The Rise of the Chief AI Officer: mandate, operating model and first 100 days

Paper Six argued that every organisation of any size needs one named person accountable for coherence in its use of artificial intelligence, while the job title given to that person remains optional.

Nothing in this article constitutes legal advice. It is offered to support discussion and further review with qualified advisers.

What did the March 2024 memorandum on Chief AI Officers actually require?

It required every United States federal agency to designate a Chief AI Officer within 60 days, and it specified the seniority, positioning and responsibilities of the role in more detail than any publicly available job description of the time. The memorandum came from the Office of Management and Budget on 28 March 2024 and dealt with governance, innovation and risk management for agency use of artificial intelligence. Paper Six was written the following day. Everything below describes the position as it stood in the last week of March 2024, and the paper stands unrevised as a record of that month.

• The memorandum required designation within 60 days of issuance and specified the post more precisely than any public job description then available.
• The requirement itself had been created five months earlier by the executive order of 30 October 2023, which directed that guidance specify the designation of such an officer at each agency.
• The specification asked for senior executive level at the largest agencies, a primary role concerned with artificial intelligence specifically, and regular engagement with the deputy head of the organisation.
• The agency governance board was chaired by the deputy head, with the Chief AI Officer as vice-chair, which placed convening authority above the role.
• The memorandum was guidance to federal agencies and bound no private company, so its force stopped at the boundary of the federal sector.

Why did the chronology matter so much?

The role was created in October 2023 and specified in March 2024, and the five months between the two documents carried the argument. A public consultation opened at the beginning of November and closed in early December, and the questions it put were the ones the memorandum later answered. Readers who took 28 March as the origin lost the middle. The memorandum itself said so, recording that it acted as required by the executive order of the previous October.

Reading the two documents together gave a fuller picture than reading either alone. The executive order created the obligation and the memorandum supplied the specification, and the consultation between them helps explain why the specification took the shape it did. That was the correction Paper Six put at the front, because the temptation in the spring of 2024 was to present a single dated document as the origin of an entire role.

Which part of the design was worth copying?

The chairing arrangement was the most transferable element, and the cheapest to reproduce. The deputy head of the agency chairs the governance board and the Chief AI Officer sits as vice-chair, which puts convening authority one level above the role. An officer who holds the vice-chair draws attendance from peers who outrank them, because the meeting already belongs to someone the organisation answers to.

That matters more than it looks on paper. A senior appointment whose meetings are optional produces good documents and little change, and the ordinary difficulty of getting busy executives into a room is one of the quiet ways such roles fail. Placing the chair one level above the officer removes the problem at no cost.

How should the operating model have been arranged?

Paper Six proposed a federated model with a small centre. A central office of two to five people owns coherence across the enterprise, while accountability for delivery stays with the business units that are actually delivering. The argument for that split was practical. Centralising delivery produces a queue, and a queue produces a reputation for obstruction within about two quarters, after which the centre spends its remaining credibility explaining itself to people who have started working around it.

The paper also set out decision rights in written form, including one that commonly goes unassigned in organisations that have otherwise thought carefully about governance, and it gave a first 100 days for anyone taking the role on. A separate section located the European position precisely as it stood that month. That section above all should be read as a snapshot, since it captures a regulatory picture on a particular date in 2024.

Should most organisations have created the post?

Most should not, and that was the paper’s least fashionable conclusion. The accountability gap is genuine in nearly every organisation of any size, and the question of who owns enterprise coherence in the use of artificial intelligence has an answer whether or not anybody has written it down. In most organisations the person capable of holding that accountability already exists under a different name, whether chief information officer, chief data officer, chief technology officer or chief risk officer. Paper Six spent a section distinguishing the role from each of those four, and that distinction earns its keep at the moment a board decides an additional post is unnecessary.

The wider misuse of the memorandum followed from forgetting who it bound. It was guidance to federal agencies, so citing it as evidence that organisations generally need a Chief AI Officer is a category error. The error was easy to make because the document was detailed, public and free to read, and a specification written for one sector remains a specification for that sector.

Read now, the value of the paper sits in its separation of two questions that were routinely collapsed into one. Ownership of enterprise coherence is a question every organisation has to answer, and the answer determines whether anything joins up. The title is a second matter. What that person is called turns on structure, sector and who already holds the ground, and it admits of several sensible answers. A government made the role mandatory across its own agencies in March 2024, and necessity everywhere else remained a question for each board to decide on its own evidence.

Frequently asked questions

How is a Chief AI Officer different from a chief information, data or technology officer?

The distinguishing question is ownership of enterprise coherence in an organisation’s use of artificial intelligence, which is an accountability that exists whether or not anyone has written it down. Paper Six, written in March 2024, devoted a section to separating the Chief AI Officer from the chief information, data, technology and risk officers. In most organisations the person able to hold that accountability already works there under one of those four titles.

What seniority should a Chief AI Officer have?

The United States memorandum of 28 March 2024 asked for senior executive level at the largest agencies, a primary role concerned with artificial intelligence specifically, and regular engagement with the deputy head of the organisation. It also placed the officer as vice-chair of a governance board chaired by that deputy head, which put convening authority one level above the role.

Does the US Chief AI Officer requirement apply to private companies?

The requirement applied only to United States federal agencies. The memorandum of 28 March 2024 was guidance to those agencies, and its force stopped at the boundary of the federal sector. Citing it as evidence that organisations generally need the post is a category error, one made easy by a document that was detailed, public and free to read. A specification written for one sector remains a specification for that sector.

How large should a central AI team be?

Paper Six, written in March 2024, proposed a small central office of two to five people, owning coherence across the enterprise while accountability for delivery stays with the business units. The reasoning is practical. Centralising delivery produces a queue, and a queue produces a reputation for obstruction within about two quarters, after which the centre spends its remaining credibility explaining itself to people who have started working around it.

What should someone do in their first 100 days as Chief AI Officer?

Paper Six, written in March 2024, gave a first 100 days for the role, and its two most portable elements were the decision rights and the chairing arrangement. Decision rights go into writing, including the one that commonly goes unassigned in organisations that have otherwise thought carefully about governance. The governance board is chaired by the deputy head, with the officer as vice-chair, which draws attendance from peers who outrank them.

Paul Forrest is a fractional Chief AI Officer and AI strategist working with UK boards, private equity portfolio companies and growing businesses. Read how the fractional CAIO engagement works, or see the full range of Ecaveo AI services.