A growing number of organizations are giving AI agents human names, listing them on internal org charts, and telling employees to treat them as “teammates” or “peers.” A startup’s AI agent called “Scout” sits on the HR team org chart, reviewing job applications and conducting first-round interviews. An engineering manager jokes that “Kevin” — the coding agent — made a mistake in the latest pull request.
It sounds progressive. It sounds like the future of work. And according to new research published in the Harvard Business Review (May 2026), it is also largely counterproductive.
“If you want people to feel like they will lose their job to AI, or can be easily replaced by AI, then put it on the org chart.” — HBR Research Team
The Study: A Randomized Experiment with 1,261 Managers
The researchers — Matthew Kropp, Julie Bedard, Emma Wiles, Megan Hsu, and Lisa Krayer — conducted a large-scale randomized experiment with 1,261 managers, directors, and executives in HR and finance roles across the United States, Canada, and the European Union.
Participants reviewed workplace documents containing embedded errors: inconsistent information, factual inaccuracies, and incorrect calculations. The critical manipulation was how the original drafter was identified:
- Group A: The document was drafted by an “AI tool”
- Group B: The document was drafted by a human employee named Alex
- Group C: The document was drafted by an AI employee named “ALEX-3”
Only the framing changed — the content was identical. The results were eye-opening.
The Five Negative Effects of Humanizing AI Agents
1. Accountability Blurs
When the AI was framed as an “employee” rather than a “tool,” personal accountability fell by 9 percentage points while accountability attributed to the AI rose by 8 points. Managers began joking about “ALEX-3’s mistake” instead of owning the error themselves.
“The blame isn’t on a person; it’s on the technology.” — Study participant
This diffusion of responsibility creates a dangerous vacuum. If everyone assumes the AI is accountable, no one is. In regulated industries — healthcare, finance, legal — this is a liability nightmare waiting to happen.
2. Escalation Increases by 44%
Counterintuitively, treating AI as an employee doesn’t make managers more confident — it makes them less confident in their own judgment. Requests for additional review increased by 44% when the drafter was framed as an AI employee.
Rather than carefully checking the work themselves, managers passed it onward. This represents real economic cost: additional review cycles, delayed decisions, and wasted senior talent time.
3. Quality Control Declines by 18%
The most alarming finding: managers in the “AI employee” group caught 18% fewer errors than those in the “AI tool” group. They were more likely to miss inconsistent logic — a contract claiming cost reduction while a spreadsheet showed an increase, or an entry-level role requiring 10+ years of experience.
The researchers link this to a phenomenon they call “AI brain fry” — mental fatigue from excessive oversight pressure. Those experiencing it scored 11% higher on minor errors and a staggering 39% higher on major errors.
4. Professional Identity and Trust Erode
Anthropomorphizing AI had measurable psychological effects on the human reviewers:
- 13% more likely to report uncertainty about their professional identity
- 7% higher concern about job security
- 10% lower trust in how AI will be used by their organization
“It makes you wonder what your role is supposed to be when AI is treated like a colleague.” — Study participant
5. Adoption Does Not Increase — At All
Here is the kicker: there was no significant difference in adoption intent between the AI employee framing and the AI tool framing. The primary obstacle to capturing AI value is getting people to actually use it — and humanizing agents does nothing to solve that.
What does drive adoption? Managerial encouragement and role-modeling. Companies leading in AI maturity are 3.5 times more likely to have managers who actively model AI use, regardless of how they frame the technology.
The Real-World Counterexample: “Scout”
The researchers profile a real AI agent called “Scout” — an AI formally listed on an HR team’s org chart, tasked with reviewing applications, conducting first-round interviews, and putting forward candidates. The HR leader described Scout as:
“It’s technically an equivalent peer on your team. That’s how it acts and behaves.”
While this may feel like forward-thinking integration, the research suggests it creates exactly the accountability and quality problems described above — without meaningfully increasing AI adoption.
Five Recommendations for Redesigning Work with AI Agents
The HBR team doesn’t just diagnose problems — they offer a concrete framework for organizations navigating the agentic transition:
1. Redefine Workflows and Human Roles Explicitly
Create clear spheres of accountability and spans of control. Redesign job descriptions to include oversight responsibility. Reset performance management to reward quality of oversight and orchestration, not just speed.
2. Make Accountability Explicit and Personal
Treat AI agents as software automation requiring clear human accountability, not as semi-autonomous colleagues. Define three governance elements:
- Decision rights — what agents can do autonomously vs. requiring human approval
- Escalation triggers — who intervenes, when, and at what cost
- Consequences — accountability carried forward to the human responsible for monitoring agent performance
3. Build Agent Management Capabilities
Teams need new training: when to rely on AI, when to challenge its outputs, and where its limitations differ from human judgment. Employees must understand the full range of tasks AI can perform — as intelligence augmentation, not just task automation.
4. Don’t Constrain Agents into 1-for-1 Roles
AI has no human limits. A single agent can operate across many workflows simultaneously. Multiple agents can reshape one job. Framing agents as human equivalents reinforces a delegation mindset that underestimates their actual capabilities.
“You’re hiring a tool, not a teammate — and that distinction matters more than you think.” — HBR Research Team
5. Lead by Demonstration, Not Declaration
Adoption follows manager behavior, not org chart design. Leaders should actively use AI in their own workflows, share what they learn, and normalize the idea that AI is a tool to amplify human judgment — not replace it.
What This Means for the Agentic Enterprise
The HBR study arrives at a critical moment. As our coverage of the State of Agent Engineering 2026 confirms, 57% of organizations now have agents in production. The question is no longer if agents will reshape the workplace but how organizations manage the human side of that transition. This organizational dimension is also what the DigitalOcean Currents survey reveals: 67% see value but only 10% scale, in part because teams don’t know how to integrate agents into existing workflows.
The research offers a counterintuitive but evidence-backed answer: the best way to integrate AI agents is not to pretend they are human. The most effective organizations will treat agents as powerful, specialized tools — and invest their cultural energy in training humans to manage, challenge, and orchestrate them effectively.
The org chart is for people. The agent dashboard is for AI. Keep them separate, and both will perform better.
Source: Harvard Business Review — “Research: Why You Shouldn’t Treat AI Agents Like Employees”, May 2026, by Matthew Kropp, Julie Bedard, Emma Wiles, Megan Hsu, and Lisa Krayer