From training to working AI solutions at TNO

In many knowledge organisations AI gets stuck in inspiration, isolated pilots or individual experiments. Employees see the potential of tools like Microsoft Copilot, but don't know how to apply them structurally within their existing processes and technology stack.

The result: a lot of energy, limited structural impact. Productivity stays dependent on individuals rather than a shared way of working across teams.

At TNO the question wasn't whether AI was relevant, but how employees could build concrete solutions themselves that add direct value to their daily work and that are transferable across the organisation.

The challenge was to move AI from inspiration and isolated pilots towards scalable, transferable solutions with structural impact on productivity and ways of working.

Instead of a classic training, we set up a Way of Working programme: two intensive four-hour sessions in which teams don't just learn, but actually develop products. We work from user stories, short sprints and direct application to their own cases, with the existing technology stack as the starting point. Copilot was central, complemented by building concrete AI agents and practical automations.

Employees define their own user stories: which concrete problem structurally costs time today? They then translate it within the same session into a working agent or automation, for example for document generation, summarisation, internal knowledge extraction or process support.

Automating is a deliberate investment. We guide participants in making their work process explicit: where is the repetition, where is the manual work, where is time lost? This way they don't just learn to use a tool, but invest in their own way of working. We work in short development cycles: define the problem, build the solution, test, improve. This lets teams keep developing new agents independently.