Ask a board to explain why it made a particular call eighteen months ago, and even a well-run one will often struggle to reconstruct the full reasoning. Not because anyone was careless, but because board packs run to hundreds of pages, meetings move fast, and by the time anyone needs to revisit a decision, much of the context exists only in the memory of whoever happened to be in the room.
Much of what gets written about AI and corporate risk focuses on security: automated threat detection, deepfake defence and keeping attackers out of the network. That work matters, but it says less about what is changing inside the room where decisions actually get made.
The useful application of AI in the boardroom may have less to do with drafting packs faster than with retaining what a busy, well-prepared board still loses track of: the precedent from eighteen months ago, the risk buried on page forty of a two-hundred-page pack, or the pattern nobody quite connected across three separate meetings.
For company secretaries, that changes the nature of the job. Their role has always involved creating order around board information, decisions and accountability. AI potentially expands how much of that institutional knowledge they can realistically retain and retrieve.
The tax on speaking a second language
On an international board, some directors are working in their second or third language while everyone else gets to think in their first.
Research published in Corporate Governance: An International Review, examining nine multinational companies, found that changing a board’s working language to English without adequate preparation could produce what the researchers described as “impoverished and silenced discussions”. Some directors struggled to contribute or voice disagreement and reported feeling socially excluded.
Research into multinational business settings suggests the problem extends beyond boards. One NBER study found managers working across a language divide lost roughly a third of shared meeting time to language barriers, with average comprehension at around 60 percent.
Nobody documents this tax in the minutes. A non-native speaker may take in less of a fast-moving discussion and say less in response, while the gap itself is rarely acknowledged.
Real-time translation offers one way to reduce that disadvantage, allowing directors to concentrate more fully on the substance of a discussion rather than devoting part of their attention to simply keeping up. For the company secretary, whose role includes enabling an effective board process, that makes translation a governance consideration as much as a convenience.
There is an important distinction, however. Real-time translation helps participation while a discussion is taking place; it is not a substitute for certified translation of resolutions or legally binding text, where precise wording matters more than speed. A sound governance process needs to distinguish between the two.
When board memory is folklore, not fact
Ask most boards what they decided eighteen months ago and the answer will often be a reconstruction: memory, personal notes and best guesses filling in the gaps.
The company secretary has traditionally been the custodian of that institutional memory. But as the volume of information boards consume grows, expecting any individual to retain the context behind years of decisions becomes increasingly unrealistic.
Making meeting history searchable helps, but ordinary full-text search only goes so far. A PDF of an old board pack may be searchable by word, but not necessarily by decision.
The more useful approach is to structure information as it is captured: connecting a resolution to its committee, date, agenda item, related actions and eventual outcome. That allows someone preparing for a major capital allocation decision, for example, to find what the board concluded when it faced a similar question before — and, crucially, why.
This matters as boards face more first-of-a-kind decisions under real-time pressure, whether around AI investment, geopolitical exposure or supply-chain risk. A board that can retrieve its own reasoning has the opportunity to learn from itself rather than repeatedly reconstructing the same debates.
For the company secretary, AI therefore has the potential to turn institutional memory from something largely dependent on individual recall into something more systematic.
An unowned decision is just a wish
Remembering a decision is only part of the problem. Making sure it happens is another.
A decision recorded in the minutes with no named owner and no deadline is not really a completed governance process. It is an intention, and intentions can disappear beneath whatever becomes more urgent the following week.
The terminology matters. A decision is the outcome the board reaches in discussion; a resolution is the specific, binding wording that gets minuted; and an action item is the delegated task required afterwards. Confusing the three can leave a board with a perfectly recorded resolution but nobody clearly responsible for carrying it out.
AI does not create accountability by itself. It can, however, make gaps in accountability harder to overlook.
An action linked back to the original minute, assigned to an owner, tracked against a deadline and returned to the agenda when overdue creates a clearer chain between what the board agreed and what subsequently happened.
