Last week, Microsoft announced it was adding ChatGPT to its search engine. ChaptGPT has grabbed a lot of the headlines recently, but in many ways, other AI is likely to have a more profound impact for managers. In their new book Power and Prediction: The Disruptive Economics of Artificial Intelligence, Ajay Agrawal, Joshua Gans, and Avi Goldfarb, economists and professors at University of Toronto’s Rotman School of Management, argue that AI could increase the power of more senior executives by giving them a greater predictive capacity.

BRINK asked Avi Goldfarb, one of the co-authors, how they think AI will change the way companies are managed.

GOLDFARB: Recent advances in AI are all about prediction in the statistical sense: Using information that you have to fill in the information that you as a manager currently don’t have. One important way that it will change what it means to manage a company is that AI can expand the span of control of leadership. 

Scaling Decision-Making

Specifically, judgment can be codified in advance, and therefore, enable a small team’s decisions to scale. For example, consider using AI in hiring decisions. Today, HR managers make hiring decisions in most companies, predicting whether an applicant fits with the characteristics of what they have determined is a good fit for the company. 

With AI prediction, this process can be centralized. The company leadership will be able to determine what the AI should predict and the predicted score that will be needed to get hired. The individual HR managers, therefore, would lose power, and the leadership would increase their span of control over hiring.

BRINK: You make a sharp distinction between prediction and judgment in decision making — will AI also take over the judgment process too?

GOLDFARB: As long as AI is fundamentally based on machine learning tools, its power lies in prediction and not judgment. A human still needs to determine what the AI should predict and what to do with those predictions once they arrive. Someday, we might have truly artificial general intelligence that can do both prediction and judgment, but that isn’t the technology we are talking about today. 

BRINK: How will this AI revolution shift the balance of power within an organization?

GOLDFARB: AI can lead to a greater centralization of power in the leadership of an organization, as I noted above. There is also the possibility that AI will upskill millions of people, enabling them to participate more fully in the economy. For example, an AI for diagnosis could shift the balance of power in medicine from relatively well-paid doctors to other medical professionals, including nurses, pharmacists and technicians. 

I think ChatGPT and other generative models are exciting opportunities for upskilling. Millions of Americans find their careers constrained by their ability to write well.

BRINK: Is there a risk of concentrating power in fewer and fewer hands?

GOLDFARB: Yes, this is a risk, but it could also expand power, through upskilling. It is not predetermined. It depends on how AI evolves and how companies and governments choose to use it. My view is that it is more likely that AI will empower many people to get more out of their skills. 

BRINK: What is your view of ChatGPT?

GOLDFARB: I think ChatGPT and other generative models are exciting opportunities for upskilling. Millions of Americans find their careers constrained by their ability to write well. Giving everyone the tools to write better means that the other skills they have become more valuable. Of course, the picture is not entirely rosy. Even in the most optimistic scenarios, while millions might gain, thousands will lose the advantage they hold in writing. That adjustment is difficult and should not be ignored. 

BRINK: Will it matter if we cannot distinguish between what is AI-generated and what is not? (such as essay-writing in the case of a natural language AI like ChatGPT)

GOLDFARB: I don’t think it should matter. We don’t worry whether numbers were added to a calculation using a calculator or in our heads. Or if a hole was dug with a hand-held shovel or a machine. AI is a tool, and it will be a useful tool for helping us write, draw, and do a number of other activities. 

BRINK: Do you believe in the need to have a code of ethics or conduct for those who use AI?

GOLDFARB: I think the regulatory environment needs to be clear. Otherwise, a code of ethics will be difficult to enforce.

BRINK: Will the AI revolution really lead to an improvement in our productivity, given how little productivity has shifted during the digital revolution so far?

GOLDFARB: If AI achieves its potential and enables new systems of organizing, then yes. That is a core thesis of our book, Power and Prediction. AI’s potential is extraordinary, but, like other general purpose technologies, it will take time to figure out how to use it most effectively. 

BRINK: How would you design a production system from scratch to take full advantage of AI?

GOLDFARB: AI disrupts by decoupling prediction from judgment. In Power and Prediction, much of the book is dedicated to describing how to take that decoupling and use it to design a new production system. With machine prediction, the humans doing the judgment can change. Furthermore, the workflow might need to be reorganized so that different parts of the organization can work together.