We Said We Wanted Critical Thinkers. Then We Eliminated Where They Come From

A few weeks ago I wrote that organizations keep demanding critical thinking while failing to define what it actually is. The if-this-then-that chain. Pattern recognition. Hypothesis testing. Skepticism as method, not skepticism as identity.

Here’s the follow-on problem nobody wants to say out loud: we’re not just failing to define the skill. We’re systematically eliminating the conditions where it develops.

The World Economic Forum published a piece this month that should be required reading for anyone running a talent strategy. The argument is straightforward: the leaders organizations will need in five years are the entry-level hires they’re not making right now. Harvard research cited in the piece found that at companies actively adopting generative AI, entry-level hiring dropped 80% per quarter since 2023. Not a dip. A structural shift.

The WEF frames this as a pipeline problem — fewer rungs on the ladder, fewer people developing into leaders. That framing is right as far as it goes. But it stops short of the harder point.

Entry-level roles were never only about output. They were practice reps. The junior analyst who builds the model manually doesn’t just learn the model — they build the instinct for when the model is wrong. The new hire who drafts the memo and gets it marked up learns judgment by watching their thinking get challenged. The early-career professional who handles the messy client situation develops the pattern recognition that no training course can replicate, because pattern recognition only develops when you’ve seen enough patterns to recognize one. You can’t shortcut that. You can’t hire for it in someone who never had the chance to build it. And you absolutely cannot prompt-engineer your way into a workforce that has it.

I’ve spent twenty-plus years watching organizations confuse activity with development. Keeping someone busy is not the same as building their capacity to think. What entry-level work provided — at its best — was a structured environment for low-stakes failure, iteration, and the gradual internalization of cause-and-effect thinking. That’s exactly what gets lost when the task gets handed to an AI model and the human’s job becomes reviewing the output instead of producing it.

Reviewing someone else’s thinking is not the same cognitive act as generating your own. The unwritten danger of AI.

The WEF piece offers a five-point framework in response: redesign entry-level work into “judgment loops,” build AI-era apprenticeships, hold managers accountable for development, and so on. I don’t disagree with any of it directionally. But frameworks don’t fix cultural defaults, and the default right now is efficiency over development. The organizations cutting entry-level roles aren’t doing it because they forgot about leadership pipelines. They’re doing it because quarterly output looks better without those headcount costs, and the consequence won’t show up on any dashboard for another five to seven years. By the time it does, the people who made the decision will be somewhere else.

That’s what makes this a slow drain rather than a trend. Trends get measured. A slow drain goes unnoticed until the tank is empty.

So here’s the question worth sitting with if you’re in talent, L&D, or running a business unit: when you hand a task to an AI tool instead of a junior hire, what are you actually optimizing for? If the answer is this quarter’s throughput, you’re probably right in the short term. But somewhere downstream, you’ll need someone who can look at an AI output and know, instinctively, that something in the chain doesn’t add up. That instinct doesn’t appear on demand. It gets built through repetition, failure, and the kind of low-stakes thinking work that entry-level roles used to provide.

We said we wanted critical thinkers. We just stopped building them.