The fix for AI's ROI problem was never better AI - it was better measurement

A new CIO analysis reframes the whole 2025 panic over AI returns. The problem, it argues, was rarely the technology; it was that projects were built in a way that made success impossible to prove. Start a deployment against a workflow with no trusted metric, and you can neither confirm value nor defend the budget when questions come.

The piece points to a shift already underway: enterprises are steering AI money toward use cases with clean data and known baselines, and Futurum’s survey of enterprise IT leaders shows hard financial measures overtaking soft productivity claims as the yardstick buyers use. The suggested next step is a pre-agreed formula that converts a KPI movement into an actual dollar figure.

That idea of translating the metric into money is the heart of what Uprovd does. We insist the outcome and its proof metric are defined before spend begins, so the return is measurable by design rather than reconstructed under pressure later.