Perspectives from both sides of the board table

Artificial intelligence can help corporate and nonprofit boards prepare more efficiently, analyze complex information, identify emerging risks, test strategic assumptions, and ask sharper questions of management. It can also streamline agendas, meeting summaries, follow-up tracking, and director education, giving boards more time for strategy, oversight, and informed debate.

Adoption is already accelerating. Morgan Stanley reported in August 2026 that 68% of S&P 500 companies had at least one non-executive director with AI skills, while 41% of executives in a separate survey identified human review in higher-risk situations as the most important guidance for employees using AI. These findings reinforce the central governance principle that AI should enhance — not replace — human judgment, ethics, and accountability.

Boards should establish approved uses, protect confidential information, validate material outputs, require human review of consequential decisions, and begin with a few low-risk, high-value applications before expanding as their knowledge, controls, and confidence grow.

One of the most immediate applications is board preparation. AI tools can summarize lengthy board packets, compare current reports with prior versions, identify changes in assumptions, and extract decisions, risks, and unresolved questions. A director serving on the audit committee could receive a briefing focused on financial controls and risk indicators, while another director could concentrate on strategy, talent, or customer issues.

AI can also help shape agendas around strategic priorities, summarize meeting discussions, and track follow-up commitments. These applications should not replace careful reading or the corporate secretary’s governance responsibilities. Their value lies in reducing administrative burden so directors can spend more time debating the matters that require experience, perspective, and judgment.

Real-World Example: Diligent reports that ELCO Mutual Life and Annuity reduced board-preparation time from one or two days to one or two hours after adopting Diligent Boards. Diligent attributes the improvement in part to AI-powered board-book summaries and automated meeting minutes. ELCO’s general counsel also reported that executive committee meetings fell from roughly two and a half hours to an average of 30 minutes, crediting agenda time limits within the broader platform. Because the performance claims come from Diligent’s customer materials, they should be treated as vendor-reported rather than independently validated.

AI can expand the board’s ability to evaluate strategic choices — organizing market intelligence, testing assumptions, and developing scenarios across economic, regulatory, technological, or workforce change. When management presents a major investment or acquisition, AI-assisted analysis helps directors explore alternative outcomes and surface questions a traditional presentation might miss.

The same capabilities strengthen risk oversight: detecting anomalies in financial and operating data, monitoring emerging threats, and connecting risk indicators across the enterprise. But the board should govern AI as enterprise transformation and risk — not merely an IT project — with visibility into where AI is used, what value it is expected to deliver, who is accountable, and how outcomes and risks will be measured.

Real-World Example: PwC’s July 2026 guidance advises boards to oversee AI as an enterprise transformation, not a technology initiative, and identifies six areas for attention: governance and accountability; strategic positioning; talent and culture; the evolving workforce; risks and controls; and monitoring outcomes and risks. This is professional guidance, not evidence of outcomes at a particular company, but it offers a credible framework for board questions and oversight.

AI offers new ways to develop directors and executives — personalized learning plans, explanations of unfamiliar topics, questions for management presentations, and simulated boardroom situations involving strategy, ethics, cybersecurity, or crisis response. Directors can practice reasoning through difficult choices before facing them in real time.

For executives and emerging leaders, AI can support coaching, role-play challenging stakeholder conversations, and suggest development activities tied to defined competencies. It can make learning more accessible and responsive, but it should not make employment decisions or replace the context, empathy, and accountability of experienced leaders and coaches. Confidentiality is especially important whenever leadership assessments or employee information are involved.

Real-World Example: Advantexe describes an AI Board of Directors debriefing tool embedded in its business simulations. Participants present their strategy, decisions, and results to AI personas with perspectives such as finance, marketing, operations, ESG, and innovation; the personas challenge assumptions and provide immediate feedback. The example confirms that this type of simulation is commercially offered, but Advantexe’s article does not provide independent outcome data demonstrating improved on-the-job performance. It is best presented as an illustrative learning application, not proof of effectiveness.

