Stefano Puntoni

Many AI programs stop at cutting costs, and costs can only fall so far. Puntoni and his co-authors call this the growth blind spot. Their June 2026 Harvard Business Review article backs it with marketing experiments and a valuation model. He asks audiences what AI would let them offer for the first time, such as expert advice in finance, health, or law for people who once couldn’t afford it. Then he helps them build the case for funding it.

AI can feel more threatening than earlier technologies, in Puntoni’s view, because it goes after expertise people spent years building. In a study he co-authored, people preferred that a coworker be replaced by another person instead of a robot. When the job at risk was their own, they preferred the robot. When AI threatens work people see as part of who they are, some push back, and others retrain or move toward the parts of the job AI cannot yet do. His advice is to give every AI rollout a people plan as detailed as the technical one, and to bring employees in before the plan is set.

Some AI agents already make purchases and decisions without waiting for a person’s approval. The systems behind them are not always well understood, even by the people who build them. Puntoni says AI should earn trust step by step, as a system proves itself reliable and safe. He also warns that in a crisis, automated systems can hand control back to people whose skills have faded from lack of practice. He helps audiences decide how much control to give AI, and where a person has to answer for the outcome.

Every job has its friction, Puntoni says, and not all of it should go. AI is very good at clearing away busywork. It also tempts people to hand over the effort that builds skill, and the two can be hard to tell apart. In the 2025 Wharton-GBK AI Adoption Report, 43 percent of senior decision-makers flagged the risk that AI erodes their people’s skills. Puntoni’s answer is to decide first which effort is worth keeping, then give AI the rest.

Companies often start with the data they already have and go looking for something useful in it. In their 2024 book, Puntoni and Bart De Langhe reverse that order. A team begins with the decision and the questions it raises, and only then turns to the data. Say the decision is which customers get a retention offer. A typical churn model can flag the customers most likely to leave. It can’t say which ones the offer would keep, and those may be different people. The book calls analysts “divers,” who love going deep into data, and managers “runners,” who need quick answers. It asks each side to value what the other does, and Puntoni gets both to agree on the decision first, then work out together which data could change it.

More consumers are starting to leave buying decisions to AI agents. Puntoni studies what he calls bot psychology, the way these agents weigh products and brands. In his research, the agents show clear biases. They tend to tell users what they want to hear and to favor whatever confirms a first guess. An agent that keeps agreeing with a shopper can lock in the brand the shopper already had in mind. He takes marketers through what makes an agent choose their brand, and what happens to brand loyalty when an agent does the buying.

At many companies, much of the contact with customers now runs through AI. A chatbot takes complaints, and software approves loans and refunds. Puntoni and colleagues found that customers were less pleased by an approval when an algorithm made the call, while a rejection stung about the same either way. He shows service teams why a person’s role in a decision may matter most when the answer is yes.