You feel way more productive with AI — but when researchers actually measured it, several studies found the opposite was true.

This explores the gap between how productive AI makes people feel and what measurement actually finds. The short answer from the corpus: the feeling is real, but it reliably runs ahead of the numbers. The clearest case is a randomized trial with experienced open-source developers. Before starting, they predicted AI would make them 24% faster. In fact it made them 19% slower, and they still believed afterwards that it had helped. Outside experts in economics and machine learning made the same overestimate Do AI coding tools actually speed up experienced developers?. Executives show a similar pattern at company scale: the gains they perceive are larger than the gains they can measure, partly because revenue lags behind day-to-day operational improvements Do AI productivity gains feel larger than they actually measure?.

Where does the difference go? Mostly into cleanup. In a Workday survey, 85% of users said AI saves them 1 to 7 hours a week, but nearly 40% of that time went back into fixing errors and checking outputs. Only 14% consistently came out ahead Where does AI's time savings actually go in practice?. A Zapier survey found the same thing: 92% reported a productivity boost while spending about 4.5 hours a week revising AI work. Oddly, the heaviest and best-trained users reported both the biggest gains and the most cleanup How much time do workers really spend fixing AI mistakes?. A deeper explanation is that AI doesn't remove time so much as move it. Hours shift from doing the task to writing prompts and working out whether the output is right, and that shift is easy to feel as "faster" even when the total doesn't drop Does AI really save time, or just change how we spend it?.

A second gap sits between individual gains and gains for the business. McKinsey found 80% of respondents report personal productivity gains, but only 37% see enterprise profits rise. The small group of companies that do capture value redesign workflows and build human checking into AI processes, rather than just handing out tools Why do individual AI gains not translate to business results?. Narayanan and Kapoor offer a structural reason: AI speeds up the middle "execute" part of knowledge work, while deciding what to do and delivering the result stay the same or grow. Faster drafting doesn't translate one-for-one into faster finished work Does AI really compress all layers of knowledge work equally?. Benchmarks contribute to the inflated picture too. AI agents win contest-style tests but struggle with long, real occupational workflows, so the headline scores overstate readiness for actual work Why do agent benchmarks not predict real economic value?.

One twist runs the other way. Some real gains may be hidden rather than exaggerated. In Anthropic's interview study, most professionals and nearly all creatives said AI saved them time, yet about 70% mentioned the stigma of using AI at work Why do workers hide productivity gains from AI use?. So the honest picture is not simply "overhyped." Self-reports overstate net time saved, organizations struggle to turn individual speed into results, and some genuine gains are kept out of view. The corpus doesn't give a single percentage for how much is real. What it shows is that the answer depends less on the model than on whether anyone counts the verification work and redesigns the job around it.