INSEAD’s AI Bet: Teach The Manager, Not The Model

There was no shortage of artificial intelligence expertise on display last weekend in San Francisco. There was no shortage of intelligence, full stop.

Over two days at INSEAD’s AI Forum Americas September 18-19, researchers, investors, executives, entrepreneurs and technologists moved well past the now-familiar conversation about generative AI. They talked about AI-driven drug discovery, autonomous vehicles, robotics, organizational design, geopolitics, careers and the increasingly complicated division of labor between human and machine intelligence.

The forum’s title, “AI’s Next Frontier: Agents, Robots, and the Physical AI Economy,” signaled where the conversation was headed: from what AI can generate to what it can do. Machine learning systems have long spotted patterns and made predictions from data. AI agents can now pursue goals across multiple steps, using tools and other systems to act. And those capabilities are moving into the physical world through autonomous vehicles, industrial systems, drones and robots.

The most useful conclusion from the forum was also the least comfortable. No one in the room could credibly tell an MBA student what this terrain will look like five years from now. What attendees got instead was a close look at the speed and breadth of the change from the people building, using and teaching the technology.

Marc Tessier-Lavigne, co-founder, chairman and CEO of Xaira Therapeutics and Stanford’s president emeritus, opened with a conversation on AI-driven drug discovery. Later sessions ranged from automation and augmentation frameworks and human reasoning with LLMs to geopolitics, agentic AI in healthcare, board governance, Google research, AI workflows, Waymo and physical AI, careers and robotics. The agenda made its own point. AI is no longer a tech-sector topic. It is the operating condition of business.

Across those very different domains, one idea kept surfacing. Executives have to engage with AI, but they don’t have to become experts in every model, architecture or technical discipline. They need enough fluency to recognize what matters, enough judgment to challenge what they’re given and enough adaptability to assemble the right expertise around a changing problem.

For management education, that distinction matters.

NOBODY HAS THE MAP – BEWARE ANYONE WHO SAYS THEY DO

For years, business schools could teach technology as a functional domain: information systems, data analytics, digital transformation and electives for students headed into tech. Generative AI upended that model. When AI can do research, write software, design workflows, handle customers, forecast outcomes and operate machines, the question is no longer whether an MBA can teach you to use a given tool. It’s how a graduate adds value alongside systems that are faster and hold far more information than any human brain.

The forum offered no single prescription. One of its most consistent warnings, in fact, was against becoming prescriptive too soon. The models, the applications and the economics are all changing, and fast. So are regulation, labor markets and geopolitics.

Lily Fang, INSEAD’s dean of research and innovation, notes that the West has often developed software AI and robotics on partly separate tracks, while China is pursuing physical AI – the integration of robotics and artificial intelligence – with real intensity. The labor implications cut both ways. Automation threatens jobs, but as many economies approach or pass peak working-age population, Fang speculates it may also become necessary to fill the gap.

MORE THAN A COLLECTION OF ELECTIVES

INSEAD has long branded itself “The Business School for the World,” with locations across Europe, Asia, the Middle East and North America. In AI, that footprint is more than branding. Different labor markets, regulatory regimes, cultures and industries are likely to experience the technology differently.

The school has also been building institutional AI infrastructure rather than treating the subject as a set of electives. Its Human and Machine Intelligence Institute, or HUMII, launched in 2025 around research, education and impact, with human agency at the center. INSEAD has run AI Forums across its locations, launched an AI-first entrepreneurship lab, commercialized AI-enabled educational products and begun weaving AI content across the curriculum. In September, the school said its immersive learning portfolio had passed 40 AI-powered experiences. It has also drawn significant philanthropic support for HUMII and AI learning at its San Francisco Hub.

None of that means INSEAD has solved the problem, and Fang, who leads HUMII, is candid about it.

“We don’t have all the answers,” she says. “Does anyone?”

That may be the right starting point. Fang points to the proliferation of AI forms that have gained public attention since ChatGPT – from LLMs to physical AI and robotics – and the diverging paths in the U.S. and China. In her view, the LLM space may be maturing after its early explosion of attention. LLMs remain consequential for students, researchers and knowledge workers, but the frontier is moving quickly into other domains.

The educational response can’t be to chase every product release. It has to build capabilities that outlast the release cycle.

A NEW AI CORE: THREE THINGS EVERY MBA SHOULD KNOW

To that end, INSEAD is developing a required AI core course for all MBA students. Fang describes three areas as nonnegotiable.

