A few months ago, I sat in a meeting room in Melbourne where someone asked: "Should we just try it and see?"

Twenty minutes later, someone was building a pilot.

A year earlier, I had been in a similar meeting in Utrecht. Someone suggested an idea. Nobody said yes. Nobody said no.
Instead, we scheduled another meeting to align on the framework for evaluating the idea.
Nothing moved for three weeks.

I used to think one was progress and the other was bureaucracy.

After running an AI practice across both cities, I changed my mind: both are wrong.

But they are wrong in opposite directions.

One culture can spend too long discussing the perfect approach before learning anything. The other can move so quickly that nobody stops to ask whether they are solving the right problem.
AI makes both mistakes more visible.

Because AI rewards experimentation, but it punishes unexamined assumptions.

This talk explores what I learned building AI solutions across two very different engineering cultures: why some teams cannot get started, why others start too quickly, and what practices help both move faster without becoming reckless.

We will look at the role of regulation, governance and engineering discipline, and why the most effective teams create their own guardrails before someone else creates them for them.

You will leave with practical ways to recognise your organisation's AI failure mode and the habits that help turn experiments into reliable systems.

Everything in this talk comes from running New Orange's AI practice across Melbourne and the Netherlands: the same technology, the same questions, and very different people.

No code, some charts, several uncomfortable mirrors (I hope...).

Suitable for a mixed room; works as an afternoon or end-of-day slot.

Based on the Substack essay of the same name (September 2026). Figures (ABS, auDA, Eurostat, Startup Genome) are current as of September 2026 and will be refreshed before the talk.