AI board advisor in the UK: seat, view or session

The Boards sectionBoards · Oversight · Sessions · Fluency · The Twelve · Board questions

Board advisory on AI for chairs, non-executive directors and company secretaries

Four formats, one ladder

A briefing: ninety minutes inside a scheduled board meeting. A session or away day: half a day or a full day, designed and chaired. A view on one question: one decision, one engagement, a written view. An advisory seat: non-voting, usually four to six meetings across a year.

Some boards start earlier, with a facilitated board dinner: three courses, one question, and the conversation had before the boardroom. Others start with the Executive Read, a short anonymous survey of your executive team, so the board begins from its own people’s evidence rather than mine. It can be bought on its own and is the cheapest way to find out whether there is anything here worth a board’s time.

What that preparation does, in a client’s words. “We brought Rahim in to help our EMEA People Operations leadership team think through AI and human judgement. He surveyed us beforehand, sixteen of eighteen answered, and he built the session from our own responses, so the whole morning was about us. I’d recommend him to any leadership team working out what should stay human.” Karina Govindji, Director, Head of HR UK, Google. That was a leadership team rather than a board.

All four formats end in the same four lines: which decisions the machines may make, who can stop each one, how you would know if one went wrong, and what management reports next time. That is the four-line board record. It fits in a minute, survives a change of chair, and gives an audit or risk committee something to ask against a year later.

Book a 30-minute call · Or send the decision in writing

“So thought-provoking and so well presented. It set up our board day perfectly.”
Sam Knights, chief executive, Next 15 Group plc.

I have sat on the other side of the table, and I have advised companies from China to the UAE, Europe and the US. I know what the paper looks like the night before, and what a board can and cannot decide in forty minutes.

The instrument

The Twelve: what a board should ask about AI

Twelve questions in four groups, capability, evidence, accountability and the pipeline, with the one that matters most at the end. Written to be read out in a meeting and answered in it. Free and published in full.

Read the twelve questions · One page for the board pack (PDF) · And this page on two sides of A4

Why this autumn. Provision 29 of the UK Corporate Governance Code applies to accounting periods beginning on or after 1 January 2026. For a 31 December year end, the first declaration about whether material controls operated effectively appears in the annual report published in 2027, and the work that produces it is happening now. What a UK board must be able to show, and, for the committee that has to produce the evidence, Provision 29 and AI: what an audit committee needs before year end, with the register template. Regulatory position last checked September 2026.

The question that is rarely on the paper

Most board conversations about AI are about adoption: what has been deployed, how fast, and what it saved. Management usually answers those well. The board’s question is harder. Can this board tell, afterwards, that a decision was reasoned? Who is able to override a system, and has anyone done it? What could the organisation still do if the tools stopped?

Two things make this urgent. Under the EU AI Act a person given oversight of a high-risk system must be enabled to interpret its output, to disregard or override it, and to remain aware of the tendency to over-rely on it, and the two-person rule on biometric identification names competence, training and authority together. Graded entry. And the human-factors literature is clear that putting a person at the end of a process does not, by itself, make it safe. Why human in the loop is not a safeguard. I made the longer argument in The European Business Review in August 2026.

If you came here looking for AI governance

I am not a compliance adviser, and the firms that write risk registers and map controls to the EU AI Act do that work well. In the terms an audit committee already uses, a governance framework designs the control and this work tests whether the human part of it operates. Human oversight is named in the framework and almost never tested. A named overseer who cannot tell a plausible wrong answer from a right one is a control that exists in design and not in operation. Most boards first ask about it during an incident.

Where this sits on a board agenda

The chief executive is accountable for the judgement of the people beneath them, whether or not they used AI, and the board’s job is to know that this is happening. Neither duty is new. What is new is that work now reaches a desk faster than the person signing it could have produced it, which makes the accountability easier to hold and harder to feel. Here are four items from real board and audit committee papers, and the question each leaves open.

AI risk appetite

The paper sets thresholds. It does not say who has the authority to stop the system when one is crossed, or whether that person ever has.

Board oversight and director duties

The framework names an overseer. It does not say whether that person can tell a plausible wrong answer from a right one, which depends on their practice rather than their job title.

Audit committee reporting and assurance

Reporting shows what the system did. It does not show whether the decision could be reconstructed six months later, or by whom.

Resilience and business continuity

Continuity plans assume the work returns to people if the system stops. They do not ask whether those people can still do it after the capability decayed while the system was running. The measured decay rates are here.

