The Apprenticeship Problem Was Never What We Thought It Was

McKinsey's global managing partner recently told HBR's IdeaCast that AI is highly effective at linear problem solving but far less capable of genuine out-of-the-box thinking.

He is probably right.

He also unintentionally described much of what junior consultants have spent decades doing.

Bob Sternfels made the observation while discussing how McKinsey is rethinking its talent model in response to AI — moving away from narrow economics and engineering profiles toward people better able to reason through ambiguity, exercise judgment, and frame problems creatively.

The discussion has largely been interpreted as a warning about the erosion of the traditional consulting apprenticeship model. AI is automating the research, modelling, synthesis, and slide-building work that historically occupied junior consultants. If the bottom of the pyramid disappears, the industry risks losing the mechanism that produced future senior practitioners.

That concern is real. But it rests on an assumption that deserves closer examination:

What the Pyramid Got Right — and Wrong

The traditional consulting pyramid was not irrational. It solved several important problems simultaneously: it created leverage, lowered delivery costs, exposed junior consultants to real client situations early in their careers, and did, over time, produce many exceptional senior practitioners.

And it contained genuine moments of structured thinking. Engagement kickoffs at the best firms were often explicitly pedagogical — the team would work through the overall question they had been hired to answer, decompose it into sub-questions, and map what evidence was needed to resolve each one. That is problem decomposition in its most useful form — the kind of structured reasoning that Kepner and Tregoe had been formalizing since the 1950s and that McKinsey's own problem-solving methodology had codified into firm practice.

But the mechanism through which judgment developed was always less deliberate than firms liked to believe — and far more dependent on which partner you happened to work for.

Those structured kickoff moments receded once execution began. What filled the weeks and months that followed was largely the delivery work: the data gathering, the model building, the slide construction. Junior consultants learned primarily through exposure — to meetings, to client politics, to partner decision-making, to difficult commercial situations. Sometimes that exposure produced remarkable judgment. Sometimes it produced highly capable analysts who had mastered execution but never fully internalized how senior practitioners actually structured ambiguous decisions.

The distinction matters. Because proximity to judgment is not the same thing as learning judgment. And an episodic scaffold at the start of an engagement, however well designed, is not a curriculum.

What Harvard Understood Earlier

More than a century ago, Harvard Business School made a different educational bet.

Rather than teaching management primarily through lectures, it built its curriculum around cases: compressed simulations of messy business decisions with incomplete information, competing incentives, ambiguous trade-offs, and no clean answer. The objective was not merely to reach a recommendation. It was to develop the reasoning process required to navigate ambiguity under pressure.

The underlying insight was profound: judgment develops less through repetitive execution than through deliberate engagement with difficult decisions.

Over the following decades, a broader movement toward making business reasoning more explicit gathered force. Porter gave strategists a rigorous framework for competitive analysis. Kepner and Tregoe formalized problem decomposition and decision analysis as teachable disciplines. Raiffa's work on decision theory under uncertainty made trade-offs, probabilities, and consequences explicit enough to examine and challenge systematically. Each contribution pushed in the same direction: the idea that at least part of what organizations treated as instinctive judgment could be externalized, structured, and improved.

The consulting pyramid never fully operationalized this insight. It assumed that enough exposure to difficult work would eventually produce senior judgment organically. Sometimes it did. But often the learning process remained accidental, uneven, and heavily dependent on individual mentorship quality.

It is worth noting that HBS itself was not immune to the same problem. The pure case method — resistant to frameworks at the beginning and takeaways at the end — could leave students unclear about what they were supposed to have learned. The judgment that developed often came less from any single class than from the sheer cumulative volume of cases and the pressure to cut through to a defensible position under time constraint. That is a real form of learning. But it is also, in its own way, learning by exposure rather than by design.

The Honest Accounting of Grunt Work

The question is not whether analyst-level work built discipline. It often did.

The question is whether the day-to-day execution work — as distinct from the structured thinking moments at an engagement's beginning — reliably developed the capabilities that ultimately make senior practitioners valuable.

Much junior consulting work historically involved cleaning and restructuring client data, building presentations from senior direction, conducting market research, formatting models, and running analyses designed by someone else. These tasks require intelligence, diligence, and stamina. They develop rigour. But they do not necessarily develop problem framing, ambiguity navigation, political judgment, strategic reframing, or the ability to recognize when an analytically correct answer is commercially wrong.

Those capabilities were developed in the structured moments — the kickoffs, the partner reviews, the client conversations where assumptions were challenged in real time. But those moments were episodic. The ratio of structured thinking to execution in a typical engagement was heavily weighted toward execution. And the quality of the structured moments varied enormously depending on the partner, the team, and the engagement type.

Which meant the consulting pyramid was never purely a learning system. It was also a staffing and economic system. And while it did produce outstanding senior leaders, the consistency with which it did so was far lower than firms implied — and far more dependent on individual luck than institutional design.

What AI Is Actually Disrupting

The standard narrative is that AI is destroying a functioning apprenticeship machine.

That is only partially true.

AI is unquestionably compressing the bottom of the pyramid. Many of the tasks historically assigned to junior consultants can now be performed dramatically faster by AI systems. The economic implications are enormous.

But AI is not destroying a perfectly designed capability-building model. It is exposing the fact that the industry often conflated doing the work with learning from the work. Those are not automatically the same thing.

This creates a more important strategic question than most firms are currently asking: if AI handles much of the repetitive analytical work, how does the next generation of senior judgment get developed deliberately rather than accidentally? That is the real apprenticeship problem.

The Mechanism the Pyramid Was Missing

The firms that navigate this transition successfully will likely rethink consulting development much more explicitly. They will stop assuming that exposure alone reliably produces judgment. Instead, they will build environments that make reasoning visible.

That means structured decision frameworks, simulation environments, probabilistic reasoning tools, negotiation walkthroughs, and deliberate engagement debriefs — systems that externalize how experienced practitioners actually think, not just what they conclude.

A junior consultant working through a structured commercial decision is not simply generating an answer. They are learning how trade-offs are framed, how uncertainty is handled, how competing incentives are balanced, and how reasoning changes under different assumptions. That creates something the traditional pyramid often lacked: pedagogically dense learning.

Mentorship also changes fundamentally. Instead of inferring a junior consultant's thinking indirectly over months of work, senior practitioners can review explicit reasoning processes directly — what assumptions were made, what trade-offs were prioritized, where uncertainty entered, why a recommendation changed. The object of mentorship becomes the structure of thought itself.

The Unexpected Opportunity AI Creates

Ironically, the same technology compressing the bottom of the consulting pyramid may also make professional judgment easier to teach systematically than before — reducing repetitive volume while increasing the density of deliberate reasoning exposure. That is a very different future from the one most industry commentary currently imagines.

The firms that thrive will probably not be the ones trying to preserve the traditional pyramid in slightly modified form, nor those that simply become leaner structures with fewer junior staff. The firms that succeed will likely be the ones that ask a more fundamental question:

That is ultimately the question AI is forcing consulting firms to confront. Not how to preserve the pyramid. But whether the pyramid was ever the optimal mechanism for building judgment in the first place.