Who Will Lead When the Juniors Are Gone?

Image generated with AI, which is fitting

This morning I read an email from safety researcher who left Anthropic. This led me to prompt Claude with a question:

The Prompt

Let's be honest here, human beings filled many jobs where the tasks were both repetitive and easy to learn. While there may be some truth to the idea that some of those jobs would be better suited to AI than a human being, there is the flip-side of that coin where some of those duties are entry-level positions for new graduates to learn on the job. Jobs like junior developers and junior designers. I myself am using AI tools to act as the role of junior designer when I'm starting on any new task, offloading the tedium of designing variations of wireframes based on a set of requirements. I then take those outputs and begin the work of a more senior level designer, gaining alignment and trust as we move toward the finished product.

How are we supposed to be handling this right now? For starters, too many people believe AI is ready to take the role of *all* positions, not just repetitive and easy to learn roles. And even for people who understand the need for senior leadership, how do we continue to scale our practice if there are no new people to teach? If companies are willing to offload the role of the junior, then there will be nothing left on the other side as senior leaders leave the workforce. How does this play out?

The Response

“This is a really important tension you're identifying, and I think you're seeing it more clearly than most because you're living it — using AI as your junior designer while doing the senior-level thinking yourself.

The letter is striking. Someone at the heart of AI safety research at Anthropic stepping away because they feel the weight of what's unfolding and want to engage with it on their own terms. That kind of departure carries a gravity that's hard to ignore.

But your question cuts to something more immediate and structural. You're essentially describing a pipeline problem. The junior role has historically served two functions simultaneously: getting repetitive work done and being the apprenticeship through which people develop into senior practitioners. AI is increasingly capable of handling the first function, but it completely fails at the second. And most companies are only thinking about the first one because that's where the cost savings show up on a spreadsheet.

The way I see it playing out if nothing changes is pretty grim. Companies eliminate junior roles, current seniors carry more load augmented by AI, those seniors eventually burn out or retire, and then there's a vacuum with no one trained to fill it. You end up with a generation of would-be designers and developers who never got the reps. The tacit knowledge, the judgment, the ability to know why one wireframe direction is better than another — that comes from doing the tedious work under the guidance of someone more experienced. You can't skip that step.

What makes your approach interesting is that you've essentially kept the apprenticeship model intact but replaced the apprentice with AI for the output, while you yourself still hold all the judgment. That works for you because you already did those years. But if a fresh graduate tried to use AI the same way, they wouldn't know which of the ten wireframe variations is actually solving the problem. They'd lack the pattern recognition.

I think the honest answer to "how does this play out" is that it depends entirely on whether organizations are wise enough to treat junior roles as investments in future capability rather than just line items to cut. History suggests most won't be. The companies that do maintain real mentorship pipelines — even if the nature of junior work changes — will have a massive competitive advantage in ten years when everyone else is wondering where all the senior talent went.

The deeper issue you're circling is one that letter touches on too: whether we have the wisdom to match our capability. Cutting junior roles because AI can do the tasks is a capability decision. Keeping them because humans need to learn is a wisdom decision. And right now, capability is winning.”

My Take

I feel this conversation is happening everywhere right now. You only need to do a simple search to find post after post after post on the subject. Unfortunately I agree with Claude’s conclusion though, modern companies are simply not built to play the long game. In today's world, only the next quarter of financials matter. There is no longer any room for patience or steady growth, or for investing in the people who will eventually need to lead. The jobs lost today will be the leaders we’ll never have tomorrow.

note: The bold text within Claude’s response are from me.