Chief AI Officer Job Description Template 2026

Last updated: September 27, 2026

By Mike Carter, Managing Director, KORE1

A chief AI officer job description must spell out the reporting line, the AI budget the seat controls, which deployment decisions it can approve or stop without a vote, and the regulations it answers for. Base pay in our salary guide runs $280,000 to $400,000 at mid-market companies and climbs to $400,000 to $550,000 at enterprises, with bonus and equity on top. Most postings we read cover the pay and skip the other four.

Can I stop a model from going live?

That was the question. The second finalist asked it too, nearly verbatim, and both of them withdrew within a week of getting the answer.

The employer was a regional health plan outside Columbus, Ohio, about 900,000 members, and its posting for a chief AI officer ran to thirty-one bullets. Every one was a verb. All thirty-one. Drive, champion, evangelize, partner, foster. It asked for a PhD “preferred,” ten years of machine learning leadership, board presence, and “passion for responsible AI.” What it never said was who could say no to a model. The honest answer, which the CEO gave the finalists over dinner, was the chief risk officer’s model committee, which met monthly, and on which the new CAIO would hold a nonvoting seat. The job was accountable for AI and had no authority over it. Both candidates had run that movie before. Neither wanted the sequel.

We rewrote the posting with them in February. It lost nineteen bullets and gained four sentences. The CAIO reports to the CEO, vice-chairs the AI governance committee, owns a $6.2 million budget line, and can pull any model out of production on its own signature, with the committee reviewing that call afterward. The search closed in twelve weeks. The person who took it told us the fourth sentence was why she applied.

Mike Carter here. I run client searches as a managing director at KORE1, and executive AI seats have gone from an occasional request to a regular one since 2024. KORE1 earns its fee only if a client hires a person we put forward, so factor that in as you read. None of it requires hiring us. If you want the search itself (sourcing, interviews, the offer), our guide to running a chief AI officer search covers that ground. What follows is about the posting itself, and it assumes you are hiring a true officer. A builder who manages a dozen engineers is a different posting, which I come back to near the end, and our chief AI officer staffing team sorts the two apart on the first call more often than you’d guess.

IBM’s 2026 CEO study, which surveyed 2,000 chief executives in 33 geographies between February and April, found that 76% of their organizations have a CAIO. In 2025 the same question returned 26%. IBM reported those numbers in May. Tripling in a year means a lot of job descriptions got written in a hurry, and it shows.

The Only CAIO Job Description a Government Ever Wrote

There is exactly one chief AI officer job description with the force of policy behind it, and almost nobody in the private sector reads it. On April 3, 2025, the Office of Management and Budget issued memorandum M-25-21, which replaced the 2024 version and told every federal agency to retain or designate a CAIO within sixty days. Section 3 lists what that officer must do. It reads like a posting written by a lawyer, which is roughly what it is.

Skip the federal plumbing. What matters is how specific it gets about authority, because that is the part corporate postings leave vague.

Three details stand out. The memo says the CAIO must be “positioned highly enough to engage regularly with other agency leadership,” naming the Deputy Secretary. It lets the CAIO certify small, time-limited pilots so they can go ahead before the full risk practices apply, with each certification tracked centrally. And on waivers, the exceptions to the minimum practices for high-impact AI, it is blunt. “The CAIO’s responsibility under this paragraph shall not be delegated down to other officials.” One sentence, and the seat has teeth.

Here is how we translate it for companies. The right column is language you can lift.

You won’t use all seven. Nobody does. A 400-person software company doesn’t need a waiver process, and writing one in would be theater. But notice the shape. Every row is a noun the seat owns, not a verb it performs. “Champion AI adoption” can’t be checked by anyone. “Owns the inventory” can. It has an owner and a location, and in a year somebody can open it and count.

The memo also allows an agency to hand the title to a sitting CIO, CDO, or CTO, provided that person “has significant expertise in AI.” Companies copy that move all the time. Does it work? Sometimes. More on that below.

Decision Rights Belong in the Posting

Serious candidates for this seat read a posting like a term sheet. Adjectives get skipped. They hunt for authority, and if it isn’t on the page they assume it doesn’t exist. Usually they’re right.

So put it on the page. We walk every CAIO client through this grid now, before anyone drafts a word, and I’ll warn you the middle column is ours, pieced together from what leadership teams admitted once we pushed, which was rarely what the org chart said.

Look at row two. “Nobody, in practice” isn’t a joke. Ask a leadership team who can switch off a misbehaving model the evening before a long weekend and the room tends to go quiet, until somebody offers that the vendor probably can. Probably. That silence is the job. Put the answer in the posting and the people who have actually held that power will spot it in a second, while the ones who have mostly given conference talks about it will hesitate, and the hesitation tells you something too.

