Why do people feel much more productive with AI than the revenue or profit numbers actually show?

This explores why AI feels like a big productivity boost while the hard numbers show something smaller, and where the gains go missing. The corpus points to one main pattern: the speedup people notice is real, but it is local. It happens at one step of the work, and the rest of the work, the organization and the measuring tools don't change to match. A survey of 750 executives found that perceived gains run ahead of measured ones, partly because revenue lags behind operational improvements Do AI productivity gains feel larger than they actually measure?. McKinsey's numbers make the gap starker. 80 percent of respondents report personal productivity gains, but only 37 percent see enterprise profit improve. The small group that does see profit gains got there by redesigning workflows rather than by using better models Why do individual AI gains not translate to business results?.

One reason the feeling outruns the result is that AI often doesn't remove time from a task. It moves it. Time once spent doing the work goes into writing prompts, reading outputs and checking whether they're right Does AI really save time, or just change how we spend it?. That work feels different and often lighter, so it registers as savings even when the clock disagrees. Narayanan and Kapoor make the same point at the level of a whole job. AI shrinks the middle 'execute' layer of knowledge work, but deciding what to do and delivering the result stay the same size or grow Does AI really compress all layers of knowledge work equally?. You feel the drafting get faster. You don't feel the extra reviewing, coordinating and sign-off it creates downstream.

Gains also leak where humans and AI meet. In one experiment, when the AI got more accurate on a question, the people using it captured only about half of that improvement. Sometimes they did worse than the AI would have done on its own Why does assisted accuracy capture only half the LLM gain?. Context matters too. The gains show up when people apply skills they already have, and they disappear when people use AI to learn something new When does AI actually boost worker productivity?. If a speedup on familiar tasks gets generalized to all the work someone does, the total gain will look bigger than it is.

Here is the twist you might not expect: some of the gap may run the other way. In Anthropic's interview study, 86 to 97 percent of workers said AI saved them time, yet around 70 percent mentioned the stigma of using AI at work Why do workers hide productivity gains from AI use?. Saved time that stays hidden can't show up in company metrics. It may be absorbed quietly rather than turned into more output. Benchmarks have a similar problem. AI agents win contest-style tests but struggle with long, real occupational workflows, so the yardsticks behind many expectations measure contests rather than work Why do agent benchmarks not predict real economic value?.

The practical takeaway is that 'is AI making us productive?' is partly a question about where gains are allowed to land. Individual speedups don't add up to organizational results unless someone redesigns the workflow, the team norms and the metrics around them. That is why Microsoft argues the next frontier is collective rather than individual productivity Can AI boost how teams work together?. Microsoft offers this as a design mandate, though, without evidence yet that team-level gains appear.