Long before "generative AI" was a phrase anyone used in a board meeting, I spent a few years obsessed with a different question: why do most good ideas inside big organisations go nowhere? The vehicle for that obsession was a programme called EXPLORE, which I designed and led at ICI Pakistan. It engaged more than 900 people, generated over 200 validated ideas, and ended up delivering more than PKR 300 million in recurring annual impact.

I bring it up not for nostalgia, but because almost everything EXPLORE taught me about turning ideas into outcomes maps directly onto the challenge of getting AI adopted today. The technology is new. The human problem is not.

Before AI, there was EXPLORE

EXPLORE was an intrapreneurship programme, a structured way to surface ideas from across the organisation, test them, and back the best ones to delivery. The temptation with these things is to measure success by activity: how many ideas submitted, how many workshops run, how many people attended. We learned quickly that activity is a vanity metric. The only number that mattered was impact that survived to the P&L.

We didn't have an idea shortage. No organisation does. We had an adoption shortage. Almost everyone does.

Ideas are cheap; adoption is everything

This is the lesson I wish more leaders internalised before launching AI initiatives. The current moment has produced an avalanche of ideas, every team can name ten things AI could do for them. That abundance feels like progress. It isn't. The constraint was never the supply of ideas. It's the organisation's capacity to actually absorb change.

I see companies running AI ideation workshops and walking out with a hundred sticky notes, mistaking that for momentum. A hundred ideas that nobody owns is worth precisely nothing. One idea that a real person is accountable for, with a budget and a sponsor, is worth more than the other ninety-nine combined.

What actually moved the needle

Looking back at which EXPLORE ideas delivered and which evaporated, the pattern was clear, and it had nothing to do with how clever the idea was.

  1. Ownership beat brilliance. Ideas with a committed individual owner outperformed objectively better ideas that belonged to a committee. Diffuse ownership is where momentum goes to die.
  2. Sponsorship absorbed the friction. Every worthwhile change hits resistance, budget, politics, the early period where it's worse before it's better. A senior sponsor who'd take that friction on their shoulders was decisive. Without one, the first obstacle was usually the last.
  3. Small, real beat big, theoretical. A modest idea actually running in one business unit taught us more, and built more belief, than a grand idea trapped in a strategy document.
  4. Visible wins gave permission. The first deployment that demonstrably worked did something no presentation could: it gave everyone else permission to believe and to act.

The transferable insight

People don't adopt change because you've proven it's logical. They adopt it when someone they trust has gone first, taken the risk, and come out better. Your job as a leader isn't to win the argument. It's to engineer that first credible win.

Applying it to AI

So when a leadership team asks me how to drive AI adoption, my answer often surprises them, because it's barely about AI. Pick one high-value use case. Give it a single accountable owner. Find it a senior sponsor who'll absorb the early friction. Ship something small and real rather than planning something big and theoretical. Then make the win visible, so the rest of the organisation gets permission to follow.

That's not an AI strategy. It's a change strategy, and it works for AI for the same reason it worked for EXPLORE: technology doesn't adopt itself. People adopt it, and people are moved by ownership, sponsorship and proof, not by potential. The organisations that remember this will pull ahead. The ones chasing a hundred sticky notes will keep wondering why their AI ambitions never quite materialise.