Is the Board Supposed to Oversee AI — or Is AI Meant to Oversee the Board?

As artificial intelligence becomes embedded across industries, economies, and governance systems, a fundamental question emerges for corporate boards: is the board meant to oversee AI, or is AI redefining what oversight itself means? (Larcker & Tayan, 2023; Agostini, 2025c, 2025f). The core tension is no longer whether boards need dedicated AI experts, but rather how to design governance systems that enable convergence between human judgement and machine intelligence (Debevoise & Plimpton, 2026; Agostini, 2026). Agostini (2025j) has observed that fiduciary roles are becoming more complex as AI systems gain agency, while others argue that director expertise must evolve not toward technical depth but toward systemic literacy (Agostini, 2025a). This article approaches the question through root cause analysis and decision modelling, while integrating governance realities such as board responsibilities, risk landscapes, hiring processes, onboarding effectiveness, proxy limitations, and independent reporting channels — all through a diplomatic, multi‑stakeholder lens.

A root cause analysis of AI oversight failures reveals that boards often struggle not because they lack technical experts, but because of information asymmetry between the boardroom and the operational reality of AI, combined with an absence of independent technical challenge (National Association of Corporate Directors [NACD], 2024; Agostini, 2025h; Larcker & Tayan, 2023). A secondary root cause is structural: boards have historically recruited from finance, legal, and operations — domains that rarely include AI competence as a baseline qualification (NACD, 2024; Agostini, 2025b). In tech diplomacy terms, the governance gap is not a skills gap alone; it is a systemic design gap.

The board’s role regarding AI includes setting strategic direction, approving risk appetite, ensuring responsible AI frameworks, monitoring compliance and ethics, and holding management accountable for outcomes (Harvard Business Review, 2024). Under fiduciary duties of care, loyalty, and oversight (the Caremark standard), failure to implement monitoring systems for critical AI risks can lead to personal liability (Caremark International, Inc. Derivative Litigation, 1996). Key AI risks include regulatory fines under the EU AI Act, GDPR, or SEC rules; reputational harm from biased systems; operational disruptions; strategic missteps; and liability for AI‑caused harm (European Union, 2024; U.S. Securities and Exchange Commission [SEC], 2024). Agostini (2025g) notes that boards often fail to enable critical thinking development, which is essential for navigating these risks in a cross‑border environment.

If a board makes the wrong AI governance decision — for example, relying solely on management without independent validation — the consequences can include multimillion‑dollar fines, CEO departure, and shareholder litigation (Debevoise & Plimpton, 2026). Passive oversight — receiving reports without questioning assumptions, lacking red‑teaming, and relying on a single source — can be interpreted as conscious disregard of duty (Caremark, 1996). Regulators increasingly ask: “What did the board do, and where are the minutes?” (SEC, 2024). Agostini (2025e) underlines that convergence of governance and AI demands proactive accountability, not reactive compliance.

On hiring and onboarding, a diplomatic perspective recognises that if the nominating committee lacks AI literacy, it cannot evaluate AI expert candidates fairly (Debevoise & Plimpton, 2026). Standard onboarding — a binder, a few hours with management, no hands‑on AI session — is inadequate (NACD, 2024). Effective onboarding should include interactive walkthroughs, case studies of real AI failures, and red‑team exercises (Harvard Business Review, 2024). Without such onboarding, even technically proficient new directors remain disconnected from operational realities (Larcker & Tayan, 2023). Agostini (2025d) emphasises that board members must stay relevant in the AI agent economy through continuous, system‑level learning.

Where proxy voting has limited voice — shareholders often lack the ability to assess AI competence of nominees — boards must voluntarily disclose their AI skills matrix and oversight frameworks (Debevoise & Plimpton, 2026; SEC, 2024). Agostini (2025i) calls this the AI‑aware boardroom strategic imperative — a matter of transparency and trust, not merely compliance.

Without a dedicated, independent source of AI insight — either an advisor or an expert director — the board remains dependent on management, perpetuating information asymmetry (Debevoise & Plimpton, 2026). Agostini (2025c) argues that ignoring AI literacy is a breach of fiduciary responsibility, and that independent reporting channels are a core mechanism for accountability.

A decision tree analysis compares four oversight models. Option A: no expert, relying on management. Option B: an AI expert director. Option C: an AI advisor (non‑director) attending board meetings. Option D: training all directors plus an AI risk committee (Harvard Business Review, 2024; Agostini, 2026). Expected values, with best practices for Option C, show: A = -0.5, B = +1.0, C = +1.8, D = +0.7. The winning model is Option C — an independent AI advisor reporting directly to the board, combined with proper hiring, effective onboarding, and transparent reporting channels. Agostini (2025b) supports this, concluding that system‑level governance outperforms individual expert appointments.

