Why Europe Should Learn from the UK's ‘Agile’ AI Regulation — Lessons for Germany and the EU AI Act

Sunday Reflection — 2026-06-18

How much regulation is too much? That’s the question I keep coming back to, now that the EU AI Act is in its teeth-baring phase and Brussels is betting hard on top-down compliance. Meanwhile, I watch the UK take a step back, pivot, and work sector by sector, making it up as they go. And honestly—that’s not a criticism. It might be the only sane response to the speed and unpredictability of AI’s progress.

What’s clear is that the EU is all in on order: fixed categories, mandatory risk assessments, exhaustive reporting, the bureaucratic answer to an engineer’s panic. Germany, especially, loves nothing more than harmonized frameworks and definitive answers. But AI isn’t humming along a fixed track—it’s exploding sideways and upward at once. The UK, unshackled from the Brussels spreadsheet, is taking cues from the world of Agile. They’re launching a policy, seeing what breaks, calling in stakeholders from healthcare, finance, and transport, then iterating. Less grand architecture, more practical patchwork. From my vantage point, it feels more honest—and more likely to keep up.

If you’ve run an IT project—especially in Germany—you know the drill. Years of requirements workshops. Sign-offs from every committee. By launch, your spec is already out of date. I’ve taught AI courses to professionals who arrive tense and leave frustrated: “How do we stay compliant when the rules are obsolete in six months?” My answer rarely satisfies anyone. But in the UK, I see policymakers acting more like software teams: launch a minimal viable rule, push it into the wild, collect feedback, patch, and repeat. You don’t get that from Brussels.

Here’s a concrete example. Last year, one of my clients—a German logistics company—struggled with integrating an AI-driven scheduling tool. The compliance demands were exhaustive, nested inside the EU’s mountain of regulation. Their UK competitors, by contrast, adopted a more nimble approach: they worked directly with the UK’s sectoral regulator, got pilot approval, and iteratively expanded the tool’s feature set as real-world results rolled in. The difference? The Germans spent a year locked in risk assessments; the Brits iterated product features in response to both user feedback and regulatory check-ins. Who won? The market decided, and it wasn’t Berlin.

The paradox is this: regulation must protect, but it also must adapt. The UK’s model—sector-specific pilots, feedback loops, empowered practitioners—mirrors what makes Agile work in software. It demands humility from regulators and places trust in sector experts, not just legislators. Germany and the EU risk smothering innovation in their bid to eliminate risk entirely. And in AI, where uncertainty is wired into the codebase, that’s a losing bet.

The immediate lesson: importing some UK-style Agile thinking could future-proof the EU’s regulatory project. We need to leave room for adaptation. Top-down edicts may feel safe, but they’re blunt tools for a world moving this quickly. Are we ready to govern AI, or just to slow it down? That’s the question Europe—and especially those of us on the ground here—can’t ignore any longer.