Boards need not become technology experts, but every director should develop enough AI literacy to understand its opportunities, limitations, and risks. Governance should define approved tools and uses, protect confidential and personal information, require validation of important outputs, and establish escalation procedures when results appear inaccurate, biased, or inconsistent with company values.

Board and committee responsibilities should be explicit — oversight may involve the full board as well as audit, risk, technology, compensation, or governance committees. Management should provide regular, forward-looking reporting on major AI initiatives, investments, incidents, workforce effects, and performance. The goal is not to slow innovation but to ensure it advances strategy without creating unmanaged risk.

A practical starting point is two or three low-risk applications — summarizing public information, preparing director education materials, or generating questions for strategic discussion. Establish safeguards, test the outputs, measure whether the tools improve board effectiveness, and expand only after the organization learns what works.

Real-World Example: Microsoft, Google, and IBM each publish first-party materials describing how they organize responsible-AI development and deployment, with documented practices including governance roles, risk assessment, human oversight, testing, documentation, monitoring, feedback, and accountability across the AI lifecycle. These are self-reported and do not independently establish how consistently or effectively every practice is implemented. The NIST AI Risk Management Framework 1.0 adds an independent, voluntary structure organized around four functions — Govern, Map, Measure, and Manage — for managing AI risk.

AI is changing how leaders work and how they learn. It can summarize information, identify patterns, generate alternatives, and provide immediate support before a difficult conversation — extending customized development beyond a small group of senior executives. Leaders can use it to test assumptions, explore unfamiliar issues, and prepare for new responsibilities.

Yet AI cannot set aspirations, earn trust, understand the full human context of a choice, or accept responsibility for the consequences. The strongest leaders will use AI to think with them, not for them. Development must combine digital fluency with curiosity, ethical judgment, empathy, and discernment — the ability to ask useful questions, evaluate AI responses, recognize missing context, and know when human expertise must take precedence.

Organizations should first define the capabilities their strategy requires, then determine where AI can help. Effective development combines AI-enabled practice with feedback, real-world assignments, mentoring, coaching, and reflection.

Real-World Example: The Center for Creative Leadership uses its HiFi Conversation Analytics™ tool in a senior-leadership business simulation. The technology measures patterns such as speaking time and interaction across the group, while human experts interpret the findings, supply context, and guide development — demonstrating how AI can make behavior more visible without replacing human judgment, wisdom, or connection.

Executive coaching traditionally depends on a confidential relationship between a leader and an experienced coach. AI offers a different kind of help: immediate support for reflection, difficult conversations, presentations, goals, and problem-solving. This accessibility can broaden development — managers who may never receive individual coaching can gain structured support between formal sessions.

But availability is not wisdom. AI sees only the information provided and may miss organizational history, culture, power dynamics, or blind spots. Its responses can sound confident while remaining incomplete or inappropriate. Human coaches build trust, notice patterns, ask uncomfortable questions, interpret context, and provide accountability. AI may be an effective sparring partner; a skilled coach understands the whole person.

A thoughtful model combines both: AI for preparation, repetition, and follow-through; a human coach for deeper insight, values, relationships, and consequential choices. Organizations should also establish clear rules for confidentiality, approved platforms, data storage, and human review.

Real-World Example: Microsoft partnered with Coachello to embed leadership skills into the daily work of more than 150 managers, giving them real-time support through a hybrid AI and human coaching platform in Microsoft Teams. Coachello reports that 90% of participants gained greater clarity about their development path. Because the performance claims come from Coachello's customer materials, they should be treated as vendor-reported rather than independently validated. The case illustrates the reach and immediacy of AI coaching while reinforcing the value of pairing technology-enabled support with human coaching for deeper context and accountability.

