Insights from Our Co-Host
Helen van Orton
I'm a passionate advocate for both AI governance and using AI in your personal governance practice. I spend a significant part of my time helping directors build the fluency to govern and use it well. But being an advocate does not mean being uncritical, and conversations like this one are valuable precisely because they force us to separate the genuine opportunities from the genuine risks, rather than collapsing both into a generalised anxiety about "AI in the boardroom."
The opportunity is real and specific. Directors are rarely short of information. What we are short of is time to reflect on it, and fresh perspectives that we don't have around the table. Used well, AI is exceptionally good at both. I think of it like a dice sitting in the middle of the boardroom table: every director looking at it sees a different combination of numbers, shaped by their own experience, but none of us can see what's facing down. The real skill in using AI as a director is learning to ask it to show you the face you can't see. What would a regulator flag here that we haven't considered? What would someone with a people and culture lens notice that a commercially-minded board might miss? That is a fundamentally different use of AI to feeding it a board pack and asking for a summary, and it is the use case I would encourage every director to build into their own preparation.
The risk is equally real, and I don't think boards have caught up with it yet. AI is trained, structurally, to agree with you. Research published in Science earlier this year found that AI systems affirmed users' positions 49% more often than humans did (Sharma et al., Science, 2026). If every director on a board is running the same board papers through the same tool, we should expect exactly what we would expect from any process that pulls toward a statistical average: convergence, homogenisation, and a loss of the very diversity of thought that makes a board valuable in the first place.
The legal landscape is also moving faster than most boards realise. In ASIC v Bekier [2026] FCA 196, the Federal Court made clear that AI can assist directors in processing and analysing board materials, but cannot replace the personal, non-delegable duty to exercise independent judgment. In the United States, Fortis Advisors LLC v. Krafton, Inc. (Delaware, 2026) saw a CEO's AI chat history become part of the discoverable court record once litigation commenced. Safe harbour protections, where they exist, tend to depend on the director having exercised genuine personal judgment above and beyond whatever the AI produced. None of this means avoiding AI. It means using it as a thought partner to interrogate, not an oracle to defer to. It also means being conscious that very little of what you do with AI, even outside a formal board meeting, is as private as it might feel.
If there is one thing I would want every board to take from this conversation, it is that AI strategy should never be treated as a technology project. The question is never "what is our AI strategy?" It is "what is our organisational strategy, and where can AI genuinely help us deliver it?" The boards getting real value from AI are the ones investing as heavily in their people, giving them permission to experiment, to fail safely, and to be retrained into higher-value work, as they are in the tools themselves. Get that right, and AI becomes a genuine source of competitive advantage. Get it wrong, and it becomes an expensive distraction from the judgment boards actually exist to provide.
AI is no longer a peripheral technology topic for boards, but neither is it yet a settled part of how most boards operate. It sits, for the great majority of directors, in an uncomfortable middle ground: widely used in personal and executive life, while cautiously (often deliberately) avoided in board work itself. It is only beginning to be addressed as a formal governance question.
BoardOutlook recently convened a Global Directors Council roundtable bringing together non-executive directors and chairs from listed, private, regulated and mutual organisations across Australia, New Zealand and the United States. Participants brought direct experience from banking, insurance, mining, retail, energy, technology and financial services boards, alongside specialist backgrounds in regulatory affairs, cybersecurity and AI governance itself. The discussion drew on BoardOutlook’s benchmarking data from approximately 150 boards globally to establish context, before opening to a candid conversation under Chatham House Rule. The roundtable was co-hosted with Helen van Orton, founder of Directorly and a practising director across banking, insurance and technology.
The picture that emerged was consistent: directors are individually moving quickly, often significantly outpacing the formal policies of the boards they sit on. Several participants described using AI extensively in their advisory or executive careers while applying a much stricter, more conservative standard to anything touching board material. The gap between personal fluency and board-level policy sits at the centre of this paper, alongside the emerging legal, governance and cultural questions that boards will need to resolve.
