In this article
- What are the best AI prompts for team AI training?
- What prompts should you use in a first team AI session?
- How do you build a team prompt library?
- How do you audit your team's AI skills with a prompt?
- How do you check AI use against your company AI policy?
- How do you measure whether team AI training is working?
- What does good team AI training look like?
- AI prompts for team AI training: questions people ask
The short answer
The best AI prompts for team AI training are practice drills that improve people's own prompts, a prompt library starter, an AI skills audit, role-specific exercises, a policy check and a progress measure. They work best inside live training on real tasks, with an agreed AI policy, rather than as a prompt list handed out once.
The best AI prompts for team AI training aren't the clever ones. They're the drills, templates and checks that turn a room of curious, slightly nervous colleagues into a team with shared habits. A prompt list gets read once. Practice on real work sticks.
I run AI training for UK businesses most weeks, in person and online, across ChatGPT, Claude, Microsoft Copilot and Gemini. These are the eight prompts I'd give any leader planning to train their team, whether you run the session yourself or bring someone in. I'll also be honest about where a prompt list stops being enough.
What are the best AI prompts for team AI training?
The best AI prompts for team AI training are practice drills, a prompt library starter, a skills audit, role-specific exercises, a policy check and a way to measure progress. Each one does a training job, not just a work job. Here is the set.
| Prompt | Training job | When to use it |
|---|---|---|
| 1. Fix my prompt drill | Builds the core prompting habit | First session, first 20 minutes |
| 2. Lazy vs better prompt | Shows what good looks like | First session |
| 3. Prompt library starter | Captures what works | After every session |
| 4. Teach me | Self-paced learning | Between sessions |
| 5. AI skills audit | Sets a baseline | Before any training |
| 6. Role-specific exercises | Makes practice relevant | Planning each session |
| 7. Policy check | Makes the rules usable | Day to day |
| 8. Measure progress | Shows whether it worked | Six weeks after |
One rule for every prompt here: don't paste confidential information or anyone's personal data into an AI tool your organisation hasn't approved. In training, use anonymised or made-up material.
What prompts should you use in a first team AI session?
In a first team AI session, use prompts that make people improve their own work in front of them, starting with a prompt they typed this week. Demonstrations impress people. Doing it themselves changes what they do on Monday.
Prompt 1: The fix my prompt drill
I'm learning to write better AI prompts. Here is a prompt I typed this week: [paste a short, vague prompt you actually used]. Before rewriting it, ask me up to five questions to find out what's missing: who the output is for, what material you need, the format, the length and anything to avoid. Wait for my answers. Then rewrite the prompt using my answers and explain, in three bullet points, what you changed and why. Finally, run the improved prompt so I can compare the two outputs.
Why this prompt works
People learn more from fixing their own prompt than from reading someone else's, and the questions teach them what to include next time.
Prompt 2: The lazy prompt versus the better prompt
Put these side by side on the screen, run both on the same report, and let the room see the difference.
Summarise this report.
You are briefing a busy operations director who has five minutes. Summarise the report I paste below. Output: - Three bullet points: what it says - Two bullet points: what it means for a [size] UK [sector] business like ours - One question we should ask before acting on it Under 150 words. Plain British English. Quote the page or section for each point. If something is unclear in the report, say so rather than filling the gap. Report: [paste here]
Then ask everyone to run the better version in two different tools, say ChatGPT and Copilot, and compare. It's the quickest way I know to show that the prompt matters more than the tool. My guide to AI prompt frameworks gives people a structure to remember afterwards.
How to run the first session
- Ask people to bring a real task.Something from this week: an email, a report to read, a plan to draft. Anonymised where needed.
- Start with the drill.Twenty minutes on Prompt 1 before any theory. Confidence first.
- Show the pair.Run the lazy and better prompts live so everyone sees the gap.
- Practise on the real task.Each person works on their own task with a colleague, then shares one result.
- Save what worked.Run Prompt 3 on the best examples and start the team library that afternoon.
How do you build a team prompt library?
You build a team prompt library by turning prompts that already worked on real tasks into templates, not by downloading a list of 500. A library of fifteen prompts your team actually uses beats a long one nobody opens.
Prompt 3: Prompt library starter
You are helping our team build a shared prompt library. Below is a prompt that worked well for me and the output it produced. Turn it into a reusable template with: - A short title - When to use it (one sentence) - The prompt, with [square bracket] placeholders replacing anything specific to my case - What material to paste in - One tip for improving the result - A warning about any data that must not go in Keep the prompt under 140 words and format it for a shared document. My prompt: [paste] The output: [paste]
Why this prompt works
It turns one person's good result into something the whole team can reuse, with the data warning built in.
Prompt 4: The teach me prompt
I'm a [job title] who has used [ChatGPT, Claude or Copilot] a little. Teach me how to [skill, for example get reliable summaries of long documents]. Explain one idea at a time in under 100 words. After each idea, give me a small exercise based on the kind of work I do: [describe your work in one line]. Wait for my attempt, then give me honest feedback and move to the next idea. After five ideas, give me a checklist I can keep. Don't move on until I've tried each exercise.
Why this prompt works
It makes the AI a patient tutor that sets work and marks it, rather than a lecturer that dumps ten tips at once.
How do you audit your team's AI skills with a prompt?
You audit your team's AI skills by using a prompt to design a short, anonymous self-assessment, then running it before training so you have a baseline. Without one, you can't tell later whether anything changed.
