Three years after ChatGPT, individuals are using AI tools constantly and capturing real productivity gains. Enterprise AI initiatives, by and large, are not. The explanations on offer - bad use cases, immature technology, poor change management - don't survive contact with the evidence. So what's actually happening?

The pattern is older than AI. It's repeated across the last 250 years, with steam, electricity, computing, and the internet. Each time, organisations divided into two camps based on a behavioural posture toward the new technology - and the camp that captured compounding value wasn't the one that invested most aggressively. It was the one that did something specific and counterintuitive that the other camp didn't.

AI is producing the same divide, faster and more sharply. This talk explains the pattern, identifies what distinguishes the organisations capturing value from those stuck in disappointing pilots, and makes the case for a different question: not "where can AI replace people?" but "where can AI multiply them?" That shift sounds simple. The historical evidence suggests it isn't, and that the organisations that get it right will widen the gap on the ones that don't.

Whether you're committing budget to your first AI initiative, defending your third, or trying to explain to your board why the returns haven't matched the investment, this talk offers a frame for understanding what's actually happening - and what to do about it.