Many teams hear about AI from headlines and vendor videos, and the reaction swings between excitement and suspicion. A short internal demo can settle both. It shows colleagues what the tool does with work they recognise, where it stumbles, and what the company is actually proposing to do.
This article is for an owner or team lead who wants to run that session in one hour without a presenter, a slide deck or technical knowledge. It assumes you have already looked at one tool and one task.
Choose one real task
Pick a task that your team does every week and understands well. Good examples are summarising a long email thread, drafting a reply to a routine customer question, turning rough notes into a tidy checklist, or sorting a list of requests by type.
Avoid tasks that need confidential data, carry legal weight or cannot be undone. If you are unsure, apply the five-question check on tasks to keep away from AI first. If you have not yet picked a first use case, start with choosing a first AI use case.
Use real material, with names and private details removed. A demo on invented examples is easy to dismiss with "our work is different".
Prepare in thirty minutes
You do not need much preparation, but you do need some.
- Write down the task in one sentence and what a good result looks like.
- Collect three to five real samples, including one awkward one.
- Run the task yourself once, privately, so nothing in the session is a surprise.
- Note how long the task takes a person today. This is your only comparison.
- Check that the tool and account you will use are the ones the company would actually allow.
Decide who will drive and who will judge. The person who normally does the task should judge the result, not the person who is enthusiastic about the tool.
The sixty-minute plan
| Time | What happens |
|---|---|
| 0–5 min | State the task, why it was chosen and what the session will not cover |
| 5–15 min | Show the task done by hand or as it works today, so everyone shares the starting point |
| 15–35 min | Run the same task with the tool, live, on two or three samples |
| 35–45 min | Show a failure on purpose: the awkward sample, a vague instruction or a missing detail |
| 45–60 min | Discuss and decide |
Run it live
Resist the urge to show a pre-recorded or polished result. Live use shows the real wait, the real wording of the instruction and the real corrections. Let colleagues suggest changes to the instruction. Participation is the point of a demo; people trust what they have touched.
Show the limits
Include a result that is wrong, vague or oddly worded, and ask the room to spot the problem. This teaches the habit that matters most: reading the output before using it. If the tool produces a confident but incorrect statement, say so plainly. It is better that the team hears this from you than discovers it on a customer email. The same honesty helps later when you train staff on a new tool.
Keep the claims small
Do not promise time savings you have not measured. "This took four minutes instead of fifteen on three samples" is a fair statement. "This will transform our work" is not.
Decide before people leave
The last fifteen minutes are the reason for the session. Ask for one of three outcomes:
- Try it: a small trial on this task for two weeks, with a named owner and a person who reviews every output.
- Park it: not now; write down what would need to change, such as better samples or a different tool.
- Drop it: the task is not a fit, and that is a useful result.
Write the decision, the owner and the date for the next look on a single page and send it to everyone the same day. If you go ahead, agree the rules for use in the same message, for example with a one-page usage policy for AI tools.
Common mistakes
- Showing five tasks instead of one
- Using a task nobody on the team actually does
- Presenting only the best result
- Letting the sceptic or the enthusiast dominate; ask quieter people what they noticed
- Ending without a decision or an owner
- Using data that should not leave the company
Checklist
- One real task, described in one sentence
- Three to five real samples, one of them awkward
- A private test run before the session
- A person who does the task today acting as judge
- At least one failure shown on purpose
- A decision (try, park or drop), an owner and a review date, sent in writing
Sources
- Practical guidance based on common facilitation and change-management practice for small teams, offered as a planning aid. The timings are starting points to adapt to your own team. No client results or market figures are claimed in this article.