How boards should govern artificial intelligence

Strong governance accelerates AI deployment rather than slows it down. The STAR framework makes oversight practical and quarterly.

By Henk S. de Jong and Sampsa Samila

While AI is now central to corporate operations, most boards lack a structured way to oversee how their organizations deploy it, manage its risks or capture its value. That gap has gone from being a compliance problem to a leadership problem, which is getting more expensive by the quarter.

The numbers are consistent across surveys from 2024 to 2026: AI adoption is widespread, but most boards have no formal AI oversight structure, and a majority of directors struggle to keep pace with the technology they are meant to govern. In jurisdictions from the U.S. to Europe and Asia, legal doctrine is clear: Where a risk is central to operations, boards have a fiduciary obligation to put oversight systems in place. For AI, that obligation is already active.

Two ways boards fail in AI governance

Our paper on how boards should respond to AI, written with Robert Maciejko and Christoph Wollersheim, identifies two recurring failure modes:

• Under-engagement. Boards don’t understand enough to challenge management, so investment decisions get approved without scrutiny or are deferred indefinitely. The result is value leakage: scattered experiments, no scaling discipline, and competitors, especially AI-native entrants, capture what incumbents leave on the table. This failure is invisible in the short term but devastating over a three-to-five-year horizon.
• Over-acceleration. Boards push for rapid deployment before the controls, data infrastructure and operating model are ready. The result is value destruction: algorithmic bias, data breaches, misleading public claims and regulatory action. All this is costly, visible and hard to reverse.

Most organizations are exposed to both problems at the same time. They move too slowly to capture value and too carelessly to manage risk. This is the uncomfortable truth our paper addresses.

5 classic governance responsibilities in an AI context

We propose that boards ground their approach to AI in five familiar governance responsibilities, applied to a new context:

1. Purpose, ethics and compliance

When companies deploy AI to make or shape decisions about customers, employees, risks or resource allocation, they are overseeing choices about what the firm optimizes for and what it’s willing to let a machine decide. Corporate purpose must be the starting point for AI governance, not an afterthought.

Compliance is about what is legally permitted. Ethics is about what is responsible, given the corporation’s values and commitments. Purpose is about what is worth doing: which uses of AI serve the corporation’s objectives and which undermine them, even if legal and arguably ethical. GenAI may reduce the cost of execution, but the difficulty of deciding what the firm should and should not do remains a board responsibility.

2. Business model and strategy

The strategic question for boards is whether the organization is capturing value commensurate with its investment and risk exposure. For most companies, the honest answer is not yet. The gap between deploying AI and making money from it is real, and boards should be clear about why. Boards should also distinguish value creation from value capture. AI may generate productivity gains, but those gains don’t automatically flow to the firm. They may leak to customers, vendors or even competitors.

This leads to the broader strategic question that many boards haven’t confronted yet. AI may change what your company does internally, but it can also reshape the competitive environment you operate in: lowering barriers to entry, commoditizing capabilities that incumbents have treated as durable advantages, and compressing the competitive cycle from decades to months.

3. Assets, capabilities and capital allocation

A sound AI strategy can still fail if the organization lacks the assets, capabilities and capital discipline to execute it. In AI, the binding constraints are often organizational, and that’s why we treat execution capacity as a separate board responsibility.

GenAI often creates an illusion of simplicity at the point of use. What appears to employees or customers as a straightforward conversational interface may depend on substantial hidden investments in data architecture, evaluation systems, compliance infrastructure and specialist expertise. For boards, this means oversight must extend well beyond adoption metrics to the assets and governance systems that make reliable deployment possible.

4. Risk profile

AI changes the firm’s overall risk profile, across strategy, operations, compliance, reputation, cyber and conduct. The board’s job is to understand how AI alters the risks it already governs.

Good AI governance assesses the consequences of each use case, applies controls sized to the risk and maintains an auditable record of the decisions made. Boards should require management to classify AI use cases by risk tier, matching controls, human oversight and audit mechanisms to each tier. High-risk systems need explicit human decision checkpoints, escalation triggers and audit trails designed in before deployment.

5. Leadership selection, evaluation and succession

AI changes what boards should look for in senior executives. Boards should now:

• evaluate whether executives can redesign workflows and operating models, rather than just sponsor pilots.
• orchestrate across business, technology, legal, HR and risk functions.
• distinguish genuine strategic value from hype.
• manage vendor dependency and technical uncertainty.
• communicate credibly with employees about automation, reskilling and the changing nature of work.

The CEO and top team must lead organizations in which judgment, data, automation and human work are being recombined at speed. Succession planning must reflect these changed requirements.

STAR tool for practical board oversight

With those responsibilities in mind, we introduce the STAR framework to make oversight practical and quarterly. Here are the four dimensions of STAR, the questions every board should ask itself, and what boards should be aiming for: