Everyone was using AI. No two people the same way.
More than seventy people and one shared method for working with artificial intelligence, built on the tools the company already had and on the cases each team brought along.
The challenge
Artificial intelligence was already inside the company. There were corporate tools and, on top of that, people were using commercial models on their own. Access was not the problem.
The problem was fragmentation: every person had built their own method. Results varied widely between teams, one person's good find never reached anyone else, and there was no shared criterion about what could be handed to a machine, what always had to be checked, and what should never leave the company's own tools.
And there was a gap underneath it. Nobody had measured where the time actually went. Without that, any training ends up being a catalogue of tools that is forgotten within two weeks.
Approach
The brief was not to teach a tool but to leave a shared method: one way of working with artificial intelligence, the same for everyone, resting on the corporate tools already paid for rather than on whatever each person had come across.
Before designing anything we interviewed department and team heads, so the work was built on real cases rather than generic examples. From there we ran several working sessions focused on two things: adopting that common method with the company's own tools, and identifying and working through the practical cases each group of participants brought along.
- Prior interviews with department and team heads, to design around real cases
- One common method for working with AI, resting on the corporate tools already paid for
- Shared criteria on what to delegate, what to always review, and what never to take outside the company's own tools
- Working sessions focused on adopting that method
- Identifying and working through the practical cases each group of participants brought along
- Systematic capture of the tasks that fill the day, during the sessions themselves
- Aggregate analysis afterwards, ranking them by how often they were actually mentioned
- A retrospective afterwards and supporting material with step-by-step guides
Key outcomes
70+
people took part, from very different areas of the company.
1
shared method for working with artificial intelligence, documented and resting on the corporate tools already paid for.
1
inventory of tasks ranked by how often they were actually mentioned rather than by hunch.
3
priority efficiency pools on which to decide where to automate first.
The inventory made clear where the time goes, and it was not where people assumed. The heaviest load is not transactional work but repetitive knowledge work. It is worth saying that this is an aggregate reading of recurrence, not a statistical count; it is good enough to prioritise, not to boast about precision.
Training and diagnosing at the same time
The most useful part of the programme was not in the script. By asking each group to bring its own cases, the sessions doubled as a data-gathering exercise about the company's real work. By the end, the organisation had more than trained people: it had a shared method and a map of where its time goes, which is exactly what you need to decide where to automate next.
Got a process that hurts?
Tell us about it in half an hour. No forty-page deck. If we are not the right people, we will say so and point you towards someone who is.