Search for a Chief AI Officer job description and you will find a document written for a company in California. It asks for a PhD, a publication record, experience scaling a machine learning platform, and a vision for artificial general intelligence. It has no register, no regulator and no board.
A UAE board reading that document cannot tell whether the person it describes would keep the company out of trouble.
Here is the version that answers the question a board here is actually asking.
Position
Chief AI Officer. Two or three days a week for the first twelve to eighteen months, moving to full time only if the evidence supports it. Reports to the Chief Executive, with a standing item on the board agenda.
Purpose
To give the company one accountable name for artificial intelligence. Today that name does not exist, which means the register does not exist, the policy is a memo, and the board has no standing answer when a regulator, a bank, a partner or an auditor asks who signs.
The four deliverables
One. The register and the assessments
Maintain a current inventory of every AI and machine learning system the company uses, whether built, bought or embedded in something else.
For each system the register records the purpose, the personal data it touches, the vendor and what the vendor does with that data, the named internal owner, and the oversight mode. The Central Bank of the UAE described the oversight modes as human in the loop, human on the loop and human out of the loop. Use those three words. They are now the vocabulary a UAE examiner recognises.
Run and keep current the impact assessments for any system that affects a person's money, health, employment, housing or legal position. DIFC Regulation 10 requires risk and impact assessments for high-risk processing and requires the register to record the necessity and proportionality of each activity.
This is the deliverable that everyone agrees is important and nobody does, because nothing breaks until the day somebody asks to see it.
Two. One production use case
Select one workflow, with the finance function, where AI can move a number the company already reports. Own it end to end: the vendor decision, the data work, the change to how the team works, and the measurement.
Report the result at ninety days whether or not it is good. A programme that only reports wins teaches the board nothing and is worth nothing.
Three. The board reporting line
Produce one page each quarter. The same page. Systems on the register, systems added, systems retired, assessments due, incidents, the state of the one production use case, and the regulatory position.
The purpose of the same page is that a board can see movement. A new deck every quarter hides movement, which is why new decks are popular.
Four. Policy and literacy
Write the internal policy on what staff may put into which tools, in language a sales manager will read once and remember. Then run the training that makes the policy true.
Abu Dhabi's Department of Health made this a formal requirement in its Responsible AI Standard of October 2025, which asks for continuous AI literacy programmes tailored by role, with clinicians, data scientists and administrators each trained differently. The principle travels beyond healthcare. Training a finance team and a warehouse team with the same slide deck produces two teams that ignore it.
What this seat does not own
It does not own the data platform. That is the CTO or the CIO.
It does not own model development headcount, at least not in year one. In a mid-market company here there is no model development headcount.
It does not own procurement of every tool with AI in the marketing copy. It owns the register that records them, and a veto on anything that fails an assessment.
It does not own the mandate to make the company an AI company. Any job description containing that sentence should be rewritten before it is posted.
Requirements
Has taken at least one AI system into production inside a real operating company and can describe what went wrong. This is the only requirement that reliably predicts the work.
Can write and defend a register and an impact assessment. Can sit in front of a regulator, a bank or an auditor and answer questions without a lawyer translating.
Understands the UAE regulatory position well enough to know which rules apply to which entity in a group that holds an onshore licence, a DIFC entity and an ADGM vehicle at the same time, and knows to confirm the specific requirement with the regulator in writing.
Has run a change programme with people who did not want the change. AI programmes fail in workflows, not in models.
Not requirements
A doctorate. A publication record. Deep learning research experience. Familiarity with a specific vendor stack. Prior use of the title, which barely existed in the private sector here two years ago.
Why the seat is fractional first
IBM's Institute for Business Value, with Oxford Economics, surveyed 2,000 CEOs across 33 geographies between February and April 2026 and found 76% of organisations now have a Chief AI Officer, up from 26% a year earlier. That figure is global. It publishes no Gulf cut.
What it tells a board here is that the title is becoming standard, not that a full-time appointment is the right first move.
A full-time appointment commits a senior package to a function the board cannot yet evaluate, on a technology the company has not finished deciding about. Two or three days a week from someone who has done it once already carries all four deliverables and produces the evidence a full-time job description would need. Twelve to eighteen months later the company can write that description from what it learned.
That is the same argument the Data Protection Officer went through a decade ago, and it ended the same way.
One caution about named regulatory positions
Regulation 10 provides for an Autonomous Systems Officer where processing is high risk. Whether a part-time or non-resident leader can hold a specific named regulatory position depends on that regulator's conditions on residency, approval and presence.
Confirm the specific requirement with the regulator, in writing, before the seat is scoped. Nothing on this page is legal advice.
Where to go next
If the seat is not yet agreed inside the company, read who owns AI. If it is agreed and the question is sequence, read AI transformation in the UAE.
If the description above is the job and you need the person, read the register. If you have already done this work inside a real company here, claim a page.