You’ve probably seen it happen at your own company: someone names the AI agent, gives it a seat on the org chart and starts talking about it like a co-worker. It feels harmless, even a little fun. New research suggests it’s quietly undermining the one thing your business can’t afford to lose: accountability.
A large-scale study led by Boston University’s Emma Wiles found that managers reviewed work less carefully when they believed it came from an AI “employee” rather than an AI tool. The output was identical in every case. Only the label changed, and that was enough to shift how closely people checked the work.
For leaders racing to fold AI into daily operations, that’s a warning worth taking seriously.
Why the ‘AI Employee’ Trend Is Spreading So Fast
Giving software a human name isn’t new. Clippy, Siri and Alexa all did it. But the current wave goes further, placing AI agents directly onto org charts as if they were staff members with roles, titles and reporting lines.
It’s catching on faster than most leaders realize. In Wiles’s survey of 1,261 HR and finance managers, 31% said they already frame AI as a teammate or employee. Nearly 1 in 4, 23%, said their company lists AI agents on formal org or work charts.
The logic makes sense on the surface. If an AI agent can complete tasks as capably as a human, why not manage it like one? The answer, according to Wiles’s data, is that the framing changes far more than vocabulary.
What Happens When You Give AI a Name and a Seat at the Table
Wiles and her research team ran an experiment to find out. They gave 1,261 managers, directors and executives five documents, each containing planted errors, and 20 minutes to review as many as possible.
Participants were split into three groups. One was told the documents came from an unnamed AI tool. Another was told a human colleague named Alex made them. The third was told an AI team member named Alex-3 made them, with 813 responses ultimately counted as reliable.
Across the full sample, the differences were small. But among managers whose companies already treat AI as an “employee,” something changed sharply: They cut their monitoring intensity by 16% when reviewing the AI teammate’s work compared to the unnamed AI tool and leaned more heavily on other people to catch what they didn’t.
The accountability shift showed up in language too. One manager described an AI agent named Kevin this way: “Kevin made a mistake. Why did Kevin make a mistake?” As the researchers point out, the responsibility for that mistake belongs to the humans who deployed Kevin, not to Kevin.
The Real Cost: Weaker Oversight, Not Better Adoption
Here’s where it gets interesting for anyone hoping the “AI employee” framing might at least boost adoption. It didn’t. Wiles’s team found no evidence that naming AI agents or giving them org-chart status improved how well people actually used or integrated the tools.
So what does this mean for you? You get the downside of anthropomorphizing AI, reduced scrutiny and diffused accountability, without the upside you were hoping for.
The escalation data makes the stakes concrete. When an additional review was free, 98% of managers in the control group requested one. When that same review carried a cost, only 45% did. Framing and incentives compound each other, and both point toward the same failure mode: skipped checks on work that still needs a human’s judgment.
There’s a human cost too. As one manager in the study put it, treating AI like a co-worker can send an unintended message to your actual staff that they’re easily replaceable. That’s a morale problem layered on top of an oversight problem.
How to Keep Accountability Human as AI Takes On More Work
None of this means you should avoid agentic AI. It means you need to be deliberate about how you frame it and who owns the outcome. Start by separating language from liability.
Keep review protocols identical regardless of the source. Whether a document was drafted by a person, a chatbot or an “AI employee,” the review checklist and time allotted should not change.
Name a human owner for every AI workflow. Wiles’s research is clear that responsibility drifts toward the AI when the AI has a name. Counter that by explicitly assigning a person, not the agent, as accountable for the output.
Audit your own team’s language. If people at your company already talk about “Kevin” or “Alex-3” like co-workers, that’s a signal to reset expectations before it costs you an error you didn’t catch.
Make second reviews free, not costly. The 98% to 45% drop in escalation requests shows people respond to friction. Remove the friction and you remove an easy excuse to skip oversight.
The key is to treat agentic AI as a powerful tool with human ownership attached, not as a personality you can hold responsible. Try this approach for 30 days: strip the names, keep the review standards fixed and see whether error rates in AI-assisted work actually change.
As AI takes on more of the operational load in your business, the question isn’t whether to deploy it. It’s whether you’ve built a system where a human is always the one answering for what it produces.
If You’re Running a Lean Team, This Matters Even More
You might assume this is a big-company problem, something for enterprises with formal org charts and HR departments. It isn’t. If you’re a solopreneur or small-team founder leaning on AI tools to draft contracts, review client work or manage bookkeeping, the same bias can creep in without you noticing.
When you start referring to your AI assistant as “my team” or giving it a name, you’re the one who absorbs the accountability gap Wiles’s research describes. There’s no HR department to diffuse it to.
The fix scales down just as easily as it scales up. Keep your own review checklist identical whether you wrote a client deliverable yourself or an AI tool drafted the first pass. Treat the label as irrelevant to the level of scrutiny the work receives because the research shows your brain won’t do that automatically.
Featured image from PanuShot/Shutterstock
