What does AI do for board directors?

 AI works on the board pack before a meeting starts. It runs analysis a pack does not usually contain, recalls what the board discussed last time, brings in expertise the room does not have that day, and checks a figure against the market outside. A director reads what comes back, tests it, and decides what it means. The judgement and the vote stay with the director.

Directors use AI on the board pack, in the hours before a meeting. It runs the calculation a paper leaves out, finds the proposal the board rejected two years ago, explains a regulation nobody in the room has worked with, and sets an internal figure against the market. The director reads the result, tests it against what they know about the business, and forms their own view. Nothing runs during the meeting itself.

 AI supports board decision-making in four ways: running analysis a pack does not contain, recalling what a board has already discussed, bringing in outside expertise, and checking figures against the wider market. Those four matter because of where boards currently lose decisions. Rigid or inconsistent decision-making processes, unclear roles and responsibilities, and poor-quality information are the three leading barriers, cited by 34%, 32%, and 29% of directors in the Board Value Index, Summer 2026, a survey of more than 400 non-executive directors, CEOs, and CFOs across the UK, US, Nordics, and Middle East. The same research found that 84% of directors had seen at least one delayed, rushed, or poor decision in the previous six months as a result.  

A director can run a calculation directly on the pack in front of them. Board packs rarely carry the workings behind a number, which is part of why information quality ranks among the barriers above.

A director working on a live proposal puts the question to IQ Experts directly: if the input assumptions for this project are 10% out, does it still deliver an acceptable payback? The sensitivity test comes back in moments. The director weighs it against what they know about the business and decides what payback threshold the board should accept.

 A director can pull the board's own history into the paper they are reading now. Take a proposal that stalled two years ago and is back on the agenda. The director asks IQ Experts  what the original business case argued and how comparable projects have fared since. It returns the original paper, the board's discussion of it, and the track record of similar projects that went ahead. The director walks into the meeting knowing what the board decided last time and why.  

A director can get a specialist view on a subject outside their own background, using expertise the boardroom does not have to hand that day.

On a proposal touching EU sustainability reporting rules, a director asks IQ Experts  what has changed and what comparable boards are doing about it. The answer comes back with the relevant regulatory position, drawn from outside the paper itself and cited. A director with no sustainability background arrives at the discussion holding a grounded view of the rules.

A director can check a number in the pack against what is happening outside the organisation.

A director questioning whether an internal cyber figure is unusual asks IQ Experts how attempted attacks have trended over three years and how that compares with market norms. The external trend line comes back alongside the organisation's own figures, each source cited. The director judges whether the internal number sits in line with the wider picture or warrants attention.

What makes those four work is what sits underneath them. IQ Experts draws on Board Intelligence's governance advisory work, more than 20 years with FTSE 100, Fortune 500, and OMX 30 boards, across more than 130 governance domains and more than 80 boardroom capabilities. A general-purpose model answers a governance question from patterns in public text. IQ Experts draws on advisory engagements with real boards.

AI has three limits in the boardroom, and each has a check a director can apply.

Over-reliance is the first. A director who treats an AI answer as final has skipped the part of the job nobody can delegate. A director repeating an AI output in the room without having tested it first has crossed that line.

Input quality is the second. An analysis only reflects the pack it reads. A thin or poorly structured paper returns a thin, generic answer, however sharp the question put to it. A weak output reads fluently and says little: a broad statement where a number should be, or the paper restated with nothing added.

Accuracy is the third. Outputs need checking, as with any AI tool. Every answer should carry a source a director can trace, whether a page in the pack or a cited external reference. An answer with no traceable source needs checking before it reaches the room, and so does one that does not match the document it claims to summarise.

Some board decisions were never analytical to begin with. Whether a reputational risk is worth taking, how much weight to give one dissenting view, what tone to strike with a struggling management team: none of these has a calculation behind it. AI can supply the facts underneath a decision like that. The decision has never moved.

IQ Experts is built around those three limits. It runs before the meeting and takes no part in the discussion or the vote, so the director does the testing and decides what to do with the answer. Every response carries a source traceable to the pack or to a cited reference.

Senior independent directors, non-executive directors, and chairs use AI for preparation most, because those roles carry accountability for outside challenge across a wide remit. A NED reviewing a capital allocation proposal does not have the day-to-day operational detail an executive director builds up from running the business. AI closes some of that gap before the meeting starts.

Governance teams gain without running the analysis themselves. A director who arrives better prepared asks less of the company secretary during pack review, which frees the governance team for work closer to the role's purpose: advising the chair and shaping the agenda.

A generic AI assistant like Copilot or ChatGPT only sees what someone pastes into it. It holds no memory of past board packs and no board-specific expertise beyond what is already public. Ask it the sensitivity analysis question from earlier and it will run the maths. It will not know the board rejected a version of this project two years ago, because that history was never in the conversation.

Building an equivalent in-house is possible, and the model is the easy part. The harder engineering problem is reading years of historic packs safely, so each director sees only what they are permissioned to see. That problem does not end at launch. Packs keep accumulating and models keep changing, so a team that builds this in-house owns the maintenance indefinitely.

IQ Experts sits inside the board portal a director already uses, with the pack history and permissions already in place. Board Intelligence configures it for each organisation during onboarding, and it keeps no record of a query once it is answered.

A board evaluating an AI tool for governance work should check two things: how the tool is permissioned, and what sits underneath it beyond the model.

On permissioning, ask who can see what, and on what basis. Does the tool distinguish between individual directors, or does every user see the same output regardless of entitlement? Does the organisation's data feed model development, or stay out of it? Where is the data hosted, and does the organisation choose?

On substance, ask what is built into the tool beyond general-purpose language generation. A chatbot wrapped in governance terminology can sound authoritative while carrying no domain expertise underneath. The question is whether its answers draw on established governance practice or on patterns learned from public text.

IQ Experts answers both. Access is permissioned to the individual director, down to the specific paper, so a director sees only material they are already allowed to see. Board Intelligence configures it for the organisation during onboarding rather than training it on the organisation's materials, keeps no record of a query once a response is returned, and runs it on ISO 27001-certified controls in a jurisdiction the client chooses.