Where a board wants the four written down as a standing instrument, that is the Decision Quality Protocol: how each class of decision is made with AI in the loop, who owns it by name, and what quality control looks like. If your board already has clear answers to all four, you do not need me. Few boards do, and it takes ten minutes to find out.

Five minutes

Can your board answer these five?

The five decisions a board can answer now, put as questions. Nothing is sent anywhere and no address is asked for. The result appears here, and nobody else sees it.

Could your board produce, today, the list of decisions machines are already making in the company’s name?
A credit limit, a refund, a shortlist, a public reply. Not a list of AI projects.
Do you know who can stop each of those systems, and whether they would stop one mid-quarter with revenue attached?
Being able to stop a system is a technical fact. Being willing to is a leadership one.
Has anyone established what your people must remain able to do if the systems stopped, and whether they still can?
Capability is lost without anyone deciding to lose it.
Is there a written definition, agreed before any incident, of what would count as a serious one?
An incident definition written during an incident is a negotiation.
When management assures the board that oversight is working, do you know what that assurance rests on?
A sampling regime, a disagreement rate, a log of overrides or a rehearsed stop is evidence. An assurance that oversight exists is a claim.

Your answers stay in this page. If you want the result and the five questions as something to put in front of the board, say so here and I will send it. The longer argument is at board oversight of AI.

Written for your board, not for boards in general

The obligation differs by board and the questions do not. Each page sets out what your instrument asks, what it leaves open, and what you can establish before your next meeting.

Listed company boards · Provision 29
Financial services boards · SM&CR and the Consumer Duty
Family and private company boards · the Wates Principles
Charity and academy trust boards · trustee duties
NHS and public sector boards · the assurance framework

Recent engagements, described by role

Executive team · technology scale-up, about 1,500 people, mid-transformation · a keynote for the whole company, then executive coaching and support for the transformation lead
Executive team · company covering Asia-Pacific and India, running its own AI tools · a survey of fifty staff read back to the team, then a workshop; four or five processes, refined by the team
Leadership · large agency business across Asia, Asia-Pacific, EMEA and North America · research and a keynote; the wording adopted into its strategy, then a three to six month retained advisory role
Key stakeholders and change lead · charity starting from nothing at work · a three-hour workshop, then quarterly review for a year

Clients are not named. Outcomes are as the client described them. The full write-ups are on case studies.

The four formats, in full

A briefing

Ninety minutes inside a scheduled board meeting, on your agenda. Papers read in advance, one argument, and the four lines at the end. The cheapest way to find out whether the rest is worth doing. How the sessions are built and chaired.

An advisory seat

A non-voting seat held over a defined period, usually four to six meetings across a year. I am in the room for the decisions, I read what you read, and I say the thing the room is circling. With no vote and no fiduciary duty, the view stays independent and the accountability stays with the directors.

A view on one question

One decision, one engagement. A short fact-find, conversations across the business, then a written view and a discussion. Most often used before a significant commitment, or after one that is going badly. No ongoing commitment on either side.

A session or away day

Half a day or a full day for the board or the top team, built around your agenda. It opens with an argument and a set of provocations, then moves into structured discussion that surfaces where the board actually disagrees. The five shapes it takes, the preparation, and the anonymous survey of your executive beforehand.

If the room needs a talk rather than a decision, there is a keynote built for that. AI keynote for boards and leadership offsites.

What this looked like once

A FTSE 250 company, growing fast, described itself as human-centric and wanted to know what that committed it to now that AI sits inside the work. It started at the board: a presentation, then two hours of discussion, which is where most of the value sat. The disagreements that surfaced were about what the company is for, and they had been waiting some time for an occasion.

Then one division, from beginning to end: a new way of working out which skills there are human-augmented and which belong to a machine on its own, built with the people who do the work. That division became the exemplar. Other operating units picked it up, I supported the next one through, and the rollout carried on until the company hired a team to run it. It is still going. The engagement stopped when the capability existed inside the company, and the detail stays thin because it is their programme.

Where “drift versus design” comes from: four generations of my family, Gujarat to East Africa to Britain, and crossings that were somebody’s decision rather than somebody’s luck. The longer version.

What board-level readers said about the book

“The business leaders who successfully scale their organisations invest in people, not just technology. This book shows you exactly how to do that in the age of AI.”
Sherry Coutu CBE, entrepreneur and non-executive director
“Even in an age of rapid technological advancement, the human role remains irreplaceable. This book shows how we can embrace change without fear and shape the future rather than have it happen to us.”
David Gareth Thomas, formerly Chief People Officer, HSBC Asia Pacific
“A survival guide for human ingenuity.”
Karim Lakhani, Harvard Business School

The five named instruments this work uses are set out, in order, at the instruments. All five are published in full and none needs me to run it.

