The AI Operator: Role, Responsibilities and How to Hire

Nearly every company has deployed AI agents. Almost none can show what they returned. In WRITER's 2026 survey of 1,200 executives and 1,200 employees, 97% of executives said their company had deployed AI agents in the past year, yet only 29% reported significant return from generative AI.

The gap is not technical. It is that nobody owns the work of turning tools into practice. That job now has a name, and this article covers what it involves, whether to promote or hire for it, and how to set it up so it does not quietly fail.

It comes from the AI Partners podcast episode where Thomas Spitz, CEO of AI Partners, hosted Théo Martin, Ecosystem Lead at Dust, and from what we see on client engagements.

What does an AI operator do?

An AI operator connects the business side to the technical side. They find the use cases worth building, get the agents adopted by the teams that need them, and keep the whole thing running once the initial enthusiasm fades.

Dust, who coined the term, describes the role as rebuilding company processes around AI rather than bolting AI onto existing ones. That is the ambition. The daily reality is narrower and more practical: identifying what to automate, training people, fixing what breaks, and reporting on what it produced.

It sits closer to operations than to engineering. The closest existing comparison is the person who runs Salesforce for a sales team. Not the one who built it, the one who makes it useful.

Why do companies need an AI operator?

Because individual adoption outran organisational adoption, and nothing in the existing structure closed the gap.

The same WRITER survey found 55% of companies describe their AI use as a chaotic free-for-all, and 79% say AI applications are being built in silos. Everyone experiments, nobody consolidates.

Managers were the obvious candidates to fix this, and they have not. Only 35% of employees say their manager is an AI champion. That is not a competence problem. A department head already has a job, and learning agentic tooling well enough to teach it is not part of it.

So the work falls to whoever picks it up, which usually means one or two enthusiasts doing it unofficially, in the margins of their actual role.

Should you promote an AI operator internally or hire one?

Promote, in almost every case. The role needs two things at once: deep knowledge of how your business actually works, and fluency with agentic tools. The second can be learned in months. The first takes years.

You almost certainly already have candidates. WRITER found that AI super-users are reported as five times more productive than colleagues slow to adopt, and three times more likely to have received a promotion and a pay raise in the past year. These people are visible. They are the ones building things nobody asked them to build.

Promoting them has a second benefit that hiring does not. It signals to everyone else that going deep on AI is a career path rather than a hobby, which does more for adoption than any internal communication campaign.

Hiring externally is the harder route. Someone with both the sector expertise and the agentic experience is rare and expensive, and they will spend their first six months learning your processes, which is exactly the part you were trying to shortcut.

Should the AI operator be a full-time role?

Yes, as soon as the programme is real. Giving a champion ten or twenty percent of their time works, and we have set up plenty of programmes that way. Beyond that it breaks, because the person still has their original job and their original targets.

This is the most common way the role fails. A company names a champion, gives them no time, no budget and no authority, then concludes a year later that the role did not work. What did not work was the arrangement.

If you cannot free up a full-time person yet, be explicit about the limit. A champion at twenty percent can run a community of practice, answer questions and spot use cases. They cannot redesign a process or maintain a fleet of agents.

Can an external partner act as your AI operator?

Yes, and it is often the right way to start. An external operator is available immediately, already knows both the business patterns and the tooling, and can be scaled up or down without a hiring decision.

We have seen engagements that began as an implementation project turn into a full-time person embedded with a single team of fifteen, simply because that team found the value obvious. That is a useful signal: if a team will fund a dedicated person out of its own budget, the role is justified.

The arrangement we recommend is hybrid rather than either-or. An external operator brings patterns from other sectors and gets things moving, while an internal champion is identified and trained alongside them. The external presence is temporary by design. The internal one is what remains.

How do you set up the AI operator role so it works?

Three conditions, and the first is the one most often missed.

Give them real means, inside real limits. Early adopters need room to build, and they also need guardrails, because the same curiosity that makes them good at this makes them likely to experiment with things nobody approved. Budget, access and a clear perimeter.

Make them teach, not just build. The operator who builds excellent agents alone has created a dependency. Train-the-trainer is what turns one capable person into a capability, and it is the difference between a programme that survives their departure and one that does not.

Measure what they change, not what they produce. Counting agents built rewards volume. Measuring adoption, time saved and the team objectives those feed into rewards the thing you actually wanted.

Watch the full episode

Thomas Spitz, CEO of AI Partners, hosts Théo Martin, Ecosystem Lead at Dust, on controlling agentic AI costs, governing a fleet of agents, and the roles that make adoption stick. The conversation is in French.

Frequently asked questions

Who should an AI operator report to?
To the business, not to IT. The role exists to change how a function works, so it needs the authority and the context of that function. IT provides the platform and the guardrails.

How is an AI operator different from a Chief AI Officer?
A Chief AI Officer sets strategy, governance and investment across the company. An AI operator works inside one function or a few, turning that strategy into agents people actually use. Large organisations need both, and they are not the same hire.

How do you find an AI operator inside your company?
Look at who is already doing it without being asked. The people building things nobody requested, running informal training for colleagues, or showing up in every conversation about tooling.

Identifying those people, giving them a method and training the teams around them is what our AI Academy programme is built for. Book 30 minutes and we will look at who your candidates are.