I sat in a strategy review last quarter where someone proudly presented an AI-powered dashboard that saved the team "hundreds of hours." When I asked what the team did with those hours, the room went quiet.
That's the problem Wharton just put numbers on.
Their 2025 AI Adoption Report found the average enterprise gets $3.70 back per dollar invested in AI. Top performers? $10.30. But the stat that should keep boards up at night: 80%+ report no measurable impact on EBIT.
The gap isn't about models or compute. It's about what happens around the model.
The orgs I've seen win are doing something different:
They rebuilt processes around AI capabilities instead of welding AI onto existing workflows. One company I advise killed three approval layers and redesigned the entire intake process. Not because AI suggested it — because AI made the old process obviously absurd.
They stopped tracking model accuracy and started tracking revenue impact. I've never seen a CEO get excited about F1 scores. I've seen them get excited about 12% less churn.
And honestly? They invested in the stuff nobody wants to present at the all-hands. Data quality. Governance. Change management. Boring work that makes everything else possible.
Most GenAI spending right now is sophisticated prototyping dressed up as transformation.
The companies winning aren't spending more. They're spending on different things entirely.
What's the actual barrier to AI ROI in your org — the real one, not the one in the board deck?