7.8%. That's how much AI lifted productivity across hundreds of engineers in one operator's measurement - tracked across three companies over real work quarters. Not a vendor study. Not a conference keynote. What actually showed up in the data when someone sat down and measured it.
The gap between that number and the "10x productivity" claims circulating in executive presentations explains a lot of the frustration showing up in workplaces right now.
The gains are real, but modest. 7.8% on a 40-hour week is roughly three extra productive hours. That's not nothing - compound it across a team and the economics can work. But it's a very different story from the "replace three people with one AI-augmented person" narrative being used to justify headcount reductions and mandatory adoption mandates.
Why the Peak Fades
The more troubling number: 66% of workers who hit a productivity peak saw those gains erode within the following quarter.
This happens for predictable reasons. AI tools are most effective on the mechanical parts of work - drafting, summarizing, searching, formatting. When you first adopt them, those tasks were already consuming significant time. Fast gains follow. But once the easy wins are captured, you're left with the hard parts of the job that AI still doesn't handle well: judgment calls, client relationships, navigating ambiguous requirements.
Skill regression is a related factor. Developers who let AI write all their boilerplate start losing fluency in the underlying language. Writers who outsource first drafts lose confidence in their own voice. At some point the tool substitutes for skill development rather than augmenting it, and performance plateaus or dips.
The Problem With Mandated Adoption
What makes this dynamic difficult is that the mandates are running ahead of the evidence. People are being pushed onto AI tools under threat of poor performance reviews or job loss, while the ROI case remains unproven even to the executives demanding adoption.
When a company mandates a tool that genuinely makes your job easier, adoption is organic. When they mandate one that adds friction, requires new skills, and delivers marginal returns - then cite aggregate productivity statistics that don't match anyone's individual experience - you get backlash.
The 10x claims aren't neutral background noise. They set expectations the actual tools can't meet for most use cases, which makes a 7.8% gain feel like failure by comparison.
Teams seeing durable improvements tend to be selective: AI handles the time-tax work - transcription, research compilation, first-pass drafts - while practitioners stay hands-on with judgment calls. Broad mandates consistently underperform targeted adoption.
The 7.8% is also an average across a wide distribution. Some people are measuring 25%, others near zero. Identifying which work categories drive those outlier gains is more useful than chasing the headline number.