AI ROI: Why Most Firms Still Can't Measure Returns

What happened

Most organisations still cannot measure the financial return on their artificial intelligence investments, but new ways of calculating return on investment are emerging, according to reporting by Computerworld. Michael Chui, a senior fellow at McKinsey, said there is a delay between adopting AI technology and capturing value from it, and that executives are increasingly pressing for evidence of returns.

The clearest data comes from McKinsey's "State of AI" survey, released in August. About 80% of respondents said AI had improved their productivity, but only 37% said its impact showed up in profits, a share broadly unchanged from the previous year. Just 6% said AI had delivered significant value, accounting for at least 5% of operating profit.

Chui described the result as a drop-off between the value individual workers gain from AI and the value organisations capture at enterprise level.

Why it matters

The gap suggests that AI's benefits are reaching day-to-day work faster than they are reaching financial results. For leaders in digital transformation, tool adoption alone does not produce returns. Value depends on how workflows, decision rights and performance measures are redesigned around the technology, and organisations that cannot link AI use to outcomes will find further investment harder to justify.

For service design, the implications are equally direct. Productivity gains that never reach customers or the bottom line are difficult to sustain under budget scrutiny, and finance and technology functions are now being asked to agree on how value is defined and tracked.

By the numbers

• 80% of respondents in McKinsey's State of AI survey said AI improved their productivity.
• 37% said AI's impact showed up in profits, about the same as the previous year.
• 6% said AI delivered significant value, accounting for at least 5% of operating profit.

The Renascence take

The headline reads as progress, but the survey is really measuring a handoff problem. Individual employees feel faster, while many organisations have not redesigned the processes that would turn that speed into margin or better service.

Productivity that lives inside individual workflows is not yet a business outcome, and it becomes a hidden cost until someone redesigns the process around it. Customer-obsessed operators should stop asking whether staff like their AI tools and start mapping where a saved minute actually lands, whether in faster resolution, fewer errors or a better commercial decision. Treat ROI as a service-design problem, with named owners for each journey step where value should appear, and measure those journeys rather than licence counts. The 6% figure is less a verdict on the technology than a signal that most organisations have not yet designed for value capture.