What Do Managers Do? An Economist's Perspective

GAICIL Insights

Every wave of automation brings the same nervous question to the boardroom: what exactly do managers do, and could software do it instead? What Do Managers Do?¹ An Economist's Perspective, by Alan Benson and Kathryn Shaw, answers with a framework sharp enough to cut through the coming fight over artificial intelligence at work. The review organizes economic research on managers around three principles, technological determinism, skill distinction and managerial self-interest, ties them to the skills managers report on LinkedIn, and draws a line that clarifies much of the literature: managers of people versus managers of projects.

That line is the key to the AI question. Managers of people hire, retain and motivate a larger set of individual contributors who have their own performance metrics, incentive pay and autonomy; think of a manager of outside salespeople or wealth managers who keep their own client base. Managers of projects monitor, coordinate and allocate interdependent collaborators working toward a common goal, such as the managers who oversee the development of a complex new product requiring cross-functional expertise. Two different jobs, two different skill sets, and, the review proposes, skills applied in different proportions depending on the production technology in which each manager is embedded.

Follow technological determinism to its conclusion and a provocative prediction appears: AI will rewrite the job description of management. If managerial work follows production technology, any tool that changes how work is measured, coordinated and evaluated changes management itself. Tools that make individual output easy to observe strengthen people management built on metrics and incentive pay. Tools that coordinate interdependent tasks shift the burden of project management. The review does not forecast which effect wins. It supplies the categories needed to ask the question precisely.

Look at the list of value-creating activities the authors identify: hiring, retaining, training, monitoring, evaluating, allocating and supervising. Recent empirical research, they write, has made clear that better managers obtain better results along these dimensions. Read that list again and it doubles as a map of where AI is already landing, from applicant screening to performance evaluation. The framing suggests such tools are best seen as changes to the technology of managing. They lower the cost of observing worker quality and speed the allocation of effort, reshaping the managerial function without abolishing it.

Then comes the realist's twist: managerial self-interest. Managers pursue their own objectives, and organizational design has to account for that. New information technology can discipline self-interest by making decisions transparent, or open fresh room for discretion. The economic perspective treats both outcomes as live possibilities and points attention at the incentives that decide which prevails. Anyone deploying AI dashboards inside a firm should ask whose interests the dashboard serves.

The most elegant idea in the review concerns where managers sit in economic models. Compared with simple firm and worker models, managers share traits of both. They make private observations of worker quality, as firms do in employer learning models, yet they also have individual marginal products, as in classic labor market models. That dual nature explains why management resists both full automation and simple measurement. A manager is at once part of the firm's information system and part of its workforce. Automate the information role, and the human role changes shape.

The review is upbeat about research, with reason. Researchers now have unprecedented access to firms and to data sources with rich, standardized information on managers across firms. The authors expect the literature to grow through researchers' creativity and to address how managers, organizations and markets will mediate the economic transactions of the future. For policymakers anxious about AI and jobs, this is the practical point: debates about work too often treat management as a black box. This agenda opens the box, measuring what managers do and how technology shifts that work, which lets firms adopt new tools productively and lets policy target real labor market frictions.

A closing principle from the review deserves a place on every AI strategy slide: seek generalizable insights across contexts. Management is heterogeneous, and so is technology's impact on it. Careful comparison across production settings will reveal where AI augments managerial value and where it reorganizes it. The firms that measure first will manage best.