- Home
- Insights
- AI Agent Manager
- When do you need an AI Agent Manager? Authority decides, not the number of agents
When do you need an AI Agent Manager? Authority decides, not the number of agents
Companies usually think about this role when the number of agents grows. What decides it is not the number but the authority the agent is given. This piece covers the signs that the role is needed, and who takes it on at what scale.
Is it the number of agents that calls for the role?
No. Numbers are a misleading measure. Ten assistants a team uses for drafting do nothing in anyone’s name; a person reads and uses the output every time. A single agent that moves ad budget between campaigns, on the other hand, spends money on the company’s behalf.
The AI Agent Manager, the person who owns the task, authority, approval steps and measurement of a company’s agents, is needed in the second case. The role is described in full in this article.
Which signs show the role is due?
If even one of the six signs below appears, the owner of the agents should now be set down in writing.
| Sign | What it means | First step |
|---|---|---|
| An agent was given write access | The agent no longer only reads; it changes something in a system. | Write down which change is tied to human approval. |
| An agent sends something outside the company | It speaks for the company to customers, suppliers or the public. | Put an approval step before sending. |
| Several teams are setting up agents | The rules differ from team to team. | Agree on a shared brief template. |
| Nobody can give the full list of agents | Some agents run off the record. | Build an agent inventory. |
| Results are described as “seems good” | No baseline was measured. | Write one success measure for each agent. |
| After an error, who should fix it was debated | Responsibility is split between IT, the vendor and the business unit. | Name a single person as the agent’s owner. |
How to keep the inventory is explained step by step in the agent governance article.
What does regulation say?
The European Union’s AI Act (Regulation (EU) 2024/1689) asks organisations that use high-risk AI systems to assign human oversight to named people. Under paragraph 2 of Article 26, those people must have the necessary competence, training and authority, as well as the necessary support.
The same article also asks the deploying organisation to monitor how the system operates and to keep the automatically generated logs under its control for at least six months. The risk framework of the US National Institute of Standards and Technology (NIST) defines governance as the function that runs through all the others.
This article was written for high-risk systems; whether an agent in your company falls within that scope requires a legal assessment. In our view it is a good yardstick for companies outside the scope as well: give oversight to a person, give that person authority, and keep records.
Should the role be a full-time position?
In most companies, not at the start. The scale of the role grows as agents spread, and it can be thought of in three stages.
- One agent, in suggestion mode. The manager who knows the work takes ownership. The brief and the success measure are written down; no separate meeting is needed.
- A few agents that act. The role becomes a manager’s written responsibility and part of their objectives. A monthly review begins and an inventory is kept.
- Agents spread across teams. Shared rules, training and record-keeping become a job in their own right. At this point a separate person is considered for the role.
What moves a company from one stage to the next is, again, not the number but the authority the agents hold.
Who can take on the role?
Someone who combines three things: they can judge whether the work is done correctly, they have the authority to decide on risk, and they know how to measure the result in numbers. In most companies that is the operations or marketing manager who runs the work today.
Giving the role to IT alone is a common choice, but in our view an incomplete one. IT knows best what the agent can access; whether the report it produced or the message it sent is correct is something only a person who knows the work can say.
The first week for whoever takes on the role usually goes into a list: which agents exist, who set them up, what they can access, who uses them. If nobody has drawn up that list before, the list itself is the clearest evidence of why the role is needed.
What does setting up the role too early or too late cost?
The cost of setting it up early is bureaucracy. Convening a committee for a single drafting assistant puts teams off trying AI at all.
The cost of setting it up late is heavier. Agents without owners accumulate, nobody knows who is responsible at the first serious error, and leadership often stops every project at once. The experience built up so far is lost with that decision.
Assessment
In our view the right moment is just before the first agent is given write access. Up to that point the role can live in a manager’s notes; from that point on it should be in writing.
- Write every agent and automation running in the company into a single list.
- Ask one question next to each: does this agent suggest, or does it act?
- Assign a named owner to every agent that acts, and write it into that person’s objectives.
- Give the owners a month to fill in the briefs; the template is here.
- Put the first monthly review in the calendar and reassess the scale of the role after three months.
To see where your organisation stands today, take the eight-question agent readiness test.
Frequently asked questions
Does a company with a single agent need an AI Agent Manager?
If the agent only makes suggestions, no separate role is needed; the manager who knows the work takes responsibility. If the agent spends money, writes to customers or changes records, even a single agent needs a named owner.
Should the AI Agent Manager come from the IT team?
Not necessarily. The substance of the role is judging whether the work is done correctly and approving risk. An operations or marketing manager who knows the work often does that better; IT supports on the access side.
Does the role need a separate hire?
Usually not at the start. The role first becomes the written responsibility of an existing manager. A separate person is considered once agents spread across several teams and the monthly review becomes a job in its own right.
Sources
Read next
All insightsWho is accountable for AI agents? The AI Agent Manager role
Agents no longer stop at answering; they act on a company’s behalf. This piece describes the role that owns those actions: what it takes on, what it is not, and where a company starts.
Read the article AI Agent ManagerWhere does governance start? Analysis · 7 min readHow is agent governance set up? Inventory first, policy second
The first thing governance brings to mind is writing a policy document. A working arrangement is built from a five-line record for every agent and a short review held once a month.
Read the article