Managers are fixing AI's mistakes more than anyone else, new survey finds
As companies adopt AI tools to increase output and offset work that can be automated, new data shows AI use is creating a heavier review and rework burden for managers and senior leaders.
Founder Reports, a business research platform, released its 2026 AI in the Workplace report, based on a survey of more than 2,000 American workers. Conducted in April 2026, the survey revealed that 57% of managers have had to fix or redo work from coworkers who relied too heavily on AI, while only 38% of individual contributors have had to do the same. And the percentage grows all the way up the organizational chart, as 63% of VPs and C-Suite executives have had to fix AI-generated work that didn’t meet their standards.
AI delivers on speed, but the impact of review and rework is often overlooked, as this article from Founder Reports reveals. Data shows that these factors should be considered, instead of only looking at the speed or output gains from AI adoption.
Why Managers Are Bearing the Burden
Eighty-nine percent of the survey respondents said they’ve used AI tools for work, and 61% use them regularly (38% daily and 23% weekly).
While AI offers speed and productivity, spotty accuracy can be the tradeoff. AI output needs human review and verification, but the burden of that work is not evenly distributed among all workers.
Managers own outcomes and responsibility for their team’s work. With AI increasing productivity and output, there’s more work to be reviewed. But two other factors also play major roles in pushing more of the rework toward managers:
1. Spotting mistakes in AI output often requires advanced expertise.
2. A lack of formal policies around AI use leaves many workers without clear guidance or guardrails.
Riken Shah, founder and CEO of OSP, a healthcare technology firm, says “the productivity gains are real, but they don't automatically come with quality.” Shah feels that his review work has changed due to AI. “I spend less time checking for completeness and more time interrogating the reasoning,” he explains. “Was the output validated? Did the person using the tool apply domain knowledge, or did they accept the model's framing wholesale?”
Part of the challenge in reviewing AI work is that it often looks correct and accurate on the surface, without the deep expertise to identify flaws. Kirill Meshyk, head of AI data collection at Unidata, explains that typical AI mistakes “can only be spotted by someone with a certain level of expertise in a particular field.”
“My team deals with structured data, so the most typical AI error we encounter is when it produces something correct-looking but false,” he says. “It’s well-formatted and sounds plausible, but the reviewer will only realize it is wrong when looking into details.”
AI Adoption Is Outpacing AI Governance
The survey showed that 44% of respondents say their company doesn’t have a clear AI policy, or they’re unsure if one exists. That means nearly half of American workers don’t have clear guidance on how they are expected to use AI for their job.
Despite the lack of governance, most companies are moving full speed ahead with AI adoption. McKinsey’s 2025 Superagency in the Workplace report showed that 92% of companies plan to increase investments in AI over the next three years, despite only 1% of business leaders labeling their company as “mature” in deploying AI.
Combined, the data from Founder Reports and McKinsey show that many companies are faster at adopting AI than they are at developing the systems to manage it. The cost of that gap falls on those responsible for reviewing the work.
Teresa Tran, COO of LaGrande Marketing, has seen firsthand the impact of effective policies and procedures. "The best thing I did was write down exactly what a finished piece of work looks like before anyone on my team uses AI to produce it," she says. At LaGrande, the solution was a one-page submission standard with specific requirements before reaching her inbox for review. The result, she says, was fewer revisions sent back, because team members knew what the finished work should look like. "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."
Executive coach and HR consultant Louise Spinks says she sees a predictable pattern when clear AI policies aren’t in place. "If the policy's too vague, people just make it up as they go," she says.
When policies don’t exist or are too vague, workers are left guessing at when and how they can use AI. Their assumptions and expectations of what good output looks like may not match what a manager expects to see. So without clear policies and procedures, managers have to review work more carefully, and the likelihood of revisions and rework increases.
Verification of AI Output Is Inconsistent
The time needed for fixing or improving AI output isn’t the only issue here. Even when the output meets expectations, simply reviewing it takes more time and effort due to increased scrutiny. From the AI in the Workplace survey, 77% of respondents said they review AI-assisted work from colleagues more carefully than fully human work, and 36% said they review it “much more carefully.”
Although skepticism exists, many admit to not thoroughly verifying facts or answers from AI tools. A recent Clear Spark Digital survey of 2,000-plus Americans found that only 17% always verify AI-generated information, and less than half (46%) ever check the sources provided by AI tools for their output.
One of the reasons why sources are rarely verified is because AI output carries a certain level of confidence, even when it’s wrong. Aimen Hallou, chief technology officer at Floxy, points out that “unlike a human, AI tools will not hesitate. They always state their answers as definitively as possible. To successfully review AI output, people need to possess even greater expertise in the relevant topic.”
With many details going unchecked by those using the AI tools and significant expertise needed to spot mistakes, most of the review work falls on managers.
Moving Forward
AI’s case for improved output is real, and it’s clear that AI will be a major part of most businesses moving forward. But data shows that currently the cost of incorporating AI is unevenly distributed, with managers bearing most of the burden.
Experts who have worked through these challenges point to clarity as the foundation of an effective AI usage policy. “An effective workplace policy has to explicitly state which data may and may not be fed into an AI tool,” Hallou says. Shah adds that review requirements must also be clear. “Be explicit about where AI output requires human verification, and make ownership of the final product clear regardless of how it was generated,” he says.
However, policies alone are not enough. Spinks points out that team members also need to know how to apply the policies. "Training matters as much as the policy itself," she says. "One can't assume people understand hallucinations, bias, data leakage or the difference between using AI as a drafting tool versus using it as a decision-making authority."
Without that training, the responsibility for catching errors continues to flow upward. As Meshyk points out, “AI doesn't remove your professional responsibility for what is done by the machine.”
Survey Methodology
The AI in the Workplace report is based on data from a survey of 2,078 U.S.-based workers conducted via Prolific.com in April 2026. Respondents were screened for U.S. residency and current employment. The sample includes full-time (80%) and part-time (20%) workers across all industries and job functions, with ages ranging from 22 to 82 (median 37). Employees from companies of all sizes are represented.
This story was produced by Founder Reports and reviewed and distributed by Stacker.