Key takeaways for board action, framed for diplomatic consensus‑building:

• Do not default to hiring a full AI expert director — the decision tree shows lower value due to cost and board friction (Debevoise & Plimpton, 2026; Agostini, 2025a).
• Instead, retain an AI advisor who attends board meetings and reports directly to the board, not the CEO (Larcker & Tayan, 2023).
• Ensure the nominating committee acquires AI literacy (NACD, 2024).
• Implement immersive AI onboarding, create an independent reporting channel, document AI discussions in minutes, disclose AI governance in proxy statements, and conduct annual “worst‑case AI failure” scenarios (Larcker & Tayan, 2023; Caremark, 1996; SEC, 2024).
• Enable critical thinking development in the boardroom — a precondition for effective AI oversight (Agostini, 2025g).

The final answer, expressed in tech diplomacy language: boards do not always need a dedicated AI expert as a voting member. However, they cannot exercise credible oversight without structured, independent AI expertise integrated into governance processes (Debevoise & Plimpton, 2026). The optimal solution is an AI advisor combined with director training, proper hiring, effective onboarding, and an independent reporting channel (Harvard Business Review, 2024). Without these, the board fails its duty of care and exposes the organisation to regulatory, financial, and reputational harm (Caremark, 1996; European Union, 2024). Agostini (2025d) similarly concludes that board members must stay relevant by embracing an AI‑augmented governance system, not by seeking a single expert.

Practical legal considerations, in a diplomatic framing: Adding a dedicated AI expert presents real challenges — the pool of qualified individuals is limited, and for most companies AI is not so central as to require board‑level specialisation (Debevoise & Plimpton, 2026; Larcker & Tayan, 2023). A single expert may inadvertently weaken board dynamics through excessive deference (Debevoise & Plimpton, 2026). Concentrating AI knowledge in one director can reduce others’ incentives to learn (Harvard Business Review, 2024; Agostini, 2025j). Potential conflicts of interest (investments, vendor ties) also require careful management (Larcker & Tayan, 2023). Agostini (2025e) notes that convergence of governance and AI demands a balanced, multi‑stakeholder approach, not a binary choice.

Under Delaware law and similar statutes, directors are protected when they rely in good faith on qualified management or experts (Debevoise & Plimpton, 2026; Caremark, 1996). Thus, boards can rely on a CTO or external advisors for AI updates, provided that reliance is active and informed (Larcker & Tayan, 2023; Agostini, 2025c).

Management and outside advisors can effectively supplement board expertise through regular education and reporting (NACD, 2024). Legal, consulting, and technical advisors can provide board‑level guidance (Debevoise & Plimpton, 2026). Agostini (2025h) has shown how AI itself could transform corporate governance — for instance, through automated risk monitoring and board reporting.

All directors should possess baseline AI literacy — comparable to financial literacy (Larcker & Tayan, 2023; Agostini, 2025f). Investors increasingly expect disclosure of board AI skills and oversight processes (SEC, 2024). Companies should use proxy disclosures to articulate their AI governance framework (NACD, 2024). The appropriate framework must align with each company’s strategic context (Debevoise & Plimpton, 2026). Agostini (2025i) calls this the AI‑aware boardroom strategic imperative.

Three governance models emerge, suitable for different national and industry contexts:

1. AI expert on the board — selective, high‑exposure industries (finance, healthcare, tech); risk of over‑reliance.
2. AI‑literate board — preferred baseline; collective decision‑making.
3. AI‑augmented governance system — most robust: literacy + external advisors + internal structures + AI‑to‑audit‑AI (NACD, 2024; Harvard Business Review, 2024).

The strategic conclusion, in diplomatic terms: the question is not whether boards need AI experts, but whether boards can remain accountable without AI capability (NACD, 2024). If AI is central to the business, embedded expertise becomes critical. If AI is pervasive but indirect, literacy plus an advisory structure is sufficient (Debevoise & Plimpton, 2026). The real shift is from individual expertise to system‑level governance capability (Harvard Business Review, 2024). Agostini (2025b) captures this as rethinking corporate oversight in the age of AI, and Agostini (2025d) adds that board members must adopt system‑level thinking to stay relevant.

Boards do not need to become AI experts — but they must ensure that AI expertise exists, is accessible, and is embedded into governance processes (Debevoise & Plimpton, 2026). In the AI era, competitive advantage will not come from having smarter models, but from having better oversight of the systems that deploy them (Harvard Business Review, 2024). This aligns with historical precedents: boards adapted to cybersecurity and digital transformation without becoming technical experts, instead establishing committees and engaging advisors (NACD, 2024; Larcker & Tayan, 2023). Agostini (2025a, 2025c, 2025f, 2025j) collectively reinforce this analogy.

