What Separates the Exceptional 10%

90% of executives report no measurable impact from AI yet. The interesting question is what the other 10% are doing differently.

69% of firms now use AI. A new NBER working paper surveying nearly 6,000 executives across the US, UK, Germany, and Australia found that 89% of them can’t point to any productivity gain from it over the past three years.

That’s not a story about AI failing. Adoption is close to universal, and the same executives still expect a 1.4% productivity lift over the next three years. They believe in the technology. They just haven’t gotten it to pay off yet.

Here’s a rough back-of-envelope on that 1.4%. It doesn’t come from sprinkling AI across every team evenly. It comes from something closer to a 10% improvement on 14% of your cost structure, concentrated. Spread thin, AI barely moves a P&L. Impact comes from concentration.

Bain and OpenAI have a name for where most companies get stuck instead: the micro-productivity trap. Individuals get faster at tasks. The gains stall at the firm level because the workflow around them — the handoffs, the legacy systems, the tacit knowledge — was never redesigned for AI in the first place.

So what actually separates the exceptional 10%? Three things, from what I see:

They chase 1-10x use cases, not 10% efficiency plays — the ones that also compound into a real competitive advantage, not just a minor cost line item.

They redeploy the capacity AI frees up. Growth stops getting staffed linearly. Savings get reinvested, not quietly absorbed back into headcount.

They have an actual operating vision for where skills and roles go next. Successful AI implementation requires new skill sets while simultaneously lessening dependence on others. This requires organizational design and investment.

Almost every company can turn AI on. That was never the hard part.

It’s not a technology gap. It’s a reinvention gap.