Prompt 5: AI skills audit
You are an L&D adviser designing a short AI skills audit for a [number]-person UK [sector] business. Our teams: [list]. Our approved AI tools: [list]. Write a 12-question self-assessment staff can finish in ten minutes. Cover: how often they use AI, for which tasks, confidence writing prompts, checking outputs for accuracy, knowing what data must not go in, and awareness of our AI policy. Mix multiple choice with a few one-line answers. Make it anonymous by design: no questions that would identify someone in a small team. Then explain how to score responses into three levels: starting, practising, confident.
Why this prompt works
It gives you a baseline before training, which is the only way to show later that anything changed.
If you'd rather map skills by role, our AI skills matrix template does that, and our free AI Capability Assessment gives you an outside view.
Prompt 6: Role-specific exercises
I'm planning a hands-on AI session for our [team name] of [number] people. Their regular tasks include: [list five real tasks]. Our approved tool is [tool]. Design six practice exercises: one for each task, plus one on checking an AI answer for accuracy. For each give: - The task in one line - Sample material to bring (anonymised or made up, never real personal data) - A starter prompt - What a good output looks like - A common mistake to watch for Order them from easiest to hardest. Each should take 15 minutes or less.
Why this prompt works
Exercises built on the team's real tasks are the ones people carry back to their desks.
Managers are usually the best people to supply the five real tasks. The best AI prompts for managers covers their side.
How do you check AI use against your company AI policy?
You check AI use against your policy by pasting the policy and the planned task into your approved AI tool and asking it to flag what's allowed, what isn't and what's unclear. It makes a written policy usable in the moment.
Prompt 7: The policy check
Below is our company AI policy and a description of how I want to use AI for a task. Act as a careful compliance colleague, not a lawyer. Tell me: 1. Whether this use looks allowed, not allowed or unclear under our policy, quoting the relevant section 2. What data I would be putting into the tool, and whether the policy permits it 3. What I should do before going ahead (anonymise, switch to the approved tool, or ask [role]) If the policy doesn't cover this use, say so plainly rather than guessing. Policy: [paste] My planned use: [describe]
Why this prompt works
It makes the policy something people consult in the moment, rather than a PDF they signed once.
How do you measure whether team AI training is working?
You measure team AI training by repeating the skills audit about six weeks later and asking people what they now use AI for. Usage and confidence on real tasks tell you far more than attendance.
Prompt 8: Measure progress
I ran AI training for [number] staff on [date]. Below are anonymised results from our skills audit before training and again [six weeks] after, plus short staff comments on what they now use AI for. 1. Show a table of each question with before and after results. 2. Highlight the three biggest changes. 3. List the tasks people say they now do with AI. 4. Suggest three practical next steps and one thing this data can't tell us. Don't claim time or money savings unless the comments state them. Data: [paste here]
Why this prompt works
The last line keeps the report honest, which matters when you take it to the board.
For the business case side, see how to measure the ROI of AI training.
What does good team AI training look like?
Good team AI training is live, hands-on and built around the work your people already do, with a policy agreed first and a follow-up to make it stick. Prompts are the raw material. Training is what turns them into habits.
| Aspect | Prompt list on its own | Live training on real work |
|---|---|---|
| Habits | Depends on who reads it | Shared across the team |
| Checking outputs | Rarely taught | Practised on real examples |
| Data and policy | Left to individuals | Agreed and applied in the room |
| Relevance | Generic tasks | Your team's actual tasks |
| Follow-through | Fades after a week | Library, owner and review |
That's the difference between awareness and hands-on training. Our Basic AI Training gets the whole team confident on real tasks, the Claude & ChatGPT Prompts, Context & Skills Deep Dive goes further for teams already using AI, and our AI Advisory and Custom Training Programme fits the rollout around your business. For the people side of a rollout, read how to get your team to adopt AI.
AI prompts for team AI training: questions people ask
What are the best AI prompts for team AI training?
The most useful are a fix-my-prompt drill, a lazy versus better comparison, a prompt library template, a teach-me prompt, a skills audit, role-specific exercises, a policy check and a progress measure. Each does a training job rather than a single work task.
Is a prompt library enough to train a team?
No. A library helps, but on its own it rarely changes behaviour. People need practice on their own tasks, clear rules on data and someone keeping the library current. See how to get your team to adopt AI.
How long should a first team AI training session be?
Long enough for everyone to work on a real task, not just watch a demo. In my experience half a day to a day works well for a first session, followed by a short check-in a few weeks later. Our guide to how long AI training for a team takes goes into more detail.
Should we write an AI policy before training staff?
Yes. People need to know which tools are approved and what data must stay out before they start practising. Read why your AI policy should come before training.
Which AI tool should we train our team on?
Train on the tool your organisation has approved and pays for, often Microsoft Copilot for Microsoft 365 businesses or ChatGPT or Claude business plans. The prompting habits transfer between tools, so the choice matters less than people think.
How do you measure whether AI training worked?
Run a short skills audit before training and again around six weeks later, and ask staff which tasks they now use AI for. Avoid claiming time savings unless people report them.
Keep reading
Train your team on their real work, not a prompt list
We run hands-on AI training for UK teams, in person and online, built around the tasks your people already do and the AI policy you agree first. Start with a free AI Capability Assessment or tell us about your team.
- Basic AI Training Your whole team, prompting properly in a day.
- Prompts, Context and Skills Deep Dive A half day on Claude and ChatGPT.
- AI Capability Assessment See where your team stands first.
Harry Lang
Harry Lang founded The Oxford AI School in Witney, Oxfordshire. He trains UK businesses and teams to use AI well, writes the Marketing Unfiltered newsletter, and is the author of Learn To Fly With AI.