Most workers feel faster with AI, but almost none of that speed shows up in company profits — why does the gain vanish?

This explores why most workers say AI makes them personally faster while far fewer companies see that speed reach the bottom line. The headline number comes from McKinsey: 80 percent of respondents report personal productivity gains, but only 37 percent see enterprise EBIT improve Why do individual AI gains not translate to business results?. MIT NANDA's finding is starker: 95 percent of enterprise generative AI pilots return no measurable value Why do most enterprise AI pilots fail to deliver returns?. The corpus suggests the problem isn't the models. Saved minutes don't turn into profit on their own. Something in the organization has to capture them.

The surprising part is where those saved minutes go. Anthropic's interview study found that most workers who save time with AI also hide their AI use because of workplace stigma Why do workers hide productivity gains from AI use?. If managers can't see the gain, they can't redesign the job around it, so the time often turns into slack instead of output. Seen this way, the gap between the individual and the enterprise is partly a visibility problem. Some gains are also smaller than they feel. Executives perceive larger gains than they can measure Do AI productivity gains feel larger than they actually measure?, and survey-based estimates put the current economy-wide gain at about 1 percent What is generative AI's actual aggregate productivity gain today?.

A second explanation is coordination. Agrawal, Gans and Goldfarb model what happens when AI makes each person better at their own task. Each person also starts deciding things a bit differently, and in a company where decisions depend on each other, those differences pile up into friction Does AI adoption require rethinking how organizations coordinate?. Ten people who are each faster don't make a faster team if their outputs no longer fit together. Microsoft Research makes a related argument: the next frontier is collective productivity, meaning tools built around shared goals rather than individual assistants. The report states this as a design goal, though, not a proven result Can AI boost how teams work together?.

A third explanation is about timing and measurement. The productivity J-curve says that firms adopting AI first spend heavily on things accounting doesn't capture well: retraining, rebuilt processes, new software. So productivity looks weak early on and only shows up later Why does AI productivity appear weak when investment is high?. Some of the margin gap may just be early. There are also limits on what kind of gain AI delivers. The boost appears when people apply skills they already have, and it disappears when they use AI to learn something new When does AI actually boost worker productivity?. Agents that win benchmarks also tend to fail at long, real occupational workflows Why do agent benchmarks not predict real economic value?. Both findings suggest that faster individual tasks don't add up to automated end-to-end work.

The practical lesson is in what the winners do. McKinsey's 6 percent of high performers redesign workflows, keep humans validating AI output, manage costs, plan their workforce and have senior leaders own the effort. MIT's 5 percent buy rather than build, give line managers the authority, and pick tools that adapt over time. Neither group wins with a better model. The corpus points to a conclusion many readers won't expect: the gap between individual and enterprise results is mostly a management problem, and part of it starts with employees who don't feel safe saying how much time AI saves them.