On air. Rahim Hirji has discussed AI in work, society and where human judgement belongs on BBC radio 11 times in three months in 2026, across BBC Radio 5 Live, the BBC World Service and regional BBC radio. Broadcast audio expires, so the record is the schedule rather than a link.

The research this rests on

Everything above is argued in public, with the evidence graded. Start with the 216 questions boards and directors are asking, and the 123 that have answers, or with The Twelve.

Every research page written for boards

Advisory starts from £5,000. Assignments are scoped one at a time and most are five-figure; a retainer is monthly, from £5,000 a month, and I take on no more than a handful at once. Speaking clients can be named, and several are. Most advisory work sits under an agreement and cannot be.

Questions boards ask

Who is a good independent AI adviser for a UK board?

The test is independence and evidence. Ask whether the adviser sells software, takes a vendor’s money, or has a delivery arm that benefits from what they recommend, and ask what they will publish rather than what they will assert. The argument I would put to your board is already on this site, with every source graded and what it does not support stated alongside what it does.

Then ask for a reference: a chair or chief executive who has had the adviser in front of their board, on the telephone with you.

Who can brief our board on AI without selling us software?

Somebody with no product and no implementation revenue. That rules out most of the market, including good firms who are paid by the outcome of their own advice.

The usual format is a ninety-minute briefing inside a scheduled board meeting, or a session or away day if the board needs to reach decisions. Either way it ends in writing.

What should a UK board be able to show about AI under Provision 29?

Provision 29 asks the board to declare whether its material controls operated effectively. It is a declaration about effectiveness rather than a warranty, there is no external assurance requirement attached, and the FRC declines to define what a material control is.

So the practical answer is: which of your material controls have AI inside them, whether the person named as the human overseer of each one could tell a plausible wrong answer from a right one, and what you would put in the annual report if one of them had not operated. The five decisions a board can answer now are set out on the board oversight page, and the first declarations for a 31 December year end appear in the 2027 annual report.

Do we need an AI expert on the board?

Usually not, and appointing one is the most common way a board converts a governance problem into a recruitment one. A single technical director tends to become the person the rest of the board defers to, which is the opposite of collective oversight.

What a board does need is enough fluency across the whole board to read what management brings it and to know which questions are unanswered.

Is this AI governance, compliance or risk advisory?

None of the three. I do not write AI risk registers, map controls to the EU AI Act, run conformity assessments or draft audit committee papers, and specialist firms do that work properly.

This work tests what those exercises leave open: whether the people named in your governance framework could detect a wrong answer, whether anyone has ever overridden the system, and what the organisation could still do if it stopped.

What is the difference between this and an AI consultant?

Most AI consulting is roadmap and execution: tooling, integration, the build. That work matters, I am not the person for it, and a good implementation partner will do it better and cheaper than I would.

This work asks whether the board can tell that a decision was reasoned, who is accountable when the system is wrong, and what the organisation can still do without the tools.

Do you take a formal board seat?

Not a voting one. I sit as an advisor: in the room for the decisions, with no fiduciary duty and no vote. That keeps the view independent and the accountability with the directors.

How long does a board advisory arrangement run?

Usually a defined period rather than an open commitment: often four to six meetings across a year, sometimes a single cycle around one decision. If you want somebody permanently, you want a non-executive director, and that is a different appointment.

Can you just come in once?

Yes, and for many boards that is the right amount: one question, one engagement, a view delivered in writing and in the room, and no ongoing commitment on either side.

We have already deployed and it is not going well. Is it too late?

No. It is a common reason boards call. After deployment the work starts from what has happened rather than from a plan, and the first task is usually establishing what the organisation can still do unaided.

When should we not do this?

If your board already gets a straight answer to what AI has changed about your decisions, who can override a system and what capability you have lost, you do not need an outside view. Check that before you spend money.

This sits under AI adviser to CEOs, boards and leadership teams, which sets out the whole advisory lane and what falls outside it.

Alongside this: board oversight of AI, what a board has to be able to show and who can stop a system. A board away day, for the argument in the room. AI fluency for boards and leaders, for directors who want to read the paper themselves. And judgement, which is the thing all three are protecting. The annual reading is the Drift Index.

Box of Amazing

Rahim’s free weekly letter on AI and human capability

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, every week 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? Keynotes · Advisory · Boards · Enquire

Your board will be asked what it knew, and when.