Leadership is often tested with little time to prepare — a crisis, a personnel issue, an ethical conflict, a major presentation, a decision under uncertainty. Instruction alone does not create readiness. Leaders also need practice.

AI can role-play an employee, customer, board member, or regulator. Leaders can rehearse difficult news, resistance to change, unexpected questions, and competing priorities in scenarios that adapt as the conversation develops. Practice can be repeated privately, tailored to a role, and scheduled close to the real event.

Still, a simulation is only as useful as its design. Weak scenarios can reinforce poor assumptions, and automated feedback may miss context, credibility, timing, or effects on real people. Confidential information should be excluded unless both the platform and process are approved. Organizations should define clear learning objectives and pair simulation with human observation or debriefing — the key question is not whether the leader completed the exercise, but what was learned and what will change.

Real-World Example: ZEALAXX AG uses Retorio’s AI-powered leadership simulations to help leaders rehearse workplace conversations and receive immediate, personalized feedback. Platforms such as Retorio and Virti allow managers to practice performance reviews, conflict resolution, difficult feedback, and other high-stakes situations repeatedly in a private environment before facing them at work. Because the performance claims come from Retorio’s customer materials, they should be treated as vendor-reported rather than independently validated. The applications demonstrate how simulated practice can build readiness and self-awareness when followed by thoughtful reflection and debriefing.

Leadership development often centers on an annual workshop, retreat, or cohort. These experiences matter, but leadership challenges rarely follow a training calendar — leaders need support while managing change, resolving conflict, or entering a new role.

AI can turn development into an ongoing practice through short learning activities, explanations, reflection prompts, and timely rehearsal. Learning can adapt to different roles, experience levels, and needs, translating broad competencies into practical actions leaders can try in daily work.

However, personalization should not become isolation. Leadership grows through relationships, shared experience, observation, feedback, and accountability — usage data alone does not prove that behavior or performance has improved. A strong model combines brief AI-supported learning with peer discussion, applied assignments, coaching, and reflection. Progress should be judged through observable behavior and business results, not completed modules.

Real-World Example: A Fortune 100 technology company developed “Knowledge Nuggets,” concise six-minute learning videos delivered directly through Microsoft Teams. AI and Microsoft Power Platform tools supported content development, personalized recommendations, automated delivery, and usage reporting. Because the performance claims come from a consulting partner’s case study, they should be treated as vendor-reported rather than independently validated. The initiative shows how development can move from occasional courses to brief, accessible learning embedded in the flow of work.

As AI becomes part of everyday work, leaders will manage systems in which people and intelligent tools contribute together. The first task is deciding what should — and should not — be delegated. AI may search information, identify patterns, generate options, or automate routine work; people remain responsible for context, relationships, value conflicts, and consequential judgments.

Leaders must also evaluate AI-supported recommendations — a polished answer is not necessarily a sound one. They should explain why AI is being introduced, how work may change, and where employees can raise concerns. Trust will depend on fairness and transparency as well as performance.

Real-World Example: Bank of America’s EricaAssist works alongside more than 18,000 customer-service representatives during client conversations. The AI summarizes why a client is calling, surfaces relevant information, and recommends next steps in seconds, while employees remain responsible for listening, judgment, and the relationship. Bank of America reports the system reduces average call time by nearly one minute. Because the performance claims come from the company’s own announcement, they should be treated as company-reported rather than independently validated.

This series is educational and does not constitute legal advice; corporate law varies by jurisdiction, entity type, and circumstance, and directors should consult qualified counsel. The duty of care asks whether directors prepared, understood the material facts, tested management’s assumptions, and exercised independent judgment before voting. Courts applying the business judgment rule generally defer to informed, disinterested decisions made in good faith — even when the outcome is disappointing. Protection weakens when the record suggests grossly inadequate preparation, conflicts, bad faith, or a failure to deliberate.

Good process means receiving materials early enough to review them, identifying missing data and downside risks, asking direct questions and ensuring management answers, using independent expertise when needed, allowing adequate time for discussion rather than treating approval as a formality, and recording the principal issues considered and the reasons for the decision.

