluff. Do NOT make up policy or pricing not in the knowledge base. Knowledge base: [PASTE OR ATTACH] Customer question: [PASTE]
Job posting drafterHR / hiring manager
Draft a job posting for [ROLE] at [COMPANY]. Use the role intake provided. Structure: hook (1-2 sentences), what you'll do, what you'll bring, what we offer, how to apply. Avoid generic corporate language and gendered/coded language. Inclusive tone. Max 350 words. Role intake: [PASTE]
Agent Templates
Ready-to-clone agents your team can build in Build Lab 1 or 2. Each one comes with setup steps your facilitator can read aloud.
Meeting Notes to Actions
Universal note → action item processor with conservative owner/date assignment.
- Create a Claude Project named 'Meeting Notes to Actions'.
- Upload your team's preferred action-item format and 2-3 example outputs.
- System prompt: 'Convert any meeting notes or transcript into our standard action-item format. Conservative — flag unassigned owners. Never invent owners or dates.'
- Test with last week's meeting notes.
System prompt (copy/paste)
You convert meeting notes or transcripts into our standard action-item format. Always include owner, action, due date. If owner or due date isn't specified, mark as 'unassigned.' Never invent. Output structure: summary, decisions, actions, follow-ups, open questions.
SOP Generator
Turns process interviews and walkthroughs into reproducible SOPs.
- Create a Claude Project named 'SOP Generator'.
- Upload 2-3 existing SOPs as templates.
- System prompt: 'Produce SOPs in our standard format. Flag missing detail that would block a new hire from running the process.'
- Test by recording a 5-minute walkthrough of a small process and pasting the transcript.
Internal FAQ Bot
Answers staff questions from internal policies, handbook, and process docs.
- Create a Custom GPT named '[Company] FAQ Bot'.
- Upload your handbook, policies, SOPs, and internal docs as knowledge.
- Instruct: 'Answer only from uploaded knowledge. Cite source documents. If not in knowledge, say so and recommend who to ask.'
- Share with team via shared workspace. Restrict to staff.
Voice Memo to Action Items
Captures a 1-5 minute brain dump and converts it into structured tasks, decisions, and questions.
- Create a Claude Project named 'Voice Memo to Actions'.
- Upload 1-2 examples of how you like your action items formatted.
- System prompt: 'Convert dictated voice memo transcripts into structured output: what I'm doing, decisions made vs open, my tasks, delegated tasks, questions, and one thing I'm avoiding that I should name. Honest tone — call out hedging.'
- Workflow: record memo on phone → paste transcript → get structured output → drop tasks into your task manager.
Printable Handouts
Print these and bring them to AI Day. Each is designed for one-page-or-less printing and built for actual humans to fill out by hand. Click any handout to open the print view.
Facilitator Script (Team)
Hour-by-hour CEO-readable script with talking points for the whole day.
Open print viewUse-Case Checklist
Score AI opportunities by impact, ease, and risk during workflow mapping.
Open print viewPre-Work Survey
Send to the team 3 days before. Surfaces friction and AI experience.
Open print viewWorkflow Capture Worksheet
What we built, what it solves, time saved, what's next.
Open print viewWhat I Learned Reflection
Each person captures their own insights and one habit to keep.
Open print viewLeadership Recap Template
Auto-fillable from the worksheet. Hand to your CEO at the end of the day.
Open print view30-Day Roadmap
Commit to one habit, one workflow, and one experiment for the next 30 days.
Open print viewIndustry Prompt Library
All your industry prompts, formatted for printing or distribution.
Open print viewAI Vocabulary Cheat Sheet
Tokens, context, RAG, agents — every term explained in one printable card.
Open print viewTool Stack Map
All ~40 AI tools by category, with free-tier notes. Printable single page.
Open print viewFree-Tier Quick Reference
What's free where, and which tier to start with for each tool.
Open print viewROI Calculator
Calculate hours and dollars recovered. Most teams under-count by 50% — this fixes that.
Open print viewAI Usage Policy Template
Copy this. Edit the [BRACKETS]. Email to your team. The policy nobody else gives you free.
Open print viewShould I Use AI for This?
Decision tree to run any task through before pasting it into AI. Saves more time than any prompt.
Open print viewHow This Compares to Other Free Workshops
Honest grid: this playbook vs Google, HubSpot, Atlassian, Liam Ottley, Allie Miller.
Open print viewPrint one set per attendee. The full handout pack is ~10 pages.
Open All-In-One Print ViewImportant Compliance Notes
- —Don't paste customer PII, financial data, trade secrets, or anything covered by an NDA into consumer AI tools without checking your company's data policy. Use enterprise/team plans where data isn't used for training.
- —AI drafts, humans send — never auto-publish or auto-send AI-generated communication to clients without human review until you've measured quality over time.
- —Keep an audit trail for any AI used in regulated workflows. If your industry has specific rules (HIPAA, FINRA, FERPA, ITAR, etc.), check the industry pack for that vertical instead of using this universal pack.
- —Hallucination check: AI will confidently invent citations, statistics, dollar amounts, and policy details. Always verify anything specific before relying on it.
- —If your business spans multiple regulated industries, default to the strictest applicable rule set, not the most convenient.
What teams like yours have done
A 40-person services firm cut weekly status-update writing from 90 min to 15 min by using a meeting-notes-to-actions agent.
An ops team at a 25-person company built an internal FAQ bot from their wiki and reduced 'how do I' Slack questions by ~70%.
A solo operator running three product lines uses a voice-memo-to-actions Claude Project as their daily standup with themselves — closes the day with a tracked task list in 4 minutes flat.
When DIY isn't enough
Want Chelsea to run AI Day for your team?
Same playbook. Custom agenda built around your maturity stage. Real workflows shipped during the day. Custom recap deck for leadership. Path into ongoing implementation if you want it.