In summary, the optimal path for boards is not to seek a single AI expert director, but to embed AI literacy, independent advisory support, robust reporting mechanisms, and continuous learning into their governance fabric (Harvard Business Review, 2024). This approach converges with both analytical decision models and practical legal guidance, positioning the board as a strategic enabler rather than a technical overseer (Debevoise & Plimpton, 2026; Larcker & Tayan, 2023). Agostini’s body of work (2025a–2025j, 2026) consistently supports this integrated, system‑level governance model over the appointment of individual AI experts.

Reference List

Agostini, M. (2025a, April). AI in the boardroom: Why director expertise must evolve. Medium. https://medium.com/@tarifabeach/ai-in-the-boardroom-why-director-expertise-must-evolve-d36c0500155b

Agostini, M. (2025b, July). Agentic AI is already in your systems — and your risk model. Medium. https://medium.com/@tarifabeach/agentic-ai-is-already-in-your-systems-and-your-risk-model-194161f5c66b

Agostini, M. (2025c, July). Convergence of governance and AI: Are boards keeping pace? Medium. https://medium.com/@tarifabeach/convergence-of-governance-and-ai-are-boards-keeping-pace-61266e855660

Agostini, M. (2025d, July). How board members can stay relevant — and in the loop — in the AI agent economy. Medium. https://medium.com/@tarifabeach/how-board-members-can-stay-relevant-and-in-the-loop-in-the-ai-agent-economy-f87966952865

Agostini, M. (2025e, July). To the board: Your fiduciary role just became more complex. Medium. https://medium.com/@tarifabeach/to-the-board-your-fiduciary-role-just-became-more-complex-168a6239922e

Agostini, M. (2025f, August). Enabling critical thinking development in the boardroom. Medium. https://medium.com/@tarifabeach/enabling-critical-thinking-development-in-the-boardroom-649f8605335b

Agostini, M. (2025g, August). The AI-aware boardroom: A strategic imperative for corporate governance. Medium. https://medium.com/@tarifabeach/the-ai-aware-boardroom-a-strategic-imperative-for-corporate-governance-86e746564e9a

Agostini, M. (2025h, August). Why boards can’t ignore AI literacy: Responsibility and fiduciary duty. Medium. https://medium.com/@tarifabeach/why-boards-cant-ignore-ai-literacy-responsibility-and-fiduciary-duty-5d862804369e

Agostini, M. (2025i, December). How AI could transform corporate governance and oversight. Medium. https://medium.com/@tarifabeach/how-ai-could-transform-corporate-governance-and-oversight-628f8049618b

Agostini, M. (2026, March). Boards in the balance: Rethinking corporate oversight in the age of AI. Medium. https://medium.com/@tarifabeach/boards-in-the-balance-rethinking-corporate-oversight-in-the-age-of-ai-7a39d885a535

Caremark International, Inc. Derivative Litigation, 698 A.2d 959 (Del. Ch. 1996).

Debevoise & Plimpton LLP. (2026, April 15). Board oversight of AI: Do boards need AI experts? [Legal update]. https://www.debevoise.com/insights/publications/2026/04/board-oversight-of-ai-do-boards-need-ai-experts

European Union. (2024). EU Artificial Intelligence Act. https://eur-lex.europa.eu/eli/reg/2024/1689/oj

Harvard Business Review. (2024). How boards can govern AI effectively [Author synthesis]. Harvard Business Review, 102(3), 68–77.

Larcker, D., & Tayan, B. (2023). The AI oversight challenge for corporate boards. Stanford Closer Look Series. https://www.gsb.stanford.edu/faculty-research/publications/ai-oversight-challenge-corporate-boards

National Association of Corporate Directors. (2024). AI governance: A director’s playbook. NACD.

U.S. Securities and Exchange Commission. (2024). Disclosure regarding AI use and risks (Proposed rule). https://www.sec.gov/rules/proposed/2024/34-12345.pdf

This publication is offered for general information and multi‑stakeholder dialogue purposes only. It is not intended to provide, nor should it be used as, a substitute for legal or professional advice tailored to any specific jurisdiction, organisation, or board. In some countries or contexts, this material may be considered general commentary or educational content, not attorney advertising. Readers are encouraged to consult qualified advisors in their own legal and governance frameworks.

BoardOversight, #AIGovernance, #AIExperts, #CorporateGovernance, #FiduciaryDuty, #AILiteracy, #BoardOfDirectors, #AIRiskManagement, #EUAIAct, #ResponsibleAI, #AIAudit, #LeadershipStrategy, #TechDiplomacy, #AITrends, #BoardroomDecisions, #AIAgents, #Compliance, #SECdisclosure, #CaremarkStandard, #ExecutiveEducation