Home / Boards / NHS and public sector boards
The Boards sectionBoards · Briefings and away days · Private dinners · AI oversight
By regime · NHS and public sectorA public board already lives inside an assurance architecture: a board assurance framework, a risk register, internal audit, and a code of governance written for the sector. AI does not arrive as a new row in that framework. It arrives inside a procurement, a supplier’s product upgrade, or a clinician’s workaround, and the framework finds out afterwards.
Rahim Hirji advises chief executives, boards and leadership teams, in person worldwide and online. He is an independent advisor on AI and human judgement, the author of SuperSkills (Kogan Page, 2026), and has run, grown, bought and advised businesses with AI in them. He also speaks, to boards, leadership teams and conferences. He founded the skills platform EtonX, later acquired by Eton College, and led Quizlet’s international growth across more than 60 countries.
Published, broadcast and bylinedPublished by Kogan Page, part of Hachette UK · Reviewed in Director, the Institute of Directors’ magazine, one of its top picks for Q4 2026 · A regular commentator on AI for the BBC, including the World Service and Radio 5 Live · Bylined in the New York Observer and The European Business Review · The record
“For directors navigating this transitional era, SuperSkills serves as an essential, reassuring compass.”
NHS provider boards work to a code of governance for the sector and to a board assurance framework that names risks, controls and sources of assurance. The structure is better than most private boards have.
What it assumes is that a control named in the framework is a control that operates. For a control whose operative part is a human reading a machine’s output, that assumption is the thing to test rather than the thing to record.

Whether the clinician or officer named as the human check has the time, the information and the standing to disagree with the system. Standing matters more here than anywhere: a junior member of staff overruling a tool the trust has bought is making a career decision as well as a clinical one.
Whether a decision could be reconstructed for a coroner, an inquiry or an ombudsman six months later, and by whom. And what the service could still do if the system were unavailable for a fortnight, which is a continuity question that has usually been answered on paper and not tested.
Which systems in use today produce an output a person is expected to act on. How each of them arrived: procurement, pilot, supplier upgrade, or nobody quite knows. Who is named in the assurance framework as the control, and whether that person knows they are.
And the override count, which in this sector is the single most useful number a board can ask for. Staff who have never disagreed with a system have either never had reason to, or have learned that disagreeing is expensive. Those are very different findings and the board should know which it has.
I do not write AI risk registers, map controls against the EU AI Act, run conformity assessments, build governance frameworks or draft committee papers. Firms that do that work properly exist and several of them are very good. This is the question those exercises leave open: whether the people named in your framework could actually detect a wrong answer, whether anyone has ever overridden a system, and what the organisation could still do if it stopped.
Regulatory position last checked: September 2026. If you are reading this much later, check the dates before you rely on them.
Comparing speakers? How to choose an AI keynote speaker, in five steps, and the 100 best AI keynote speakers, by topic and country.
No, and clinical safety has its own standards, specialists and regulators. This is the board-level question those processes leave open: whether the human part of the control operates, and what the service could still do without the system.
As a source of assurance against controls that already exist, rather than as a new risk. Most boards find the gap is not a missing row; it is a row whose source of assurance is the existence of a policy rather than evidence that it worked.
A great many organisations discover otherwise. Models arrive inside products the organisation already buys, through supplier upgrades that need no new decision. The first useful exercise is usually a list of what is already in use rather than a list of what was approved.
Yes, and public sector work is scoped and priced for it. What I will not do is bid as an implementation partner, because I am not one and the independence is the product.
I reply myself, within a working day, and we talk before anything is proposed.
Enquire or email rahim@thesuperskills.com
The argument is at board oversight of AI, the twelve questions are at what a board should ask about AI, and the offer is at board advisory. The parent page for all of this is AI keynote speaker. Browse every topic, audience and region, or take the speaker pack to whoever is running the day. Every talk so far is on the speaking record.
The pages this session draws on:
If this was useful and you want a small amount of it each week, that is what the letter is for. Every week, an essay and ten links to broaden your view of the world. Read by 25,000 people, since 2017. Free, and one click to stop.
Opens Substack to confirm. No pitch in it, unsubscribe in one click, and nobody follows up because you read something.
Running an event, or responsible for how AI arrives in your organisation? AI keynote speaker · Advisory · Boards · Enquire