Case study · AWS Berlin · 9 months

Nobody was asked to join.

Every session was optional, inside working hours, and no individual’s AI usage was ever tracked. This is what happened over nine months — and the method WonderLead now runs with companies.

Optional is not the weak version. Optional is why it spread.

1 → 7Teams, then three organisations
9 moOne floor in Berlin
0Usage dashboards in phase one
How optional AI practice spread from one organisation to three at AWS Berlin over nine months.

Party Rock, Console Platform, UX. Learners became the teachers. Nobody was mandated in.

The situation

A director was enthusiastic about AI. There were Lunch and Learns. There were Slack posts. Inside the teams, almost no one engaged.

Seven pizza teams across Berlin, Amsterdam and the UK. Licences were not the problem. People were hesitant, some went quiet, others opted out. The tools were there. The practice was not.

The first attempt, which failed

I built a multi-agent scrum system and got resistance. I had led with the technology — the same mistake as everyone above me.

That failure is the point. If you start with a demo of what the model can do, you get applause and then nothing. People will not learn something they are afraid to be bad at in public, and a dashboard tracking their usage only makes it worse.

What changed

Bottom-up sessions. Optional. In working hours. No usage tracking. Deliberately no metrics in phase one.

We started where adoption actually happens: in one person’s daily habits, and in a team’s shared sense that it is safe to be a learner. Every role was in the room — engineering, design, product — on real work, not fictional exercises.

What happened

One team to seven. Seven to three organisations sharing a floor in Berlin. Different products, same problem. Nine months.

People who had never touched AI were building on-call agents and design-system automation, and demoing them. Learners became the teachers. A cross-org guardrails working group formed because people wanted one, not because a slide said they should.

How it was measured

Behaviour first, outcomes after. We did not count seats. We watched who showed up, who asked, who demoed, and who taught.

Behaviour over nine months

Voluntary attendance
High
Questions in the open
High
Peer demos
Rising
Orgs opted in
3

Qualitative, on purpose. Phase one had no usage dashboard. The numbers that mattered were people choosing to come back, asking in public, demoing work, neighbouring teams asking to join.

What it ended with

The findings were handed to AWS’s senior principal for AI literacy, who agreed to apply the approach. On the last day, the director thanked me for the method.

WonderLead exists to bring that method to other companies: a practical way to get real people to change how they work, drawn from having actually done it.

The people who lived it

What they said, after the looking glass.

“

Her ability to connect people, encourage curiosity, and create psychological safety around learning new technologies had real impact on AI literacy and adoption.

Olga Madejska
Head of AWS Design Systems
“

She single-handedly created an environment where we could safely explore, experiment, and adopt cutting-edge technologies — and actively helped her peers understand how to integrate AI into their daily workflows.

Kamil Bladoszewski
SDE II, AWS
“

She led a large AI learning series across multiple teams — well organised, with regular guests, demos, and practical examples. It created a useful space for people to learn, ask questions, and see what others were building.

Michael Dowse
Senior SDE, AWS · Cloudscape
✦  Run the same sequence  ✦

Before you book anything, take the checklist.

The ten moves, in the order you make them, plus the spread story and the measurement graph. Work email. The pack arrives, then you decide.