Strategy / Jun 13, 2026 / 8 min
The AI ROI Debate Is Asking the Wrong Question
Executives keep asking whether AI has ROI. The better question is whether the organization has changed enough for ROI to be captured.
The market is impatient with AI pilots that promise transformation and produce scattered productivity anecdotes. That impatience is healthy. But the question 'does AI have ROI?' is too blunt. It assumes value should appear automatically once a capable tool enters the building.
Most AI value is trapped behind organizational decisions. Will the process change? Will managers accept different roles? Will compliance approve new decision rights? Will IT expose the right data? Will leadership retire the manual report if the AI workflow works? ROI depends on those decisions as much as on model quality.
This is why individual productivity can feel real while the P&L barely moves. A worker saves two hours, then fills the space with more meetings, more drafts, more coordination, or more work that no customer values. The company captures activity, not economics.
The measurement system has to become more surgical. Pick a workflow. Establish the baseline. Define the economic mechanism: shorter cycle time, fewer errors, higher conversion, reduced backlog, lower risk, better retention, more capacity. Then change the process enough that the gain cannot evaporate into organizational noise.
The winners of the AI ROI debate will not be the companies with the broadest tool access. They will be the ones that make value legible, assign owners, retire obsolete work, and turn productivity into an operating result.