Generic AI training produces informed people who change nothing

The problem with most AI training is not the content — it is that it is about AI in general rather than about your invoicing process, your service desk or your quality data. Sessions built on your own material end with a decision or something working, which is the only version that survives the week after.

  • A decision or an artefact. Each session ends with something concrete: a shortlist, a policy, a working prototype, a process redesign.
  • A shared vocabulary. So that the executive, the manager and the engineer are arguing about the same thing, which is not currently the case in most companies.
  • Calibrated expectations. What these systems genuinely do well, where they fail quietly, and what that means for your specific processes.
  • Capability that stays. Deliberate transfer, so the second project needs less of us than the first.
Talk to us about this

What is delivered

  • Executive session

    Half a day: where value realistically is, what it costs, what the regulatory position means for you, and how to tell a real proposal from a fashionable one. No tooling, no demos of other companies.

  • Manager and process workshop

    One or two days on your own processes: mapping where decisions are made, where AI would help and where it would not, ending with a ranked shortlist the room agrees on.

  • Engineering workshop

    Two to three days hands-on with your data: retrieval, evaluation, agents, guardrails and cost. People leave having built and broken something.

  • AI usage policy

    The practical one: approved tools, data classes in plain language, what may leave the building, and an amnesty for what has been happening already.

  • Prompt and pattern library

    For your actual tasks, tested and versioned, rather than a collection of general tips that decay in three months.

  • Follow-up clinic

    A session some weeks later on what people tried and where it went wrong. This is where most of the learning actually lands.

How it runs

  1. 01

    Pick real material

    Your documents, your processes, your awkward cases. Preparation is most of the work and it is what makes a session different from a course.

  2. 02

    Run short and hands-on

    Minimal lecture, mostly doing. People remember what they built and forget what they were told.

  3. 03

    End with an artefact

    A decision, a policy, a prototype. A session that ends with enthusiasm and no artefact has produced nothing.

  4. 04

    Come back

    Weeks later, when the practical obstacles have appeared. The follow-up is where the training converts.

A good fit when

  • Different parts of the business have wildly different ideas of what AI can do.
  • Staff are already using AI tools and there is no policy.
  • Your engineers are capable but new to evaluation, retrieval and agent patterns.
  • You want internal capability rather than a permanent supplier.

Not the right service when

  • You need a certification programme. That is a training company, not us.
  • Nobody can spare the time. A half-attended workshop is worse than none, because it produces the appearance of alignment.
  • The decision has already been made and the session is meant to sell it internally. We are the wrong people for that.

Frequently asked questions

Do you use our real data?
Yes, wherever the data rules allow — and that is the point. Working on your material with its actual mess is the difference between a session that changes something and a session people enjoyed.
How many people should attend?
Executive sessions work up to about a dozen. Hands-on workshops work best at six to ten; beyond that people watch rather than build, and watching does not transfer.
Can this be remote?
Executive sessions work well remotely. Hands-on workshops are meaningfully better in a room, mostly because the useful conversations happen while people are stuck, and remote formats hide being stuck.
In which languages?
Polish or English, and mixed groups are fine. Written material is provided in the language the group works in day to day rather than the language of the session.
What about the AI policy — is that legal advice?
No. It is a practical operating policy: approved tools, data classes, what may leave the building, and a log. Where obligations under the AI Act or data protection law need a legal opinion, we will say so rather than improvise one.

Keep reading

Other services