For company secretaries, the opportunity is not to hand responsibility to an algorithm. It is to spend less time manually reconstructing that chain and more time ensuring it remains intact.
Boards have more to oversee, not more people to do it
The scope of board oversight has expanded significantly. ESG disclosure, cyber governance, geopolitical risk and AI itself have all competed for board attention, usually without a corresponding increase in the number of people supporting the process.
EY’s analysis of Fortune 100 filings found the share of companies identifying AI as a distinct focus of board oversight rose from 16 percent to 48 percent in a year. The proportion assigning that responsibility to a specific board committee rather than the full board increased from 11 percent to 40 percent over the same period.
This expansion creates a practical challenge for governance teams. More information has to be gathered, understood, connected to previous decisions and converted into a board process that remains manageable.
That is where AI becomes more interesting than simple document generation. Summarising information, surfacing relevant historical material, tracking actions and helping people navigate increasingly large bodies of board content can potentially reduce administrative load without transferring responsibility for the underlying judgment.
The distinction matters. Efficiency is useful; governance still requires human ownership.
The board pack is becoming a different kind of tool
Board technology is consequently moving beyond simply replacing paper with PDFs.
PwC’s survey of corporate directors found that more than a third of boards had already incorporated AI into their oversight work. As adoption grows, the question is becoming less about whether AI appears somewhere in the board process and more about how it is governed when it does.
For company secretaries assessing AI-enabled governance tools, the useful questions are therefore fairly practical. Can the accuracy of an output be checked? Is there an audit trail? Can a person correct the record? Can users understand where the information came from?
These questions matter more than the novelty of the underlying technology because board information carries consequences.
Who is accountable when AI gets something wrong?
A 2025 EY analysis of Fortune 100 risk disclosures found that around one in five companies identified AI hallucinations, inaccuracies or bias as a material risk in their own filings.
A hallucination and an omission are different failures. One creates something false; the other leaves something relevant out. A governance process needs to be capable of catching both.
The harder question is responsibility. If an AI-assisted summary misrepresents a discussion, who is accountable: the technology provider, the company secretary who circulated it, or the board that relied on it?
A client alert from law firm Kane Kessler argues that board secretaries and counsel should remain responsible for finalising minutes, with AI treated as an administrative aid rather than a substitute for judgment or governance responsibility.
That principle provides a useful boundary. AI can help create, organise and retrieve information, but responsibility for the formal board record needs to remain identifiable and human.
What it takes to trust AI at board level
Trust in an AI tool at board level ultimately comes down to a handful of specific questions, and their apparent dullness is precisely the point.
Where does the data sit, and under which jurisdiction? Who inside and outside the organisation can access it? Is that access role-based and logged in a way that can be audited?
Can a director or company secretary override or correct what the technology produces? Does that correction create a timestamped record rather than silently overwriting what existed before?
For European boards operating under data-protection regimes such as GDPR, these are not technical footnotes. They go directly to confidentiality, accountability and the integrity of the board record.
A sophisticated AI capability is of limited value in a governance setting if nobody can explain precisely how sensitive information is handled or how an incorrect output is corrected.
Governance has always had a memory problem
The company secretary’s job has always involved being the board’s institutional memory: knowing what was decided, why it was decided and what was supposed to happen next.
AI does not necessarily remove that responsibility. In many respects, it increases it.
Technology can retain more institutional information than any one person realistically could, surface relationships across years of meetings and make forgotten actions harder to ignore. But someone still has to determine whether what the technology produces is accurate, appropriate and fit to become part of the board’s record.
That may be the more significant way AI changes the company secretary’s role. The job moves gradually from being the person expected to remember everything towards being the person responsible for ensuring the organisation’s systems of memory can be trusted.
Boards that treat this purely as a technology question may spend the next few years comparing features. Those that treat it as a governance question — about who remembers, who is accountable and who can be trusted with sensitive material — will ask much sharper questions of every technology they use.
Babur Mirza is Global Head of Sales at Sherpany, a meeting management solution for boards and executive teams.