Participants converged quickly on a useful framing: AI in the boardroom is not one conversation but three. The first is AI as a driver of organisational performance — how the business uses AI to grow, compete and operate more efficiently. The second is AI in the governance domain — what the board is accountable for in overseeing management's use of AI, its risks and questions of ethics. The third, which was the focus of the conversation , is AI as a tool directors themselves use to prepare, challenge and exercise judgment more effectively.
Most boards have real, structured time allocated to the first conversation and increasing time on the second. Almost none had a considered, board-endorsed position on the third. Several directors observed that they are, in effect, running two entirely separate relationships with AI: an enthusiastic, well-developed one in their professional and executive lives, and a cautious, largely improvised one in their capacity as a director. The consistent view was that this gap will not close on its own, and that boards which address it deliberately, rather than leaving individual directors to make their own calls, will be better placed both to capture the benefit and to manage the risk.
A strong, practical theme concerned how boards actually govern AI use amongst their own directors, as distinct from how they govern AI within the organisation. The clearest pattern was that boards functioning well on this front had reached explicit, collective agreement — typically calibrated to the comfort level of the most cautious director or company secretary, rather than the most enthusiastic — about what is and is not appropriate. Where that agreement existed, practice was described as considered and consistent. Where it did not, several participants used the phrase “Wild West” to describe the resulting variation.
A related point was that disclosure of AI's role in a piece of thinking is more valuable to a board than concealment. Presenting an argument as one's own spontaneous insight, when it was substantially AI-generated, was seen as harder for the rest of the board to interrogate on its merits than the same argument presented candidly as AI-assisted. Boards that build a culture where “I asked AI about this, and here's what came back” is a normal, unremarkable contribution, (rather than something to be embarrassed about or to conceal) were seen as better positioned to use AI productively while retaining the challenge function a board exists to provide.
A cautionary theme running through much of the discussion concerned AI's tendency to agree with its user, a well-documented characteristic of current large language models and one cited in the room as making AI roughly 49% more likely to affirm a position than a human would. Several participants connected this directly to a risk of homogenised board thinking: if every director is running similar board papers through similar tools with similar prompts, the natural tendency of AI to converge on a statistically likely, broadly agreeable answer works directly against the diversity of perspectives a well-composed board is designed to provide.
This was not treated as a hypothetical risk. Multiple participants described boards, including their own, that had trialled AI summarisation tools on board papers and switched them off after limited use, specifically because of this concern — the issue was not accuracy of the summary, but the loss of each director's individual reading and independent perspective on the underlying material. One exchange captured the concern precisely: if management increasingly relies on AI to draft board papers, and directors respond by using AI to digest and challenge them, the risk is a conversation that is, in substance, AI talking to AI, with no genuine human perspective anywhere in the loop.
Practice on AI and the company secretarial function varied more than on almost any other topic discussed, and there was no consensus reached in the room. Approaches ranged from full retention of AI-assisted meeting transcripts under active oversight, through record-and-delete models with a short retention window overseen closely by the company secretary, to an outright prohibition on any recording or AI note-taking at board level.
What was consistent was the risk framing. Retention obligations differ meaningfully by jurisdiction and by entity type — government and statutory boards, in particular, face specific restrictions on what may be deleted and when. Legal privilege in relation to AI minute-takers being present for privileged discussions has not yet been tested in any jurisdiction represented in the room, and participants were encouraged to treat that as a live, unresolved risk rather than a settled question. There was strong agreement that directors should not independently run unauthorised recording or transcription tools in parallel with the board's official process, and that boards should think explicitly about which conversations, particularly those that are legally privileged or where directors need to be able to speak with complete candour, warrant an AI minute-taker being switched off entirely.
The discussion pointed to recent case law in Australia and the United States as early signals of how courts are approaching AI and director judgment. In ASIC v Bekier [2026] FCA 196, the Federal Court found that AI can assist comprehension of complex material but cannot be relied upon to displace the exercise of independent judgment. A recent Delaware case, Fortis v Krafton, was raised as a cautionary example closer to home for dealmakers: once litigation commenced, the CEO's AI conversation history became part of the discoverable court record, undercutting the company's own earnout-avoidance argument.
