What Twenty-Five Years in Transformation Taught Me About AI

I've delivered transformation programmes for 25 years, across sectors that don't forgive getting it wrong: banking, insurance, wealth management, telecommunications, and more recently film and media. Here's the one lesson from that quarter-century that applies directly to AI, right now, whether or not anyone wants to hear it.

The technology was almost never the hard part.

In 2003, early in my career, I sat in a room where a sign-off took four signatures and eleven days. Everyone in that room agreed it was ridiculous. It took another six years to actually change it. Not because nobody had the idea sooner. Because nobody had been given explicit permission to say, out loud, "we're stopping this."

That's not a technology story. That's an organisation story. And it is exactly what's playing out with AI right now, just faster and considerably louder.

Three things I look for before I trust any AI transformation plan, learned the expensive way over 25 years rather than in the last two: has anyone named who has permission to kill the old process? Does the operating model actually change, or just the tool sitting on top of it? And is there a plan for the accountability gap AI creates, not just the capability gap it fills?

The people building frontier AI models are extraordinary at the model. Most of them have never had to explain a failed rollout to a regulator, or sit in a room and tell four hundred people their roles are changing. That gap, between what's technically possible and what an organisation will actually let happen, is not a footnote to the AI story. It is the entire job.

I didn't learn that from an AI conference. I learned it from 25 years of rooms like the one in 2003. That's the perspective I bring to every engagement, every keynote, and everything I write here.