A caution. Don’t write authority you won’t actually grant. We watched a medical device maker in Minneapolis post “full authority over AI deployment” in 2025, hire against it, and then route the new CAIO’s first shutdown request through a quality board that took three weeks to meet. She left in month seven. The posting was the only place that authority ever existed.

Name the Rulebook, With Dates

“Familiarity with AI regulations” is the most common requirement line we see on these postings, and the least useful. Which regulations? Enforced when? A candidate who has actually carried the compliance load reads that phrase and concludes nobody inside has decided what applies yet.

The calendar moved a lot this year, so here is where things stood when I wrote this in late September 2026.

Sources for the table are the European Commission’s notice that the AI Omnibus entered into force, the Colorado General Assembly’s page for SB 26-189, ISO’s listing for ISO/IEC 42001, and NIST’s AI RMF page.

The ISO line earns its own short paragraph. Clause 5.3 of 42001 requires top management to assign responsibility and authority for the AI management system, and for reporting on how it performs. If your company is pursuing certification, the CAIO is the natural assignee, and the posting can say so in nine words. “You will be the named owner under ISO 42001.” A candidate who has sat through a certification audit reads that and knows the job.

The EU delay is a trap. A quiet one. A few clients have read “December 2027” as “later” and taken the regulatory line out of the posting entirely. Don’t. The hire you make this fall is the person who has to be ready by then, and fourteen months is short for inventorying systems, classifying them, and building the documentation for anything that lands in Annex III. Change the date in the line. Keep the line. If your governance program is still on paper, our piece on AI guardrails for mid-market companies is a reasonable starting point for the conversation before the hire.

Where the Seat Stops

One sentence, placed in the posting, saves more interview time than anything else in this article. It says where the CAIO’s job ends and a neighbor’s begins.

The fights are predictable. Data quality and lineage usually stay with the chief data officer. Infrastructure and model hosting usually stay with the CIO or CTO. Security of the AI stack sits with the CISO, with the CAIO setting the policy it enforces. Write it plainly, something like “the CDO owns the training data, and you own the decision to put it in front of a model,” and the finalists won’t have to ask. If you are still deciding which of those executives you need in the first place, our chief data officer search practice and the CAIO team compare notes on exactly that question most weeks.

A regional bank in Charlotte learned this the slow way. Its CAIO and its CDO both believed they owned the model inventory, and for five months each kept one. They didn’t match. Not even close. Our chief data officer job description template draws the same line from the CDO side.

Pay Ranges to Put in Writing

Post a range. Yes, even for an officer. Executives read postings too, often more carefully than anyone, and a CAIO posting without a number signals either that the band isn’t set or that it’s low.

These tiers come from our chief AI officer salary guide, which sorts the market by company size because that, more than the title, is what moves the number.

Two notes from recent offers. The posting only needs base. Bonus and equity are real, but they’re negotiated, and printing an equity figure invites the finalist to treat it as a floor. And the decision rights you wrote above move pay. A CAIO with shutdown authority and a consolidated budget is carrying more personal exposure than one who advises a committee, and good candidates price that in. Expect the top half of the tier. If your metro is unusual, the salary benchmark assistant will adjust the band for it.

Chief AI Officer Job Description Template

Replace every bracket with your own facts. Text in parentheses is coaching for whoever drafts the req, and it comes out before the posting goes anywhere public. If a section doesn’t fit your company, cut it rather than stretching it to fit.

Title

[Chief AI Officer / Chief AI Officer and SVP, AI / Chief AI and Data Officer] (Use a combined title only if the seat really owns both. A combined title with half the authority is the Charlotte problem, printed on a business card.)

Why This Seat Exists Now

[Company] uses AI in [the two or three places it matters, such as claims triage, underwriting, customer service, or clinical documentation], across roughly [N] systems, [N] of them bought from vendors. [The event that created the seat, such as a board request, a customer audit, an EU sale, an acquisition, or a program that outgrew its sponsor.] We need one officer who owns what we run, decides what we launch, and answers for both.

Reporting Line and Authority

(Candidates look for this block first. Keep it near the top.)

• Reports to [the CEO / the COO], and sits on the [executive leadership team]
• Vice-chairs [or chairs] the AI governance committee, which meets [monthly]
• Owns an AI budget of $[amount] for [fiscal year], consolidated from [the cost centers that hold it today]
• Leads a team of [N] today, with [N] approved hires in the first year

Decisions That Are Yours

• Approving AI systems for production, directly or through written delegation by risk tier
• Removing any AI system from production on your own signature, with committee review afterward
• Granting and revoking exceptions to the AI policy
• Signing AI vendor and model contracts up to $[threshold]
• [Serving as the named owner of our AI management system under ISO/IEC 42001] (Keep only if certification is real.)

