How to Turn Your Team’s Expertise Into Something AI Can Use
Make your team’s methods and context usable with AI - without flattening the judgement that still needs a person who is responsible for the result. A practical worksheet.
Published · Updated
Start with one kind of work your team already does well. Write down what someone needs to know to do it, the examples that show what “good” looks like, and the constraints you never skip. Give AI the context it can use. Keep clear where a person’s judgement still matters - and who is responsible for the result - so expertise guides the work instead of disappearing into a generic prompt.
What this worksheet is for
Use this for one recurring kind of work, not a company-wide knowledge dump. A useful private method is a strong starting point. The goal is to make the method clear enough that AI can help prepare or structure the work, while the judgement that makes your team distinctive stays with a person.
Use the five prompts below to make one method clear enough to share. Start with the work, give AI useful context, and name the decisions a person still needs to make.
Read the diagram as text
1. Name the work: What do we deliver, and for whom? What does “done” mean? When is it needed? 2. Give AI usable context: Which sources and good examples can it use? What steps, constraints and must-nots matter? 3. Keep judgement with a person: Which calls and exceptions need a person? Who is responsible for the result? 4. Agree who does what: Who prepares, runs, reviews and updates? How does the work move between them? 5. Test one real task: What will we test, when, and with whom? What will we check and improve afterwards? Use only material you are permitted to share with your chosen AI service. Try it once and note what needs to change.
Download the printable worksheet (PDF) · Get the editable text version (Markdown)
Print the worksheet or copy the text version into a shared document. Use only material your team is permitted to share with its chosen AI service. OpenAI’s guidance on clear prompts and context is a useful starting point.
Filled hypothetical example
The following is a labeled hypothetical, not a client result. It shows one completed worksheet for a specialist research memo method.
Block 1 - Name the work (filled)
Kind of work: Client briefing memo after a research sprint
Who it’s for: The client’s product and strategy leads
What “done” means: A clear memo that states findings, confidence, open questions, and recommended next steps - without inventing evidence
When this work usually happens: End of each two-week research sprint
Block 2 - Context AI can use (filled)
Sources and facts: Interview notes in the shared project folder; survey summary; agreed research questions; prior memo template
Examples of good output: Two past memos marked “shipped as-is” in the project archive
Constraints and must-nots: No invented quotes; no client names outside the brief; flag low-confidence findings explicitly
Preferred structure: Context → what we learned → what surprised us → implications → open questions → recommended next steps
Block 3 - Judgement that stays human (filled)
What the model should not decide: Whether a finding is strong enough to recommend a product change; how to frame sensitive feedback
Exceptions: Any claim that could commit the firm to a timeline or cost
Client / reputation calls: Tone when findings contradict the client’s current plan
Who is responsible for the result: Research lead for the engagement
Block 4 - Roles and handoffs (filled)
Who prepares inputs: Researcher who ran the interviews
Who runs the AI assist: Same researcher, using the shared instruction note
Who reviews: Research lead
Who updates instructions: Delivery lead for the research practice
Handoff: Request → prepare agreed sources → AI draft from instruction note → research lead review → send → note what broke
Block 5 - First test (filled)
One real task: Next sprint memo for Project North (hypothetical name)
What we will check: Every finding traces to a source; open questions are honest; recommendations do not overclaim
What we will change after: Add a short “confidence” line under each finding if review keeps inserting it by hand
Date / owner: Research lead, next sprint close
Read the diagram as text
Context supports judgement. Illustrative research-memo flow Usable context: Sources, examples and constraints; Give the method something concrete to use.. AI assist: Structure and first draft; Keep claims tied to supplied evidence.. Human judgement: Interpret findings and exceptions; Research lead reviews recommendations.. Responsible person: Stand behind the result; The research lead decides what can go out.
Common flattening mistakes
One clever prompt instead of usable context. A short prompt can look efficient and still leave out sources, examples, and must-nots. With too little context, the model may produce generic or unsupported material, leaving out what makes your method distinctive.
Hiding disagreement. If senior people disagree about what “good” means, a single instruction note will not settle it. Name the disagreement in the worksheet, or agree on one standard before you share the approach.
No owner for updates. Context goes stale. Without someone responsible for keeping the instructions useful, people return to private habits - even when the first version helped.
Treating judgement as a tip. “Use your judgement” is not enough. Say which calls stay with a person, and who is responsible when the output is not good enough.
A flattened prompt asks the model to imitate expertise. Usable context gives the model what it can use, and leaves interpretation, exceptions, and ownership with a person.
When Design or Implementation help
A filled worksheet can give the team a starting point for a supervised test. If you want help turning that worksheet into a shared way of working - or into a new client-facing capability - that is Design work, sometimes with Implementation when you also need systems connected and tested on real tasks.
If the team cannot yet choose which expertise to externalize first, an Assessment can help decide where AI could make a worthwhile difference. Use it when you need that decision. It is not a compulsory first step.
Mode Lab helps teams improve how their business works - and build new capabilities - with AI. We bring your team’s expertise and AI together to rethink how the work gets done, and help you put it into practice.
If you want to build capabilities yourself through learning and practice, Mighty AI Lab is a separate learning and community pathway for individuals and teams.
Start with one filled worksheet
Complete Blocks 1-5 for one recurring kind of work. Run the first test on a real task. Improve the context from what broke, and keep judgement and ownership explicit.
When you have one filled worksheet and want help turning that expertise into a shared way of working - or a new capability - without losing where judgement sits, start a conversation.