About twelve years ago, Dan North shook up the software community with the claim that Agile doesn't scale. A decade later, we find ourselves facing a remarkably similar problem: AI doesn't scale either.

We've moved from deploying software a few times a year to deploying multiple times a day. Now AI promises another leap in productivity. Yet many organizations still struggle to translate that productivity into lasting customer value.

The latest manifestation comes in many forms: spec-driven development, AI-native development, agentic software engineering, and whatever the next certification body decides to call it. The terminology changes, but the underlying challenge remains the same.

Scaling technology is easy.
Speeding up coding is easy.
Scaling outcomes is hard.

And when the expected results fail to materialize, the blame inevitably lands with the engineers.

In this talk, I'll explore the uncomfortable questions that most AI vendors would rather avoid. Why do AI initiatives stall after the first project? Why do productivity gains in coding fail to translate into business outcomes? How do you scale success beyond a handful of AI enthusiasts and 10x engineers? And what actually creates sustainable competitive advantage?

The answer has far less to do with the latest AI model or Copilot feature than most people think. Long-term success depends on maintenance, craftsmanship, organizational design, and our ability to continuously discover and deliver customer value.

This talk is about what happens after the AI honeymoon ends. It's Monday morning, it's raining, and despite all the AI productivity gains, customers are still leaving.