Why every board needs an AI seat soon
Boards remain slow to adopt AI where it matters most: preparation, debate and decision-making. New use cases—from LLM-powered meeting prep to live scenario testing and even AI “observers”—signal the early stages of a governance shift. Chairs who build literacy, experiment collectively and manage risks now will bank a durable strategic advantage.
Points clés
- in 2014, Deep Knowledge Ventures gave an algorithm full voting power on its investment board, an early signal of AI’s boardroom potential
- from June to September 2024, the authors interviewed 50+ board leaders from firms including ASM, Lazard, Nestlé, Novo Nordisk and Shell
- most directors had tried AI personally but had not integrated it into board responsibilities
- a Swiss chair (“Alexander”) uses ChatGPT to digest board materials and generate questions and options ahead of meetings
- an Austrian chair (“Gerhard”) ran LLM-driven scenarios on an acquisition, revealing risk beyond the board’s appetite and prompting management to add scenario analysis to proposals
- a Finnish chair (“Juho”) used ChatGPT to review outcomes of a two-day strategy retreat; its recommendations aligned with the board’s, boosting confidence
- a Dutch chair (“Catherine”) used Claude 3.7 Sonnet to re-examine four board conclusions; the AI confirmed three and triggered deeper debate on the fourth
- a Swiss industrial firm uses AI to analyse speaking time, tone and participation, advising airtime rebalancing and avoiding loaded phrases like “no-brainer”
- in 2024, UAE-based International Holding Company appointed “Aiden Insight,” developed by G42 via BoardNavigator, as a non-voting AI board observer with remarks recorded in minutes
- key risks and mitigations: information leaks (use enterprise LLMs like OpenAI and client-only models like SAP’s), sample bias (data audits and bias tools), and backward-looking anchoring (use explainable, scenario-driven models)
À retenir
Start small: let AI chew through your board pack so you don’t have to, then use it to stress-test a few strategic assumptions—no PhD or midnight coding required. Lock down data with enterprise-grade tools, broaden your datasets to avoid “management echo chambers,” and make scenario runs your new meeting warm-up. And yes, keep experimenting together—because if an AI board observer can earn a mention in the minutes, your human board can surely handle a few smarter prompts without breaking a sweat.
Sources
