Your Staff Say AI Makes Them Faster. Your Numbers Say Nothing Changed. Both Are True.
Workers report AI gains while nine in ten executives see no productivity impact. Here's where the saved time goes, and the four moves that make it show up.
Your staff say AI makes them faster. Your numbers say nothing changed. Both can be true, and two large 2026 studies show exactly how.
What the data actually says
Workers feel it. Gallup reported in 2026 (page updated September 30) that in May, 65% of U.S. employees in organizations that have implemented AI said it had a positive effect on their productivity and efficiency. Gallup's own summary: the benefits "appear to be concentrated at the level of individual tasks rather than in broader workplace systems."
Firms don't see it. "Firm Data on AI," an NBER working paper from February 2026 by economists including Nicholas Bloom and researchers at the Federal Reserve Bank of Atlanta and the Bank of England, surveyed nearly 6,000 senior executives in the US, UK, Germany and Australia. 69% of firms actively use AI. Executives who use it average only 1.5 hours a week. And nine in ten reported no impact of AI on employment or productivity at their own firm over the past three years. Looking ahead, they expect AI to lift productivity by an average of 1.4% over the next three years.
Put those side by side and you get the workplace AI productivity stall: plenty of use, real task-level speed, and almost nothing on the income statement. The Guardian reported on October 13 that some companies which rushed in are now pulling back, adding restrictions, bringing back experienced staff and increasing training.
Where the saved time goes
- Into review. A faster draft that needs careful checking moves work from one person to another. It doesn't remove it.
- Into more of the same. An hour saved on an email becomes an hour of more email. Nobody planned what the time was for.
- Into scattered experiments. Many people trying many tools, none of them owned or measured. We covered the same pattern in the MIT GenAI Divide report.
- Nowhere visible. If nobody measured the task before, nobody can prove it got cheaper after.
What separates the firms that see gains
Gallup's data points at two conditions most strongly associated with frequent AI use: workflow integration and manager support. Among employees in AI-adopting organizations, 78% of those who strongly agree their manager actively supports AI use it frequently, against 44% of those who don't. Yet only 36% of employees in AI-integrating organizations strongly agree that their manager does. Breadth matters too: 45% of employees using AI for one or two purposes reported a productivity gain, rising to 90% among those using it for seven or more.
None of that is about buying a better tool. It's management.
The operator's fix: four moves for a small firm
- Pick one task, not one tool. A task done every week that eats real hours. Write down what it costs now: hours times rate.
- Name an owner. One person who checks the output and reports the number.
- Decide where the hours go. Billable work, faster turnaround, or one less hire later. If you don't decide, email decides.
- Measure at 30 days. Cost per finished task, review time included. Keep it, fix it or switch it off.
Our AI ROI calculator runs the before-and-after arithmetic, and the 10 benefits of AI for business page shows which gains tend to reach the numbers and which stay as a feeling.
The bottom line
AI's productivity stall isn't proof that AI doesn't work. It's proof that individual speed doesn't add up to business results by itself. The firms that break the stall treat AI like any other operating change: one process, one owner, one number.
