Leading AI Isn't a Title, It's a Stage

A management committee at a two-hundred-person company read the same piece everyone's been sharing on LinkedIn: a comparison of Chief AI Officer, Chief Data Officer and Chief Information Officer, explaining which one fits which size of company. None of the three fit this year's headcount budget.

Carlos Andrés Ramírez · 2026-09-05

A management committee at a two-hundred-person company read the same piece everyone's been sharing on LinkedIn: a comparison of Chief AI Officer, Chief Data Officer and Chief Information Officer, explaining which one fits which size of company. None of the three fit this year's headcount budget. The committee closed the tab and kept going without anyone actually in charge.

That comparison is large-enterprise literature, written by executive search firms advising companies with C-suite budget, half a dozen AI initiatives running at once, and a board that already agreed to create a new seat. The mid-sized company actually asking this question, the one with two or three hundred people and a pilot running in two departments, doesn't live in that world. Borrow the question from the large enterprise and you borrow an answer that doesn't fit either.

Because the real question was never about a title. It was about stage. Who leads AI changes depending on how much of it is actually running and in how many departments, not on how sophisticated the org chart is supposed to look. Naming the role before the stage justifies it is where this goes wrong.

The symptom

Who should lead AI in a company: IT, the business, or a new role?

It gets decided badly when the answer copies what a larger competitor did, instead of counting how much AI is actually running inside the company today. The committee debates the title before it has the volume that title would manage, and ends up choosing between three abstractions with nothing tying any of them to an operating fact.

• The committee debates creating a Chief AI Officer before more than two business processes depend on an AI system for anything real.
• Technology still approves which model gets tested, even though the process left technology months ago and three different business units now run on it.
• Every department has its own pilot, and nobody on the committee can name, without checking a list, how many are running at once.
• An AI lead role exists on the org chart, with no budget of its own and no authority to stop an initiative in another department.
• A new seat gets hired before anyone decides what technology stops deciding and what each business unit starts deciding instead.

The problem underneath

The title is a borrowed answer to a question about stage.

Most title comparisons come from a context the mid-sized company doesn't have: an organisation that already moved through several stages of AI maturity and now needs to coordinate the volume it built up. Copying the title without having copied that volume is exactly what fails. It's hiring a global logistics director for a company still delivering with two vans.

And the mistake isn't only about size, it's about sequence. Technology should lead the pilot, because the risk there is technical and so is the judgement call. The business should lead once the system starts deciding something real inside a department, because the risk there is operational, not technical. A dedicated role, with its own budget and the authority to stop initiatives in any department, only earns its seat once the combined volume and risk across several departments stop fitting into the calendar of someone who already has a full-time job. Skipping that order is the expensive part.

The title leads nothing. What leads is whoever answers when the system decides wrong, and that name exists long before the seat does.

BECOME

The framework

How do you decide who leads, stage by stage?

The order matters more than the job title. Each stage has a natural owner and a clear sign that it's time to move to the next one, and none of the four requires creating a new seat ahead of time.

Pilot

One or two AI experiments running, with no business process yet depending on them for anything real. Natural owner: technology, because the risk is still technical. Exit sign: the first pilot reaches production and someone from the business starts asking for a say in how it works.

Adoption in one area

One specific department, risk, sales, operations, already relies on the system to make a decision a person used to make. Natural owner: whoever runs that department, not technology and not a corporate role. Exit sign: a second AI process shows up in a different department, deciding things by a completely different logic.

Cross-department coordination

Several departments run AI at the same time and start duplicating work or making decisions that contradict each other. Natural owner: one person with an explicit mandate from the management committee, even without a new title or a large team; it can be someone already on the committee who takes this on in writing. Exit sign: the combined volume already demands full-time attention.

Systemic risk

The number of AI systems in production, or what a single failure would do, is already enough to move the whole company's results or reputation. Only here does a dedicated seat, with its own budget and the authority to stop any initiative, pay back what it costs to create it.

The question most committees ask isn't which stage the company is actually in. It's whether they can afford to jump straight to the fourth one, borrowing the title from someone who genuinely needed it. Almost never can they, and almost never do they need to yet.

Count how many business processes depend today on an AI system to decide something real, not how many pilots are running. If the number is one, the natural owner is that department, not a new seat. If it's three or more and nobody is coordinating between them, the company already moved to the next stage, whether or not anyone decided that on purpose.

Let's decide who leads your next AI stage