The 95% no one wants to own

In 2025 MIT published perhaps the most uncomfortable study on the topic: 95 percent of enterprise GenAI pilots show no measurable P&L impact.

Ninety-five percent.

And the truly interesting thing about that number is not its size. It's the reason. MIT's finding is that the failure isn't the model.

It's the approach.

On Friday I showed the pattern in ten frames — MIT, Deloitte, McKinsey, BCG, PwC, Harvard, six independent sources pointing the same way. Today I want to start with the first of the four dimensions behind it: Intent.

Because there's a second place where you can read the same gap, and it isn't a study. It's the capital market.

In its Q2 numbers, SAP reported a record current cloud backlog — 22.9 billion euros, up 27 percent. Contractually committed future revenue. And the share is still trading below its January high.

Read those two facts together. The market clearly doesn't doubt that AI software gets sold. It doubts that it gets realised. Between the signature and the effect sits the missing half — and that gap is exactly what the 95 percent measure, seen from the other side.

Over 30 years of global transformations — 23 of them as a Senior Partner at EY — I've seen the same regularity again and again. The technology was the solvable problem. The human half decided the outcome.

That's the half no one books, no one budgets for, no one addresses first.

But without it, the rest doesn't work.

Twelve weeks. Today: Intent.

To open: 💬 what was your most expensive experience with the human half being underestimated?

I read every comment.