[Siddharth and Najvot are students at National Law University Delhi.]
The year 2023 saw the mainstreaming of artificial intelligence (AI) through the deployment of generative AI for regular consumers. AI and machine learning (ML) systems, however, have been undergoing tremendous growth and enhancements in the past decade. They have long been seen as holding the key to revolutionising commerce by reducing costs, multiplying efficiency, identifying bottlenecks and exploring previously unchartered territories. Much has also been written about the deployment of AI in the governance of companies, particularly to solve problems that plague the system due to design flaws in the conception, framework and action of corporate governance. More particularly, the solutions that AI offers to problems of board independence, efficiency, diversity and responsiveness have been well documented.
This article explores the contours of a proposed relationship between AI and the board of directors (Board). We highlight the prospects of positive governance outcomes for companies that employ AI systems and argue against the prevalent AI-scepticism. In particular, we explore the revolutionising prospects of AI in improving Board oversight and responsiveness, which we attribute to the ability of AI to provide the Board with necessary building blocks to perform these functions. We then examine Board committees in our approach, and within the existing corporate governance framework, explore the space for autonomous AI decision-making within committees on subjects that are delegated to it by the Board.
Answers to AI-Scepticism
AI deployment in the boardroom holds significant positive prospects in promoting positive corporate governance. Particularly in the Indian context of promoter-led companies, AI can herald a new wave of management scrutiny and board independence, thereby promoting good governance and bringing a value-free, independent opinion to the oversight and management of companies. The value addition on these counts to shareholders and members by prospective AI deployment has not gone unnoticed, with extensive (and now ever-growing) acknowledgement of AI’s capabilities in this regard. On the other hand, ethics concerns related to AI too have received deserved attention. The possibility of design flaws and bias resulting from data fed to the AI begs the question: are directors well-prepared to understand and oversee the advent of AI decision-making?
Information asymmetry and paucity of data remain a perennial limitation on the capabilities of the Board to effectively oversee and monitor management performance. Furthermore, the management has a self-serving incentive to obscure important information and data from the Board to foreclose the possible scrutiny. The dexterity of AI in picking up otherwise obscure pieces of information, and in analyzing extremely large datasets will return immensely useful information for boardroom perusal, thereby equipping it to perform its statutory duties.
Another issue plaguing corporate governance is the relative laxity of the Board in the management of the company, often attributable to information asymmetry and general unawareness. The positive outcomes from the implementation of AI will improve access to information and workable inputs to directors for better decision making, whereby the Board will be equipped adequately to proactively address concerns and improve governance outcomes.
Board Oversight and Responsiveness
The duties of the Board require it to be able to independently apply its mind in exercising its powers. However, as previously discussed, the Board suffers from information and intellectual asymmetry as against the company management. Effective oversight and management, on the contrary, requires the Board to overcome these massive intellectual hurdles to be able to perform their entrusted functions. Therefore, equipping the Board to reduce information asymmetry will otherwise entail large costs.
Indian company law is heavily preoccupied with promoting Board independence. One out of every three directors in a listed public company is required to be an independent director as per Section 149(3) of the Companies Act 2013 (Act). Board independence is seen, of itself, as a valuable characteristic. We, however, argue that the law’s conception of independence is incomplete since it does not supply the independent directors with the building blocks to discharge their functions. We further discuss how AI can aid in achieving this independence.
The reasons to have independent directors, however, are easily discernible from the qualifications, duties, powers and appointment procedure of an independent director. More generally, the law remains preoccupied with patent or visible dependence/relation between the independent director as a person and the company. Simultaneously, the dominant approach to achieve independence has been to promote diversity in the Board, thereby increasing intellectual independence/diversity. The assumption implicit in this view is that increased diversity would lead to divergent ideas, thereby reducing the concentration of bias-inducing factors (that is to say, directors having diverse backgrounds would have varied opinions, tendencies and approaches, which would preclude the concentration of similar ideas and promote variance on the board).
The induction of independent directors belonging to diverse backgrounds still suffers from the same design flaws – information asymmetry and unresponsiveness. The deployment of AI on the Board, we argue, can serve as a complement to this approach. AI deployment can help provide the Board with building blocks required to discharge their oversight function. AI will augment access to information and insights that would otherwise remain obscure to the Board. These inputs when made available to the Board will redeem the full value of having an independent and diverse membership, fulfilling the policy objective of having varied perspectives. Having AI outputs will also reduce the inertia in developing an opposition to a dominant viewpoint in the Board, or to the opinion of the management. The inertial reduction also bodes well with the idea of having a board that is intellectually independent of the management. Such a Board will be more capable of rationally appraising the management’s performance, since the basis to make such decisions, will no longer be out of bounds for the Board. As David Yermack points out, boardroom AI is also likely to bridge the information asymmetry at the level of shareholders and other members.
Similarly, short notice periods also put Boards in a tight position in making important decisions. This leaves directors wanting for time to digest and make sense of all the required information in a short period of time. Here too, AI systems can promote efficiency and speed in decision-making by processing information quickly and running algorithms to produce the required inputs for directors to act upon. More importantly, the ability to process data and identification of otherwise obscure details and patterns will greatly augment the oversight capabilities of the Board. That AI and ML are ever-helpful in performing such tasks in a short span of time stands proven and acknowledged. AI algorithms can be programmed through datasets in the nature of “data” and “considerations” inputs. Boards can also obtain output “decisions” produced by the AI through applying these considerations to data inputs. Such outputs can also be dry run by tweaking (akin to switching “driving modes” on an automobile) each of these inputs as required, thereby quickly switching the background considerations as per necessary requirements, all at unprecedented speeds.
Board Committees as AI Training Grounds
The deployment of AI on company Boards will herald a revolution in the framework of governance. Naturally, it will require a reworking of the present regulations. Along also come questions about the capacity of current Boards to accurately and amply oversee AI deployment. However, technological advancements cannot wait for the law to catch up. We therefore present Board committees as an avenue for the inaugural deployment of AI decision-makers. Board committees can serve as a launchpad to harness the abilities of AI, while also limiting costs by only training directors who sit on such committees. Board committees consist of directors who exercise their powers in relation to specified subjects of the company’s governance. These majorly analytical subjects are the most suitable training grounds for deploying AI decision-makers, for the scope of error remains small in these use cases, while also returning the maximum returns by optimising the AI through in-house use, which will re-wire the AI to align with and learn the company’s values and inclinations. The proposed framework also does not fall foul of a director’s duty to apply his independent mind while observing due and reasonable care, skill and diligence as provided in Section 166(3) of the Act, for the AI decision output always remains open for directorial oversight and scrutiny. The framework would entail decision-making by the AI (thereby maximising efficiency), but it would always be in the supervisory oversight of the committee in the inaugural deployment phase. This phase would not require greater regulatory response, while also serving as a stepping-stone to provide learning inputs in furthering the next stage of AI deployment.