The first is a taxonomy of AI: enough technical fluency to understand what kinds of systems exist, how they’re built and where their limits lie. An LLM can be remarkably capable and still produce inconsistent or wrong answers. Fluency, Fang says, includes knowing that confidence isn’t reliability, and that output should be scrutinized in proportion to what’s at stake.

The second is a roadmap matching categories of AI to business problems. Machine learning suits prediction, classification and optimization in data-rich settings such as engineering, finance and operations. LLMs are useful where language, synthesis, coding or unstructured information dominate. Robotics and embodied AI matter when a system has to decide and act in the physical world. The goal is to identify the problem, understand the capabilities it requires and know where to find the right technical expertise.

The third is the most interesting: what Fang calls philosophical positioning. She wants students to build the intellectual muscle to work with AI without “cognitive surrender.”

“We need to know how to challenge AI,” Fang says. “Argue with it. Set up contests. Try to give it an impossible task.”

The aim is to add friction to how we use AI. Debate, experiments, contests, verification and disagreement keep students active in the reasoning. Curiosity becomes a professional edge, and knowing the right question to ask remains a distinctly human skill.

A K-SHAPED FUTURE

The risk Fang sees is a K-shaped future for human capability. Highly educated, high-agency people may use AI to accelerate dramatically, while passive users let core reasoning skills atrophy. A task that once took a research assistant two weeks can sometimes be done in hours. That’s an extraordinary gain. But speed isn’t the same as thought, reasoning or growth.

Her concern reaches into the classroom. Fang already sees signs of AI fatigue among students, and a renewed appreciation for pencil and paper, whiteboards, flip charts and other tactile forms of learning. As more interaction becomes synthetic, the physical environment may matter more for holding attention, building memory and making knowledge stick.

It’s a striking inversion. The business school at the forefront of AI innovation is also the one that knows when to put the laptop away.

THE GENERALIST STRIKES BACK

Another theme recurred over the two days: the value of the high-functioning generalist.

Asked what a chief scientific officer needs when the field is changing this quickly, Tessier-Lavigne and Andrea van Elsas didn’t describe omniscience. They described practice, adaptability, the ability to see what the organization needs and the judgment to assemble the right team.

Alex “Sandy” Pentland, a Stanford HAI fellow and founder of MIT’s Human Dynamics Laboratory, and Michael Liou, a venture investor and adviser to AI and technology companies, pushed the argument further in a session titled “Innovation Pendulum & Shared Wisdom.” Their premise was that humanity’s advantage has never rested on individual intelligence alone. It rests on the ability to share experience, test ideas socially and turn distributed knowledge into collective understanding.

They championed truth-seeking, curiosity and high-EQ generalists over the notion that every leader must become a technical expert or risk irrelevance. Plenty of technologies have arrived with predictions of intellectual or social collapse. Each time, people have asked the same question: What can I do with this new tool?

For MBA students, that is both reassuring and demanding. Technical literacy matters. So do communication, judgment, emotional intelligence, network-building and the ability to move across domains. No manager, generalist or specialist, can be the expert in every room.

THE ABILITY TO KEEP THINKING

Vivienne Ming, chief scientist at Possibility Sciences, closed the forum with a keynote on the future of work and human potential. Her work draws a useful distinction. Chess is a well-posed problem, with clear rules and a defined outcome. Business problems often aren’t, because goals and constraints shift as leaders work through them. AI excels at well-posed problems and is getting better at ill-posed ones.

But consider the questions leaders actually face. Which market should we enter? What should we build? Which risk is acceptable? Whose interests matter? When should we stop optimizing for efficiency because resilience, trust or legitimacy matters more?

Answering them requires someone to decide what matters, weigh competing interests and own the choice. That is where business education still earns its keep.

THE HONEST, UNCOMFORTABLE ANSWER

There is no consensus on where AI is headed. That shouldn’t paralyze business schools.

The most durable AI education gives students a working taxonomy of the technology, the ability to match systems to problems, the judgment to know when not to automate and enough intellectual resistance to work with machines without surrendering the act of thinking. It teaches them to stay curious when answers are cheap and to verify when output is effortless. To tell speed from wisdom. To protect human relationships. And to understand that a prediction is not a decision.

Used well, AI can multiply human capacity by cutting drag from routine processes, accelerating discovery, widening access to expertise and freeing people for the problems that still require people.

The AI-ready MBA, then, is a leader fluent enough to use the machines, skeptical enough to challenge them, adaptable enough to keep learning and human enough to know what not to surrender.

DON’T MISS POETS&QUANTS’ 2025-2026 INTERNATIONAL MBA RANKING

© 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.