What You’ll Own in the First Year

• A complete inventory of AI in use at [Company], including AI features inside vendor software, by [quarter]
• A risk classification for every system on it, with independent review for anything high-risk before launch
• [The one or two business outcomes that justify the seat, stated as numbers, such as cutting claims-triage time from four days to one]
• Readiness for [the EU AI Act high-risk obligations taking effect December 2, 2027 / Colorado SB 26-189 / the framework that applies]
• Quarterly reporting to the board on AI spend, risk, and results

Background We’re Looking For

• Has led an AI or machine learning organization of [N]+ people with its own budget
• Has personally approved, and personally stopped, AI systems in production
• Has built or run an AI governance program under [NIST AI RMF / ISO 42001 / sector regulation such as model risk guidance in banking]
• Can explain a model’s failure to a board in plain language, and has done it
• [Industry experience, only if it truly matters, such as HIPAA-covered data or FDA-regulated software]

Helpful, Not Required

• Hands-on experience with [the platforms you run, such as Azure OpenAI, Amazon Bedrock, or Google Vertex AI]
• Experience with an ISO/IEC 42001 certification audit
• A graduate degree in computer science, statistics, or a related field (Leave the PhD off unless the seat runs research. It cuts the pool and rarely predicts the job.)

Location, Travel, and Pay

[Headquarters city, hybrid [N] days / remote with [N] trips a quarter]. Travel about [percentage]. Base salary [$low] to [$high], plus [bonus target], [equity], and benefits. (Our guide puts mid-market base at $280,000 to $400,000 and enterprise base at $400,000 to $550,000. If the seat carries shutdown authority and a consolidated budget, price in the top half.)

Posting Questions From CAIO Search Committees

Can our CIO or chief data officer carry the CAIO title instead?

Sometimes, and even the federal government permits it, as long as the person has significant AI expertise and the authority that comes with the title, not just a second line on the business card.

OMB’s memo lets agencies designate a sitting CIO, CDO, or CTO. Where we see it fail is time. A CIO running infrastructure, security budgets, and three ERP projects gives AI whatever is left on Thursday afternoons. If you go this way, move something off that person’s plate in the same announcement, or the posting you skipped writing becomes the one you write next year.

Should the posting name OpenAI, Anthropic, or a particular cloud?

Name what you already run, in the “helpful” section, and leave vendor choice open if the new officer is supposed to make it.

Listing Azure OpenAI as a requirement tells candidates the big decision has been made, which is fine if it has. It also quietly filters out an excellent officer who spent three years on Bedrock. Platforms are learnable at this level. Judgment about when not to use them is the harder thing to find.

Europe pushed the high-risk deadline to December 2027. Can the regulation line come out?

Keep the line and change the date, because the officer you hire now is the one who has to be ready when Annex III obligations apply on December 2, 2027.

Taking it out tells candidates compliance is someone else’s problem. It isn’t. The ones who have built a program before will read that as a warning.

Our AI budget is under $2 million. Is this really a chief AI officer seat?

Probably not yet, and at that budget a head of AI who answers to the CTO or COO usually attracts stronger candidates than an officer title the company can’t back with authority or money.

There’s no official cutoff. That $2 million is a rough line from our own searches, below which the officer title tends to outrun the job. Two better-fitting options exist. The builder seat one level down has its own write-up in our head of AI hiring guide, and a fractional head of AI can stand up governance at a part-time cost until the budget justifies a full-time officer. Some clients do the second and then hire the first. It works.

How long should the finished posting be?

Around 600 to 900 words, which is shorter than most CAIO postings we’re sent, since every line should be something a candidate can verify in the first interview.

The Columbus posting ran about 1,400 words before the rewrite and 700 after. Length wasn’t the problem, exactly. Thirty-one unverifiable bullets were. Cut the verbs. Keep the nouns.

What do strong candidates read first?

The reporting line, almost every time, followed by the budget and the sentence about who can stop a model.

That’s from our finalist debriefs, where we ask what made someone keep reading. The reporting line comes up first far more than anything else. Now and then somebody says pay. Honest people.

Three Sentences to Settle Before Anyone Posts

Before a single bullet gets written, get your CEO to agree on three sentences. Who does this person report to? What can they stop? How much money do they control? If the leadership team can’t agree on those in one meeting, the posting isn’t the problem, and no recruiter will fix it by finding a better candidate. We’ll say that on the first call, even though it costs us the search for a while.

Once the three are settled, the template above takes an afternoon. Most of the clients we work with on executive searches for AI leadership arrive with the posting already drafted, and most of those drafts shrink by half in the first review. Once the posting is live, our chief AI officer finalist questions cover the interview that follows.

KORE1 has recruited out of Irvine since 2005, placing people in over 30 American metro areas, and our placements hold at 92% twelve months in. Nearly every CAIO hire is a direct hire. Executive AI searches run longer than our average fill time, usually eight to fourteen weeks, and a clear authority section does more than anything else to shorten them. If you’d like a second read of your draft before it goes public, send it to our team. We’ll mark it up either way.