Hey, you. Yes, you. I’ll bet you think AI always makes your work better. Well, I have bad news: a new Harvard Business School paper just dropped, and it might rattle a few of your convictions.
Not because the study says AI doesn’t work — it works very well — but because it confirms, with actual data, that if you’re bad at understanding where you can push it (and push yourself), it can make your results catastrophically worse. And it’ll do it while you sit there convinced everything is going swimmingly.
The Study: 758 BCG Consultants
The study is compelling because it involved 758 BCG consultants working on real tasks, with and without the help of an LLM.
What emerged is that the AI capability frontier is a jagged boundary: as long as you stay within it, the machine genuinely helps you produce more output, faster, and at higher quality.
The Junior Advantage: A Spectacular Catch-Up
Even more interesting: the study measured the effect on those who started further behind in terms of skill level. Their output improved significantly more than their more experienced peers.
This proves beyond any reasonable doubt that one of AI’s core effects is flattening the skills gap between people — and doing so at spectacular speed. In other words, juniors have a massive advantage in the catch-up potential the tool provides.
Darkness Beyond the Competence Boundary
The study also documented what happens the moment you step outside the LLM’s zone of competence — and you don’t have to go far. A task that looks similar but requires a type of reasoning the model handles poorly is enough.
In those cases, counterintuitively, consultants using AI still finished faster — but made significantly more errors, and more serious ones.
The Real Skill: Knowing When to Use It
The conclusion is clear: those who decide which tasks to hand to AI (and which to keep for themselves), and who have the experience and lateral thinking to steer the machine correctly, come out ahead. For everyone else? Big problemz. 😈
Dismantling the “No More Juniors” Narrative
This seriously undermines the narrative pushed by companies that have “stopped hiring juniors” because, apparently, one senior plus a swarm of AI agents is enough to maintain productivity.
The study actually shows that a junior with solid lateral thinking skills can direct the machine just as effectively — and, cherry on top, would close the gap with the senior far faster, while costing the company considerably less in the meantime.
After reading this study, I’m more convinced than ever that the policy of cutting junior hiring is a colossal act of self-sabotage.
What do you think?