It’s 10 a.m. and Maria asks an AI tool to draft a summary. The answer comes back fast but not quite right, so she pastes the same request into a second tool to see if it does better, then a third, because the second one was missing context the first one had. Thirty minutes disappear before she starts the work she actually sat down to do.
This isn’t unusual. Glean’s Work AI Institute surveyed 6,000 full-time digital workers across the U.S., U.K. and Australia for its 2026 Work AI Index and found that 77% of AI users juggle multiple AI tools every week, a third use four or more, and 60% have rerun the same prompt across different tools because the first result wasn’t good enough. Employees say AI saves them roughly 11 hours a week more, than a quarter of a standard work week, which sounds like a real win until you learn that only 13% of those same workers think AI has meaningfully improved how their organization actually performs.
Somewhere between “I got 11 hours back” and “my company isn’t better off,” the gains leak out. Researchers have a name for where they go: botsitting. In other words, the unglamorous work of feeding AI context, supervising its output, debugging its mistakes, and cleaning up after it. It happens not after hours but during the job itself, one re-run prompt at a time period.
Saved Hours Are Spent Managing AI, Not Doing the Job
According to the report, workers spend 6.4 hours a week botsitting, 37% of all the time they spend with AI tools, which is more than the 36% of AI time spent on productive output. In short, employees now spend more time managing their AI than being helped by it.
Maria’s third tool, the one missing context, is typical. 53% of workers say the information they actually need isn’t accessible from the AI tools they are given, which leaves them serving as the connective tissue between systems that were never built to talk to each other, during the same eight hours they were supposed to be more productive.
AI didn’t create the outdated policy still floating around, or the guidance two departments swear by that contradict each other, or the answer that only exists in one manager’s head. It created what it could with the context it was given. Handing a workforce the fastest AI on the market and asking it to run on institutional knowledge nobody’s kept current is like handing someone the keys to a race car and sending them down a road full of potholes. The car doesn’t get them there any faster because it hits every pothole harder.
But there’s a second layer underneath the information gap, and it’s the one HR is actually responsible for: trust. When employees don’t have a clear enough mandate or worry that a single AI-assisted mistake could cost them their job, they stop acting like confident editors and start acting like nervous chaperones, checking and re-checking every output because being wrong feels more costly than being slow. If you give people better information but unclear stakes, they’ll still hover over every answer, and if you fix the tools without fixing that, all you’ve done is give people more reason to be anxious.
The Heaviest Botsitters are also the Most Likely to Quit
The report ties this directly to whether people stay in their roles. Workers who spend 40% or more of their AI time botsitting, the heaviest cases, are 73% more likely than their peers to be actively looking for another job, and those stuck with AI tools that lack context are far more likely to feel worn out by AI (50%, compared with 18% of workers whose tools are better supplied) and to be cleaning up after AI at least weekly (35%, compared with 24%).
This is a retention problem wearing the mask of an adoption success story: the employees doing the most invisible labor to keep AI useful are also the ones most likely to be quietly job hunting.
There’s also another risk here: Nearly seven in ten AI users, 69%, admit sometimes shipping AI-generated work they haven’t fully verified. In other words, sharing or publishing output they couldn’t confidently explain or defend if someone asked. 28% admitted to blaming AI for a mistake that was actually theirs. Not only is this a problem for compliance to solve, it’s an accountability problem showing up inside teams long before it becomes a headline.
Companies Avoiding this Drain Train People, Not Just Tool
The most useful part of the report isn’t the warning though, it’s the insights from the handful of organizations that aren’t experiencing this drain, and what sets them apart has nothing to do with which AI tool they bought. In what the researchers call “transformative” organizations, 90% of employees say their employer provides enough AI training and support, compared with 52% everywhere else. 84% say their employer formally recognizes AI skills, compared with 48%, and 90% say their employer treats AI as a real reason to redesign how work gets done, compared with 54%.
These companies aren’t using AI less, they’re training people properly, building context and verification into their workflows instead of leaving employees to supply it tool by tool. They’re also treating AI adoption as a reason to rethink how work is structured rather than just a reason to hand out more licenses.
Building that context is, in practice, a job internal communications and people teams are already positioned to own, since they’re usually the ones holding the single, current version of what’s actually true inside the company. Handing AI a clean, maintained answer instead of leaving employees to reconstruct one from memory and five open tabs is the difference between an assistant and one more thing to supervise.
The real test isn’t how many hours a rollout claims to save, it’s what happens to those hours afterward. If five free hours just get refilled with double-checking and cleanup, the rollout hasn’t worked, whatever the usage dashboard says. For HR and internal communications leaders, that’s a more useful metric to track than adoption rate: how much of an employee’s week now goes toward managing AI rather than being helped by it. Most organizations don’t track that yet and given how directly it connects to burnout and attrition, it’s worth finding out before it shows up in an exit interview instead of a survey.
Eva Spatz is VP, Head of People Experience at Staffbase. She writes and speaks regularly on the intersection of AI adoption, internal communication, and employee retention.