Team training
AI training for teams
In most companies somebody already uses AI, each in their own way and without any agreement about it. One person drafts text, another tidies data, a third uses nothing because nobody has said whether it is allowed. Training is useful not when it explains how models work but when everyone leaves knowing which of their own tasks it suits, how to check the output, and what must not be done with company information.
Discuss a training scopeIn short: what useful training looks like
Useful training is built around a role and its real tasks. A marketer, an accountant and a developer need different things, and one shared session for all three gives the least to each. Second, people have to learn to check an answer rather than trust its tone. Third, it must be clear which information may be used and which may not. And fourth: some tasks that surface during training do not suit AI at all, because a precise rule already exists for them.
When training is the right answer
- People already use AI on their own initiative and the company has no agreement about what is allowed.
- You want the team to work faster with what you already have, without building a new system.
- Somebody needs to be able to check the output — particularly where a mistake reaches a customer.
- You want to understand which of your tasks are worth giving to AI at all, and which are not.
When something simpler is enough
- The task is single and repetitive — then automating it with a rule beats teaching people to do it by hand with a new tool.
- You need a working solution rather than a skill: if a system has to produce the result, that is AI solutions work.
- The company has not yet agreed which information may be used — settle that before training, not during it.
What different roles need
Scope is set by role, because different people benefit from different things. Below is an illustrative example of how that looks. It is not a price list and not a promise about a particular outcome.
| Who is trained | What they need to learn | How it is checked |
|---|---|---|
| Marketing and content | Producing a draft from your own material rather than asking for generic text; spotting output that merely reads well. | A practical task with your own material: a draft produced, and the facts in it that needed checking identified. |
| Sales and customer service | Preparing a reply quickly and correcting it; never sending an unread message to a customer. | Several real customer cases: a reply prepared, with an explanation of what was changed in it. |
| Administration and accounting | Where AI helps with documents and where it must not be used because an exact figure is required. | A document task: extract the data and state which values must not be accepted unchecked. |
| Managers | Which team tasks are worth giving to AI, and which should be automated or left to a person. | A list of the team’s tasks with a decision against each and the reasoning for it. |
| Technical team | Where AI speeds up work with code, and why the output still has to be read. | A review task: find what the suggested code does differently from what was intended. |
How the work runs
- 01
A conversation about real tasks
Before the training we establish what people do daily and where they get stuck, and ask for a few real examples — emails, documents, reports. The sessions are built around those, so people practise on their own material rather than on invented cases.
- 02
Agreeing about information
We write down which data may be used and which may not. This is frequently the most valuable part, because without a clear answer people either use nothing or use everything. If the company has not settled it, we help put it into words.
- 03
Practice on your own work
Sessions run on real tasks: produce, check, correct. We work separately on recognising a wrong answer that sounds convincing — a skill nobody acquires from theory.
- 04
What we measure, and what comes next
We agree how to confirm the training worked: specific tasks people should be able to complete unaided. If the sessions reveal work better automated or built as software, we record that separately — it is different work, not part of the training.
What drives cost and duration
- How many distinct roles are trained — one group doing similar work needs less time than four different ones.
- Whether you have real examples to work with, or they have to be collected first.
- Whether the company has already agreed which information may be used.
- Whether one session is enough or several with practice in between — the second gives more but takes longer.
- Whether the team needs written rules afterwards, or the sessions themselves are enough.
Worth considering before we talk
- 01Which teams and which roles will be trained, and what do they do daily?
- 02Is there already an agreement about which company information may be used?
- 03Who already uses AI on their own initiative today, and how?
- 04Which work would you most like to speed up first?
- 05Who will answer the team’s questions after the training?
Let us scope the training
A note about which teams will take part and what work you want to speed up is enough. Please do not send real customer data or documents at this stage — a general description is fine. If it turns out that automating beats training, we will say so.