Chief Justice Bell's 2026 Harold Ford Memorial Lecture at the University of Melbourne Law School was also referenced, reinforcing the same theme from the bench: safe harbour and similar protections tend to depend on evidence of genuine, independent judgment being exercised above and beyond whatever an AI tool produced.
Cases and sources referenced:
Several participants shared examples of organisations experimenting with AI “personas” in and around the boardroom — from tools designed to anticipate a director's likely questions on a paper before it is tabled, through to fully synthetic digital board participants used in an advisory capacity in a small number of overseas examples. Reactions ranged from cautious interest to clear scepticism. The practical value of these tools was widely seen as unproven, and a specific concern was raised repeatedly: systems with access to an organisation's full data set risk pulling board conversation down into operational detail, rather than keeping it at the level of strategy and judgment that is the board's actual remit.
Set against that caution, participants were consistent and specific about where AI is already adding real value in their own practice. Using AI to research unfamiliar technical or regulatory areas ahead of a board meeting, to synthesise what peer organisations or regulators globally are doing on a given topic, and to pressure-test a position by adopting a perspective the board doesn't otherwise have in the room — a regulator, a specific stakeholder group, a sceptical investor — were the most frequently cited and most positively regarded applications. The distinction drawn repeatedly was between using AI to think more broadly before arriving at the boardroom, and using it as a substitute for the reading, challenge and judgment that happen within it.
Disclose AI's role rather than concealing it: Boards that normalise “AI suggested this” as an ordinary contribution find it easier to interrogate ideas on their merits than boards where AI's involvement goes unstated.
Set explicit rules of engagement, calibrated to the most cautious director: Collective agreement is important as ad hoc, individually-determined practice produces uneven and unpredictable results.
Use AI to bring a missing lens into the room, not to shortcut reading the papers: The strongest use cases involved prompting AI to represent a perspective — regulatory, cultural, adversarial — that the board doesn't otherwise have, rather than using it to summarise material directors should be reading themselves.
Give minutes, recordings and privileged discussion their own, more conservative rules: Whatever a board's general AI policy, retention periods, hosting arrangements and rules for switching AI tools off during sensitive discussion warrant explicit, separate agreement.
Name AI directly on the skills matrix: AI capability remains one of the most frequently raised concerns among directors, yet is rarely included as a distinct, named skill. Instead it is commonly folded into a general “technology” heading.
Invest in people at least as heavily as in tools: The organisations finding genuine value from AI were consistently those treating it as a change-management and capability question, not solely a technology procurement decision.
AI readiness in the boardroom needs to be measured and actively managed, not assumed. BoardOutlook's platform supports boards across the full set of processes that make this possible: board, committee, director, chair and CEO evaluations, and skills matrices that can incorporate AI as an explicit, assessed competence.
Where this becomes particularly powerful is in combination with AI itself. BoardOutlook's AI-powered tool, OutlookIQ, enables boards to interact with their own performance, composition and evaluation data dynamically, rather than through static, point-in-time reports. Drawing on a board's own underlying data, BoardOutlook's global benchmarking dataset, and an institutionalised governance framework developed in partnership with experienced chairs and directors, the platform supports thought-partnered analysis across succession planning, capability gap identification, performance oversight, and strategic priority alignment. Insights are supported by inclusion of the underlying data source, and hosting, retention and deletion controls are explicitly engineered for the sensitivity of board data.
For boards ready to treat AI readiness as a governance priority, both as a domain to oversee and as a capability to build into their own work, BoardOutlook provides the tools, data and structured process to make that commitment concrete and sustained.
For information on upcoming sessions or the BoardOutlook platform, please contact the BoardOutlook team at team@boardoutlook.com.
This paper was developed by BoardOutlook based on a Global Directors Council roundtable co-hosted with Helen van Orton. Participant contributions are reflected thematically and are not individually attributed, in accordance with the Council's confidentiality principles.