Why chief AI officers may need to think more like HR leaders

Enterprise AI is shifting from deploying individual tools to organising teams of agents, argues NTT Research chief executive Kazuhiro Gomi. His case is that leaders will need to decide how agents divide work, learn their roles and remain under human oversight, not simply pick a model and let it loose.

Watch Desk analysis

What happened

In a 6 October Forbes Councils article, Gomi says that as businesses adopt teams of specialised AI agents, questions about hiring and managing people start to apply: how many agents should handle a task, which should work together, what training they need and when a person should step in.

Gomi cites a study by NTT Research’s Physics of Artificial Intelligence Lab and Harvard University’s Center for Brain Science. He says its “Society of AI Agents” experiment found that adding agents did not keep improving performance indefinitely. The article also describes benefits from combining agents with different skills and assigning them suitable roles. These are findings as Gomi presents them; the article is an argument about organisational design, not a full account of the study’s methods.

Why it matters

If agent teams become a normal way to do business work, their performance will depend on more than the underlying models. Coordination, role assignment, oversight and the point at which a human takes responsibility all become part of the system’s design.

That gives chief AI officers and CIOs a broader job than choosing software. It also makes a familiar question newly practical: if an automated team gets something wrong, who was responsible for how it was assembled and supervised?

Our read

Gomi’s HR comparison is useful because it shifts attention from the agent count to the organisation around it. More agents are not automatically more capable, just as adding more people to a meeting is not a substitute for deciding who is doing what.

Businesses considering agent teams should define roles, readiness checks and human intervention points before scaling them. Treat the article’s research summary as a prompt to ask better questions, not as a complete playbook: the study details and evidence needed to generalise its findings are not set out here.

What to watch

• Whether further published details of the cited study clarify how its agent teams were tested.
• How organisations decide what training and readiness checks an agent needs.
• Whether human oversight has clear intervention points and named accountability.

Discussion spark: When a team of AI agents makes a consequential mistake, should accountability sit mainly with the person who deployed it, or with the leaders who designed the organisation around it?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.