Law and technology-related research has arrived in the world of corporate law, as academics begin to tackle topics like artificial intelligence (AI), increasing automation and robots, Big Data, and blockchain. In a recent working paper, I focus on a specific aspect that has not received much attention: the future of corporate management in an AI-dominated world.
My paper is divided into two major parts: The first asks whether it is feasible for AI to take over the management of corporations. I argue that it is—and will happen sooner than we might think. The second part is a thought experiment. It proceeds based on the assumption that AI will take over corporate management and explores the potential consequences, focusing on selected core corporate governance areas.
In considering whether AI will be capable of assuming corporate leadership, I first examine the nature of tasks carried out by today’s directors and managers. In a second step, I explore the capabilities of AI and whether it could assume these tasks. An important distinction in this regard is between administrative work and judgment work. The former consists of routine tasks, such as scheduling, allocation of resources, and reporting. The latter is work that requires creative, analytical, strategic, and interpersonal skills. Directors and managers engage in both types of tasks, although the board’s activities lean much more heavily towards ones that require judgment.
The division between administrative tasks and judgment work proves useful in mapping managerial tasks onto AI capabilities. The nascent literature on AI and management does not challenge the idea that administrative work will be the exclusive domain of AI in the future. There is however disagreement regarding AI’s role when it comes to judgment work. Commentators are roughly split into two camps. Those in the first—‘skeptics’—argue that machines will not be able to replace humans when it comes to judgment work, which would suggest that AI and humans will work together in leading corporations. In contrast, a second group of commentators—we can call them the ‘believers’—suggest that AI will likely match and even exceed humans in the ability to perform all sorts of tasks, including judgment work.
If the believers are correct, this would allow AI to essentially replace directors and managers and ultimately run corporations. My view is that it is more difficult to believe that humans will always be superior in some areas than imagining a future in which machines beat us at every possible task. Indeed, next generation technology that could enable management-by-machine may be much closer than we think. According to AI experts, highly advanced machine intelligence could become a reality in 20 to 30 years.
Based on the assumption that management-by-machine is possible, and will emerge at some point, the paper then explores the potential corporate governance consequences and hypothesizes a radically different framework of corporate management.
On leadership and management, the paper suggests that AI will usher in the end of the corporate board. It posits that AI will gradually replace human directors on boards, leading to ‘fused boards’ that incorporate into software or an algorithm the various roles and inputs previously provided by human. The multi-member board is set to vanish once AI can replicate the benefits of group decision-making by humans and exceed both the speed and quality of human decisions.
AI will likely also lead to fused management. This refers to the amalgamation of boards and managers, resulting in the abolition of the corporation’s traditional two-tiered governance structure. In its place, an encompassing ‘corporate management’ body could assume all of the functions of directors and managers below the board level without separating these two groups. The reasons supporting the emergence of fused management are, principally, that properly programmed corporate management software will drastically reduce agency costs, thus making the board’s main functionsto monitor managers—far less important or even obsolete. In addition, AI will not be subject to time restrictions, enabling it to carry out boards’ traditional functions and day-to-day managerial tasks.
In this new world, directors’ and officers’ personal liability will change as well. In an initial phase, when humans and AI will still work together on boards and in management, a number of challenging legal questions concerning personal liability will arise, including the extent to which human managers can and should monitor AI and to what extent they may delegate tasks to machines without exposing themselves to personal liability. In a later phase, when AI will take over completely, today’s managerial liability framework will either vanish; evolve into a system in which the AI entities can be sued; or be replaced with a system akin to product liability. Under the latter system, which is arguably more likely, corporations and shareholders, would be able to sue the developers and providers of AI software.
Finally, the paper explores the impact of an AI-dominated future on the corporate purpose. In order to manage businesses, AI will need highly specific targets, which will lead to more clearly defined aims and strict implementation of mission statements. Although AI managed entities, especially in the absence of human controllers, may provide cover for illicit activities, the rise of AI also offers the potential for significant positive changes in corporate objectives. Due to their ability to efficiently handle high degrees of complexity, AI-managed businesses will be in a better position to pursue multiple objectives simultaneously—specifically, the interests of multiple stakeholders—and optimize the outcomes of several objectives within given constraints. Shareholder-wealth maximization as a singular corporate goal may thus become obsolete. At the very least, basic corporate responsibility should be expected to improve as AI management software could be programmed to strictly adhere to all applicable laws.
While the paper does not take a normative stance but focuses on describing current and possible future developments, it seems clear that there will be a need for legal reform to accommodate changes brought about by new technologies. These reforms should be both enabling—facilitating the efficiencies and other beneficial effects of AI management—and restrictive—protecting society from negative impacts that range from loss of employment to harmful actions by rogue AI entities. From a broader corporate governance perspective, the prospect of AI management suggests the likelihood of change in the study of agency costs as an important theoretical underpinning of corporate governance theory. Agency costs between shareholders and management could be solved with AI. Instead, the development of ex ante standards for designing, controlling, and holding accountable algorithms will take center stage. On all counts, AI management seems set to initiate a new chapter for corporate law and governance.
Martin Petrin is an Associate Professor at University College London, Faculty of Laws.
This post first appeared on the Columbia Law School Blue Sky blog here.