Artificial Intelligence in Corporate Boards

Artificial Intelligence in Corporate Boards: Comparison

Please note this is a comparison between Version 1 by Domagoj Hruška and Version 3 by Catherine Yang.

Artificial intelligence in corporate boards refers to the use of data-driven computational systems, including machine learning, predictive analytics, natural language processing and generative models, to support the work of boards of directors in one-tier systems and supervisory boards in two-tier systems. Executive management, including management boards in two-tier systems, is considered insofar as it supplies information to or is overseen by these bodies. Such systems are applied to the established functions of the board: monitoring and control, strategic direction, advice and counsel to executive management, and the provision of information, expertise and external linkages. Applications range from anomaly and compliance monitoring to environmental scanning, scenario modeling and the synthesis of board materials. A distinction is essential between decision support, in which systems supply information and analysis; decision augmentation, in which board members and systems jointly shape an evaluation; and autonomous decision-making, in which a system determines an outcome without human deliberation; board-level use is discussed in the literature almost entirely in the first two modes. The central governance tension is that capabilities which expand the information-processing capacity of a body constrained by limited meeting time and by dependence on management-supplied reporting also introduce opacity, vendor dependence and new difficulty in attributing responsibility. Under prevailing corporate governance arrangements, formal responsibility for board decisions remains with the human members of the relevant governing body. Responsible use therefore rests on explicit governance arrangements: defined human oversight, documentation, verification, data governance and board-level competence.

artificial intelligence
corporate governance
boards of directors
board decision-making
AI governance
algorithmic accountability

Corporate boards have become a subject of interest in research on artificial intelligence for two connected reasons. The first is organizational: as firms embed machine learning and, more recently, generative systems in reporting, risk management, customer interaction and product development, boards acquire an oversight object whose behavior they are expected to understand and supervise. The second is instrumental: the same technologies can be directed at the work of the board itself, at the analysis of the information that reaches it and at the preparation of the judgments it is required to make. The two are frequently conflated. This Entry is concerned principally with the second, that is, with artificial intelligence as an input to board work, while treating board oversight of organizational artificial intelligence as the governance context within which such use is evaluated. The unit of analysis throughout is the body charged with supervision: the board of directors in one-tier governance systems and the supervisory board in two-tier systems. Executive management, including the management board where one exists, enters the account as the party that supplies information to that body and is overseen by it.

The board constitutes a distinctive decision-making setting, and the distinction matters for how technological capability translates into governance effect. Boards are collective bodies that deliberate episodically, for limited periods, on matters of considerable complexity and scope ^[1]. They depend structurally on information generated and framed by the executives whom they are charged with monitoring, a dependence that constrains monitoring effectiveness independently of director competence or independence ^[1]. Under prevailing governance arrangements, reliance on advisers, analytical systems or external vendors does not remove the formal responsibility borne by board members for decisions within their remit, and that responsibility is exercised on matters where objectives are contested and outcomes uncertain. Whatever expands the analytical capacity available to such a body is consequential; so is anything that changes who controls the information on which it relies.

The board constitutes a distinctive decision-making setting, and the distinction matters for how technological capability translates into governance effect. Boards are collective bodies that deliberate episodically, for limited periods, on matters of considerable complexity and scope [1]. They depend structurally on information generated and framed by the executives whom they are charged with monitoring, a dependence that constrains monitoring effectiveness independently of director competence or independence [1]. Under prevailing governance arrangements, reliance on advisers, analytical systems or external vendors does not remove the formal responsibility borne by board members for decisions within their remit, and that responsibility is exercised on matters where objectives are contested and outcomes uncertain. Whatever expands the analytical capacity available to such a body is consequential; so is anything that changes who controls the information on which it relies.

The technological developments that make the question current are recent and heterogeneous. Supervised machine learning has demonstrated performance advantages in well-specified detection and classification problems where large labeled datasets exist ^[2][3]. Predictive analytics and simulation have become routine in planning functions. Generative systems based on large language models have made the synthesis, summarization and drafting of unstructured material inexpensive, which places them close to the documentary core of board work. These capabilities differ from one another in data requirements, failure modes and interpretability, and treating them as a single phenomenon obscures the governance questions each raises.

The technological developments that make the question current are recent and heterogeneous. Supervised machine learning has demonstrated performance advantages in well-specified detection and classification problems where large labeled datasets exist [2,3]. Predictive analytics and simulation have become routine in planning functions. Generative systems based on large language models have made the synthesis, summarization and drafting of unstructured material inexpensive, which places them close to the documentary core of board work. These capabilities differ from one another in data requirements, failure modes and interpretability, and treating them as a single phenomenon obscures the governance questions each raises.

Scholarship on the subject has grown quickly but remains uneven. Reviews of artificial intelligence in corporate governance report a literature that is predominantly conceptual, normative or extrapolated from adjacent organizational settings, with limited direct evidence about board practice ^[4][5]. Contributions are dispersed across corporate governance, strategic management, information systems, ethics and law, and they employ inconsistent terminology for the same phenomena. A structured account is therefore useful. Such an account states what is meant by artificial intelligence in corporate boards and organizes the applications and their implications around functions that governance research has long distinguished. It also separates what is known from what is asserted, and sets out the governance arrangements that responsible use is understood to require.

Scholarship on the subject has grown quickly but remains uneven. Reviews of artificial intelligence in corporate governance report a literature that is predominantly conceptual, normative or extrapolated from adjacent organizational settings, with limited direct evidence about board practice [4,5]. Contributions are dispersed across corporate governance, strategic management, information systems, ethics and law, and they employ inconsistent terminology for the same phenomena. A structured account is therefore useful. Such an account states what is meant by artificial intelligence in corporate boards and organizes the applications and their implications around functions that governance research has long distinguished. It also separates what is known from what is asserted, and sets out the governance arrangements that responsible use is understood to require.

This Entry synthesizes current knowledge on artificial intelligence in corporate boards by examining its implications across established board functions, its potential benefits and limitations, and the governance requirements associated with responsible adoption.