The Manager’s New Job: Quality Control for AI

AI tools are widely credited with speeding up work and increasing output, and rightfully so. But a growing body of data suggests that productivity gains are also creating a new burden for the people reviewing that work.

Managers, directors, and senior leaders are increasingly spending time checking and fixing AI-generated output from their teams. According to Founder Reports’ AI in the Workplace Report, 57% of managers and above have had to fix or redo a coworker’s work that relied too heavily on AI. Among individual contributors, that figure is just 38%.

This points to something most organizations haven’t fully reckoned with yet. AI may be changing what it means to manage a team.

Quality Control is Falling on Leaders

The data shows the burden of reviewing and redoing work that relied too heavily on AI clearly correlates with seniority. 53% of managers have had to clean up AI-reliant work. At the senior manager level, it jumps to 65%. Directors come in at 61%, VPs at 63%, and C-suite executives at 63%.

Interestingly, leaders are also the most likely to use AI tools regularly. C-suite executives (62%) and VPs (63%) are among the most frequent daily AI users in the survey. They’re using AI themselves and still finding that their teams’ AI output regularly needs fixing.

A separate global study from Workday, which surveyed 3,200 workers in late 2025, revealed that roughly 37% of the time saved through AI is consumed by rework. For every 10 hours of efficiency gained, nearly four are lost to correcting, clarifying, or rewriting AI-generated content.

Individual Productivity Does Not Equal Organizational Productivity

These findings help explain a disconnect that’s been showing up in broader workforce research. Individual workers report that AI makes them more productive, but organizations aren’t seeing those gains translate into measurable improvements at the company level.

Gallup‘s February 2026 survey of 23,717 U.S. employees found that workers who use AI frequently report personal productivity gains. However, only about 1 in 10 employees in AI-adopting organizations strongly agree that AI has actually transformed how work gets done at their company. Gallup noted that this tracks with firm-level studies across the U.S., U.K., Germany, and Australia, which show minimal aggregate productivity impact from AI so far.

The data from the Founder Reports survey offers an explanation for this gap. AI saves time for the person using it, but that time savings is partially offset by additional review and rework from others, particularly managers. 77% of workers in the survey say they review a coworker’s AI-assisted work more carefully when they know AI was used, with 36% reviewing it “much more carefully.” Even among daily AI users, 80% apply extra scrutiny to a coworker’s AI-influenced output.

In other words, the net productivity equation is more complicated than many organizations are accounting for.

Why Managers Are the Ones Doing the Fixing

Individual contributors generate output. Managers review, approve, and ultimately own that output. When AI enables individual contributors to produce work faster but the quality isn’t consistent, the review step becomes harder or more time-consuming.

Michael Maximoff, Co-Founder & Chief Growth Officer at Belkins, found that the nature of managerial review has shifted and become more complicated due to AI. “Before, managers were mostly reviewing execution,” he said. “Now we spend much more time reviewing whether the person actually understands the customer, the business problem, and why the message would matter to that specific account.”

And there’s little reason to expect this dynamic to ease on its own. The Connext Global 2026 AI Oversight Report found that only 17% of U.S. adults consider workplace AI reliable without human oversight. An additional 35% said reliability requires “AI plus light review,” and another 35% said it requires “AI plus dedicated oversight.” Nearly two-thirds of respondents predicted that human review of AI output will increase in the coming years.

Which Functions Are Feeling It Most

The AI in the Workplace report also breaks down rework rates by department. Marketing professionals report the highest rate of regular AI rework at 16%, followed by data analysis (13%), project management (12%), and software engineering (11%).

These are fields where AI tools see heavy use and where the consequences of errors are immediate. Someone upstream, usually a manager, is the last line of defense before that work goes out the door.

What Organizations Can Do About It

The rework burden on managers isn’t inevitable, but addressing it requires organizations to think about AI adoption differently.

The first step is accounting for review time in AI productivity estimates. If a company is measuring AI’s value only by how much time it saves the person using the tool, it’s capturing an incomplete picture. The downstream cost of review, editing, and correction must be part of the calculation. Right now, most organizations aren’t measuring that at all.

The second step is setting explicit quality standards for AI-assisted work. The Founder Reports survey showed that 45% of workers have had to fix a coworker’s AI-generated output. Without clear expectations for what “finished” looks like when AI is involved, quality is left to individual judgment.

Teresa Tran, Chief Operating Officer at LaGrande Marketing, says the best way to reduce rework is to set a clear example of what completed work looks like. At LaGrande, they established a clear standard with examples that team members can reference. “Managers who skip that step end up doing the hard work twice, once when they review and again when they send it back for a full rewrite.”

The third step is building review into the process before the output is passed on to management. This starts with personal ownership of work, even when AI is used.

Christian Espinosa, Founder and CEO of Blue Goat Cyber, a medical device cybersecurity company, points out that standards should be set before tools are deployed, not after problems surface. He believes that teams handling this well are the ones where “the manager made it clear from day one that submitting AI output without personal verification is the same as submitting work you didn’t actually review.”

The Bigger Picture

The promise of AI in the workplace has always centered on doing more with less. At the individual level, that’s happening. Workers are producing more output, faster, with the help of AI tools.

But the organizational picture is more complicated. Managers and senior leaders are absorbing a quality control function that barely existed two years ago, and most companies haven’t formally acknowledged it, resourced it, or adjusted expectations around it. As AI-generated output continues to grow in volume across every department, the question for organizations isn’t just whether their employees are using AI. It’s whether the people reviewing that work have the capacity to keep up.