Learn AI through your team’s real work.

We start from a task your people understand, then practise: select the right inputs, frame the request, check the output and measure whether the method actually helps.

  • 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.

Measure the exercise without promising an outcome
CriterionWhat we observeEvidence retained
AccuracyCorrect elements and errors caughtCompleted review checklist
CompletenessRequired information presentComparison with the reference output
EffortTime and rework neededExercise measurement
AutonomySteps repeated without assistanceReusable 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

Frequently asked questions

Is this general ChatGPT training?

No. Tools can be introduced, but the programme starts from one of your team’s tasks and an output that can be checked. The method remains useful when the interface or model changes.

Can we use our documents in the exercises?

Yes, when their use in the selected tool is authorised and participants know the rules. Otherwise we prepare anonymised or synthetic examples that keep the difficulty without exposing the data.

How do you check whether the training helped?

Before the session, we define a simple task and review grid. The group does the exercise, checks the result and compares time, corrections and required elements. We report what we observed without extrapolating a guaranteed gain.

Does the training provide an AI Act compliance certificate?

No. It can support your team’s AI literacy and record the topics covered, but it does not certify your organisation’s legal compliance. Needs depend on people’s roles, the systems they use and their risks.

Do participants need prior AI experience?

No. We adapt the vocabulary and exercises to the starting level. For a highly mixed group, it may be better to separate discovery, practical business use and advanced automation.

How long is the training and how many people can join?

That depends on the number of roles, current experience and chosen exercise. We propose a format after a preparation call instead of imposing one fixed duration on every team.

Let’s discuss your context.

A first conversation frames the need, access requirements and a realistic next step.

Scope a training session