# The Agent Got A Job Description. You Didn't. > Agents are getting seats on the org chart while the humans beside them take less responsibility, catch fewer errors and lose track of their role. - Source: https://www.aicapability.academy/insights/the-agent-got-a-job-description-you-didnt - Author: Marco van Hout, AI Capability Academy (Demadera) - Published: 2026-10-09 - Type: article Somewhere in your organisation an AI agent now has a name, a job title and a box on the org chart. The person who works next to it still has the job description they had before it arrived. One of those two roles was written this year. It was not the human one. ## The org chart is a statement of ownership An org chart is not decoration. It says who owns what. Put a box on it and you have told the whole organisation that whatever sits in that box is responsible for something. So what happens to the person in the box next to it? In May, [Harvard Business Review published a study of 1,261 HR and finance managers](https://hbr.org/2026/05/research-why-you-shouldnt-treat-ai-agents-like-employees) across the US, Canada and the EU that tested what the framing does. [Pat Brans summarised the result in Computerworld](https://www.computerworld.com/article/4213117/govern-ai-agents-like-workers-just-dont-pretend-theyre-human.html) last month: when AI was framed as an employee rather than a tool, experienced managers "assigned more responsibility to the AI. They also escalated more problems and caught fewer errors." [Irving Wladawsky-Berger put numbers on it](https://blog.irvingwb.com/blog/2026/09/should-organizations-treat-ai-agents-like-employees.html) this week. Personal accountability fell 9 percentage points. Managers identified 18% fewer errors. And they were 13% more likely to express uncertainty about their professional identity. This is not a fringe habit. In the same survey, 31% of managers said their company already framed AI as a teammate or employee, and 23% said agents appeared on organisational or work charts. Nearly a quarter of the managers surveyed work somewhere that has put the agent on the chart. The survey does not count how many rewrote the role of the person standing beside it. Ask around your own building. ## Ownership leaked, and nobody wrote it down Here is what I think happened. The agent's role was written with care: scope, permissions, a name. The human's role was left as it was. Leaving it as it was did not keep it the same. It made it vaguer. Before the agent, the manager owned the output. After the agent, the manager owns some part of it, and nobody said which part. People fill a gap like that the way people do: they assume the box next to them took over the rest. That assumption is the 9 points. It is the 18% of errors that nobody caught. The 13% is the number I would watch. Managers were 13% more likely to be unsure what their professional identity was. Read that as a person noticing their role changed and nobody telling them how. It is not a skills gap, and a prompt course will not close it. It is a design gap. As Wladawsky-Berger puts it, "how work is divided between humans and AI is a design choice". Somebody made that choice for the agent and skipped it for the human. The people building these agents say the same thing from the other side. Raja Iqbal, founder of Ejento AI, told Computerworld: "The metaphor is dangerous when it's decorative." Amy Loomis of IDC was blunter about where responsibility lands: with "the person who gave the agent access, not the agent itself." Russell Jukes, CDIO at DXC Technology, thinks of an agent "as a superpower, not as a person". None of them are arguing against agents. They are arguing against pretending the human role did not move. And here is the part that should settle the debate inside your leadership team. Wladawsky-Berger notes that adoption in the study depended far more on managerial expectations and organisational support than on how the AI was labelled. The name on the org chart does not buy you adoption. It only costs you ownership. ## Write the human's paragraph this week The capability work here is not onboarding for the agent. It is role redefinition for the human. That sounds large. It is not. You can start on Monday without a budget line. - Pick one agent already in use. Not a pilot, one that is doing real work today. - Find the person who works beside it most days. Sit with them for an hour. - Write one paragraph together: what they own, what they check before it leaves their hands, and the moment they stop the agent. - Do it with them, not for them. A role written about someone is a role nobody holds. If the paragraph is easy to write, good, you have a role. If it is hard, you have found the leak, and you found it before the auditor did. Either way, the agent already has its job description. This week the human gets one too. ## Sources - [Pat Brans, "Govern AI agents like workers but don't put them on the org chart", Computerworld (Iqbal, Loomis, Jukes, Tatsiy)](https://www.computerworld.com/article/4213117/govern-ai-agents-like-workers-just-dont-pretend-theyre-human.html) · 2026-09-14 - [Irving Wladawsky-Berger, "Should Organizations Treat AI Agents Like Employees?"](https://blog.irvingwb.com/blog/2026/09/should-organizations-treat-ai-agents-like-employees.html) · 2026-09-30 - [Matthew Kropp, Julie Bedard, Emma Wiles, Megan Hsu & Lisa Krayer, "Research: Why You Shouldn't Treat AI Agents Like Employees", Harvard Business Review](https://hbr.org/2026/05/research-why-you-shouldnt-treat-ai-agents-like-employees) · 2026-05-06