Porto Business School Dean: B-Schools Are Solving The Wrong AI Problem

Business schools have moved faster on AI than almost any other part of higher education. In three years we have written policies, retrained faculty, redesigned assessments and run more pilots than anyone can count. That deserves saying first.

It is also the easiest part of the problem because it is the formal layer of education: the part we already know how to govern. The harder question sits underneath it. Education was not working perfectly when ChatGPT appeared. Programs were too content-heavy and assessment often rewarded outputs more than reasoning. AI did not create that weakness. It industrialized it.

Before asking how AI fits into education, we should ask what we mean by learning. If learning is producing the right answer on time, a machine that does it in seconds is an unqualified win. If learning is becoming someone who can think, judge and act for themselves, the question changes completely.

We have known for decades that learning is a social act. Yet self-paced modules, asynchronous delivery and scaled online programs gradually moved the social part of learning to the edge. AI arrived at a layer already thinned out by the pursuit of scale, flexibility and convenience.

THE LAYER WE MANAGE, AND THE LAYER WE DON’T

Both layers teach. Only one of them is managed.

The formal layer is curriculum, course design, assessment, accreditation and AI policy. Visible, designed, measured. It is where most of the current friction appears: what students may submit, what counts as original work. Those problems are real. We have committees for them.

The informal layer is the conversation after a difficult case discussion. The study group at eleven PM. The message that begins, “Am I the only one who doesn’t understand this?” Tacit, social, unplanned and almost entirely unmeasured. None of our dashboards capture it.

MIT’s Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training points to what may be happening there. Its report notes decreased attendance at office hours, reduced online discussion and anecdotal declines in in-person study groups. Sloan students also reported higher AI use across task categories than MIT students overall, and undergraduates were the only group at MIT who reported feeling more replaceable (40%) than more capable (34%). Business education is a leading indicator here, not an exception.

One student choosing a model over a classmate is unremarkable. An entire cohort making that choice every day is a change in the architecture of a school. And almost no metric we keep will register it.

AI does not only change how students get an answer. It changes who they ask.

CRITICAL THINKING & TRUST ARE BUILT SOCIALLY

Judgment, critical thinking and trust are among the capabilities we say will matter more as AI makes analytical horsepower abundant. None is built alone. And much of their development happens not only in the classroom, but in the informal and social spaces around it.

Critical thinking is not a module we install in a student and switch on. It develops when we put our reasoning in front of others: in a study group, a conversation after class, a project team, an argument over a case. We defend a position, hear an objection, discover that our reasoning is weaker than we thought and revise it. Over time, the group’s habits of scrutiny become our own.

Trust develops in the same way. We learn who is reliable by watching how people think, how they behave under uncertainty, whether they take ownership, whether they help others, and whether their contribution to a group can be counted on. Those judgments are rarely formed through formal assessment. They emerge through repeated social interaction.

That is why the informal layer matters so much. It is not simply where students socialise. It is one of the environments in which they learn how to think with others, how to judge others, and how to become someone others can trust.

NYU psychologist Tessa West, writing in The Wall Street Journal, describes a new workplace problem: colleagues are unsure whether others are using AI, conceal their own use, and sometimes judge co-workers when they suspect it. The result is mistrust between people who still need to work together. Business schools therefore have a double responsibility. We are not only preparing students to use AI well. We are preparing them to work with other people in organizations where AI will increasingly mediate how work gets done. If learning becomes progressively more private (one student, one device, one model), we risk weakening not only the social processes through which critical thinking develops, but also those through which professional trust is built. A model can simulate questioning. It cannot be a community. It is an excellent tutor and a poor classmate.

This is where deans and academic leaders need to take ownership. We cannot outsource this question to an AI committee. Our job is not only to decide where students may use AI. It is to decide what kind of graduates we want to send into an AI-shaped world, which human capabilities must remain central, and what learning experiences actually develop them. If an AI strategy ends with a policy, a chatbot and a faculty workshop, we may have governed the tool while missing the institution.

WHAT USED TO BE FREE IS NOW THE DIFFERENTIATOR 

Corridors cost nothing. Social learning often happened as a by-product of putting people together. AI changes that. If students increasingly turn first to a model, social learning will happen only if somebody deliberately designs it.

That is why social learning can become a real differentiator. AI has made content abundant and individual tutoring dramatically cheaper. What it cannot commoditise is a room full of people who have to reason in front of each other. In the AI era, designed social learning is not the soft part of the offer. It is the offer.

There is also an equity risk. Designed human interaction is expensive. Left alone, it could become what students get only if their school can afford it and their lives have room for it — a strange destination for a technology sold to us as democratizing.

3 CHOICES FOR BUSINESS SCHOOL LEADERS

None of this is an argument for less AI. It is an argument for designing both layers instead of one.

Design AI as a collective practice, not only a private tool. Most educational AI use cases assume one student, one device, one chat window. Try three students and one model instead, with the group deciding what it believes and defending why.

Measure what AI displaces, not only what it enables. We track adoption, satisfaction and time saved; we should also track peer interaction, office hours, study groups and the strength of learning communities. What is not measured will not be defended.

Protect productive friction and human accountability. Not every difficulty in education is inefficiency. Some learning should still require discussion, defence, disagreement and moments when the tool is deliberately off the table. And some work should only count once a student is willing to stand behind it in front of other people.

At Porto Business School we are starting with ourselves. Starting this academic year, we will stop treating AI as an add-on to the curriculum and start designing it into the social and informal side of learning. We will also begin measuring the layer we have never measured, and publish what we find, including what fails. A proposal its own author will not test is not a proposal. It is a comment.

We have spent three years designing one layer of learning while the other was quietly redesigned for us. The layer we never had to design is now the one we most need to design deliberately. That is the opportunity of the next five years: not simply to put AI into education, but to build a better learning system around what humans and machines can each do best.

The schools that understand that first will not simply lead the adoption of AI. They will redefine what a business school is for.

---

José Esteves is Dean and President of the Executive Board of Porto Business School. This argument develops his keynote panel at the British Academy of Management 2026, “Who Do You Ask? How AI Is Making Learning a Private Act.” Survey figures are from Appendix C of the MIT report. The trust example draws on Tessa West’s recent Wall Street Journal article, “AI Is Creating New Trust Problems Between Colleagues.”

© Copyright 2026 Poets&Quants. All rights reserved. This article may not be republished, rewritten or otherwise distributed without written permission. To reprint or license this article or any content from Poets&Quants, please submit your request HERE.