Company AI Day · Multi-Industry / Cross-Functional · Stage 1
Should I Use AI for This?
Decision framework. Run any task through it before spending an hour trying to make AI do it. Saves more time than any prompt ever will.
Step 1 — Is the task suited to AI?
| AI is GOOD at | AI is BAD at |
|---|---|
| Drafting from a template or pattern Summarizing long text Reformatting / restructuring content Brainstorming options Translating between formats (notes → SOP, audio → text) Reviewing for inclusion/exclusion criteria First-pass research synthesis (with sources) Prompting you to think (rubber duck) | Anything requiring fresh judgment about your specific situation Current events / data after its training cutoff Math beyond simple arithmetic (use a calculator / code) High-stakes recommendations without verification Specific facts you can't verify (dates, prices, citations) Long-context recall (it forgets the start) Replacing your taste, voice, or relationships |
Step 2 — What's the cost of being wrong?
- Low (internal draft, brainstorm): Use AI freely, light review.
- Medium (internal decision, public-facing draft): AI drafts, you verify specifics, you decide.
- High (client-facing, regulated, contractual): AI drafts, expert reviews EVERY specific, expert signs.
- Very high (legal / medical / financial advice to a third party): AI may not be appropriate at all without enterprise tooling and an audit trail.
Step 3 — Am I using the right kind of AI?
- One-shot text task → ChatGPT / Claude / Gemini chat
- Same task repeatedly with shared context → Claude Project or Custom GPT
- Connected to other apps (trigger → AI step → action) → Zapier / Make / n8n
- Research with sources required → Perplexity
- Slides / one-pagers / sites → Gamma / v0
- Coding work → Cursor / Claude Code / GitHub Copilot
- Multi-step autonomous work → an agent (and read Module 10 first)
Step 4 — Quick gut check before pasting
Is this safe to share with this AI tool? (See AI Usage Policy)
Have I given the AI enough context (role, goal, constraints, format, examples)?
Will I review the output, or am I about to send it without reading?
Do I know how I'll verify the specific facts in the output?
Step 5 — When the answer is "No, don't use AI"
Just do it yourself. Knowing where AI doesn't belong is half the value of this framework. The other half is doing more of what AI IS good at, on purpose.
Free playbook · centraops.com/company-ai-day · © Chelsea Hulin / CH Tech Ventures
