- Content adapted to participants' roles, tools and starting level
- Exercises using authorised, anonymised or synthetic cases
- A verification method for facts, calculations, sources and produced documents
- A simple before-and-after measure on a representative task, with no promised gain
A target task, not a tour of tools
We start with a short inventory: who already uses AI, for which tasks, with which tools and data. Then we pick a useful exercise, such as summarising a file, structuring notes, drafting or extracting fields.
The programme follows the participants' level and context. A team discovering AI does not need the same vocabulary as one already automating workflows.
What participants practise
The goal is to decide when to use AI, what information to provide, how to inspect the result and when to stop for human review.
- Break down a task and define an acceptable output
- Provide context and the required format without attaching unnecessary data
- Spot an unsupported claim, an omission and false precision
- Check an output against the source, document or system of record
- Keep a reusable method rather than a collection of magic prompts
Confidential data remains an explicit decision
Before the exercise, we list the authorised tools and information categories. Personal data, client files, trade secrets and confidential documents are not copied into a public tool by reflex.
When a case cannot use real data, we use an anonymised or synthetic example. The training provides a clear escalation rule for ambiguous situations.
Measure the exercise without promising an outcome
Participants complete the same representative task with their starting method, then with the practised workflow. We can compare time, corrections, required elements and detected errors. The result describes this exercise and group, not a guaranteed company-wide productivity gain.
| Criterion | What we observe | Evidence retained |
|---|---|---|
| Accuracy | Correct elements and errors caught | Completed review checklist |
| Completeness | Required information present | Comparison with the reference output |
| Effort | Time and rework needed | Exercise measurement |
| Autonomy | Steps repeated without assistance | Reusable workflow and guidance |
What remains after the session
Deliverables match the chosen format and team context. They help repeat the task, brief an absent colleague and document internal awareness initiatives.
- Concise material adapted to the tools used in the session
- Step-by-step workflow for the chosen case
- Verification checklist and data-handling rules
- Anonymised result of the reference exercise
- List of questions or use cases that need further scoping