From Problem-Solving to Problem-Framing: What Business Schools Must Rethink in the Age of AI

There is something strange going on in education, says Barbara Oakley on Substack.

On the one hand, we have decades of cognitive research showing that beginners learn best through structured guidance: clear explanations, worked examples, immediate feedback. John Hattie’s synthesis of evidence makes this point unambiguous. “Sit back and watch” is not a pedagogy. It is an abdication of duty. Her argument: The Constructivist approach to teaching and learning is not working.

But: something equally problematic happens once learners are no longer beginners.

The cohort I know best is 25-year-olds in an MBA program. They are capable, motivated, and ambitious. By the time they arrive in business school, they have been trained—often very successfully—to solve well-defined problems. Give them the variables. Define the constraints. Provide the data. They will optimize scarce resources with impressive efficiency.

But that is precisely the issue.

As Eric Mazur has pointed out in his physics teaching, students become very good at solving problems that are presented to them. The PISA “drip problem” is a classic example: all variables carefully specified, the task neatly contained. Success depends on executing known procedures correctly.

In real life, however, we spend perhaps 80 percent of our time figuring out what the problem is in the first place.

This challenge is not new. During the Sputnik era in the United States, a physics student was asked to measure the height of a building using a barometer. His answer was simple: go to the building, knock on the housekeeper’s door, and offer the barometer in exchange for the height of the building.

Brilliant problem-framing. Useless physics.

Today, the stakes are higher.

If I understand Ethan Mollick and others correctly in their work on the “jagged edge” of technology, AI systems are already outperforming most humans in well-defined, linear domains. Even on a bad day, a large language model may outperform the majority of humans on their best day in tasks that resemble exam questions: structured inputs, bounded outputs, narrow solution spaces.

So we need to ask an uncomfortable question: how important will linear-domain skills be in the future?

Business schools traditionally train students to optimize. Given constraints, maximize output. Given data, find the pattern. Given a cost function, minimize it. These are valuable skills.

But in a world of intelligent augmentation—IA rather than artificial intelligence—the comparative advantage shifts.

Machines will clean the data. They will crunch the numbers. They will detect subtle correlations in niche datasets. They will generate ten plausible strategies in seconds.

What they do not do well is navigate messy, non-linear domains with multiple competing set points—precisely the domains that define our era: climate systems, social systems, political economies, public health, organizational culture. In these fields, “solutions” often create new problems. Interventions have second- and third-order effects. Optimization in one dimension destabilizes another.

My hypothesis is that the skills we will need most will not primarily come from those who scored at the very top of the global PISA mathematics rankings.

We will need people capable of framing ambiguous situations, defining meaningful questions, and understanding that the problem statement itself is often the core intellectual task. We will need individuals comfortable with complexity, trade-offs, and uncertainty. We will need graduates who can work with AI not as a crutch, but as an amplifier.

This does not invalidate the evidence on explicit instruction. Beginners still need guidance. Foundational knowledge still matters. Without structured input, there is nothing to augment.

But at advanced stages of education, especially in business schools, an exclusive focus on well-defined problem-solving becomes dangerous. It risks producing graduates who are highly efficient within artificial boundaries and helpless outside them.

The future will not reward those who can best solve the problems they are given; it will reward those who can decide which problems are worth solving