The duty of loyalty requires directors to place the interests of the corporation ahead of personal, financial, professional, or affiliated interests — beyond obvious self-dealing to undisclosed conflicts, misuse of confidential information, diversion of corporate opportunities, and conduct driven by an improper purpose. A conflict is not automatically a breach; the danger is an unmanaged conflict.

A director should disclose relevant facts promptly, follow the organization’s governing documents and applicable law, avoid influencing the process when appropriate, and permit disinterested decision-makers to act on a complete record. Before acting, ask: Do I or an affiliated organization have an interest? Have I disclosed all material facts? Should I leave the discussion or abstain? Would an objective outsider view the process as fair to the corporation?

Oversight is not passive. Under the line of Delaware cases associated with Caremark, directors must make a good-faith effort to establish board-level reporting and monitoring systems and must respond appropriately when warning signs appear.

Real-World Example: In Marchand v. Barnhill (2019), the Delaware Supreme Court found that allegations of no board-level food-safety reporting system could support an oversight claim against the directors of Blue Bell Creameries, where food safety was a mission-critical risk. The case underscores that for a company whose business depends on a particular area of safety or compliance, the board must ensure that area receives regular, board-level attention.

A practical oversight framework: identify mission-critical risks; assign responsibility across the board, committees, management, and advisers; set reporting expectations with useful metrics and escalation thresholds; track red flags and management’s response; and periodically test whether information reaches the board soon enough and whether actions are completed.

Cyber risk can disrupt operations, create financial and legal exposure, damage reputation, and weaken stakeholder trust. AI raises the stakes by making fraudulent messages, voices, and video more convincing. Boards should treat cybersecurity as enterprise risk: assign oversight, identify critical systems and third parties, connect technical exposure to business consequences, and require tested response and continuity plans.

Real-World Example: Attackers inserted SUNBURST malware into SolarWinds Orion updates distributed to customers in 2020. The SEC later alleged that internal assessments had identified serious weaknesses while public disclosures described risks more generally. A court dismissed most of the SEC's claims in 2024, and the SEC dismissed the remainder with prejudice in November 2025. The governance lesson is not the litigation's outcome: boards need a process that brings significant internal warnings forward, reconciles them with public disclosures, and tracks corrective action.

AI-enabled fraud exploits urgency, authority, and trust. The strongest defense is disciplined operating control: two approvals for unusual payments, independent verification through a known channel, separation of payment duties, verified vendor-account changes, and a clear way for employees to stop and report suspicious requests. Detailed controls belong in policies and procedures that can evolve as threats change—not in bylaws.

Real-World Example: In 2024, an Arup finance employee in Hong Kong transferred HK$200 million—about US$25.6 million—after a video conference appeared to include the company's CFO and other colleagues. Police said the other participants were deepfake impersonations created from publicly available material; 15 transfers were made before the deception was discovered. The case shows that a familiar face or voice is no longer proof of identity. The reported facts come from police briefings and subsequent reporting that identified Arup as the company involved.

Boards need useful governance information, not more technical detail. Directors should ask: Who is accountable? Which operations, systems, and data are critical? How are AI-enabled impersonation and payment fraud assessed? Which transactions require independent approval? How often are response and recovery plans tested? Which incidents and near misses reach the board? How are vendors and AI tools evaluated before and after deployment?

Real-World Example: MGM Resorts shut down certain systems after detecting a cybersecurity issue in September 2023, disrupting interconnected hotel and casino operations. MGM later estimated a roughly $100 million negative impact to adjusted property EBITDAR and less than $10 million in one-time expenses. The widely reported social-engineering attack on its IT help desk was not detailed in MGM's regulatory filing. The incident illustrates why authority, manual procedures, communications, financial tracking, and recovery priorities should be settled before a crisis.