AI isn’t killing junior talent... companies are.

That epiphany hit me in the middle of a conversation with a few colleagues earlier this week. We were talking about Anthropic’s new Opus 4.5 model, and there was a lot of chatter that devs were already frustrated because they’d blown through their credits again, not because they were pushing the boundaries of anything, but because they couldn’t get the model to give them what they wanted. They kept firing off prompts without stopping to think about what they were actually asking.

I picked up on the pattern that the people who usually max out their limits are the people who treat AI like some genie in a lamp that should just “get it” from vibes. They expect the model to fill in the blanks between what they mean and what they type, and when the model inevitably misses, they waste another prompt... and another... and another. It’s not a capacity issue. It’s a direction issue.

That’s when the larger realization hit me. The people who struggle most with AI aren’t just the ones with the least experience. It’s the people who never learned how to give precise direction in the first place. And the people who thrive are the ones who already spent years developing the underlying craft behind the work AI is now able to accelerate.

When you’ve done the manual version of a job, actually lived inside the strategy, the design, the writing, the engineering, you build a memory for patterns and outcomes. You develop a sense of what “good” looks like and why. So when you start using AI, it feels like having a bottomless bench of people you can guide. You understand what to ask for, what to correct, and how to steer the whole thing toward something real.

I felt this firsthand recently. I was working on a new product alongside our core one, and normally the process of turning insights into strategy, strategy into story, and story into a visual identity is a long and messy human exercise. This time it compressed dramatically. What usually takes weeks happened in hours, and the quality actually jumped. Not because the model did the thinking for me, but because I’ve done enough of this work to know how to move through the thinking with it.

That’s the real divide AI is opening. Seniors accelerate because they’ve built the reps. Juniors aren’t falling behind because they lack value. They’re falling behind because they’re being pushed straight into a directing role without ever having done the work they’re directing. It’s like skipping the “learn to play” part and being asked to coach the team on day one.

This doesn’t mean junior roles disappear. It means they shift. In product design, for example, applying a design system used to be the perfect junior task. Now AI does it instantly. But the first draft is always a mess, and someone still needs to look at it and figure out what’s wrong, why it’s wrong, and what to tell the model to fix. That’s where the judgment lives. That’s the new craft.

Most companies, though, are interpreting AI as a reason to shrink teams and rely only on seniors who can “do more with less.” And yes, a small senior team plus AI looks efficient in the short term. But once every company has access to the same models, the same assistants, the same workflows, that advantage evaporates. When everyone can “do more with less,” the new competitive pressure becomes “do far more, far faster.” That shift forces companies back into the same truth they thought they could escape... you need more people who know how to wield the tools, not fewer.

There’s a pattern here that shows up outside of tech too, and basketball is a good example. In North America, the talent pool is so deep that teams can afford to cut players quickly and cycle through options until they find the perfect fit. In Europe, the pool is smaller, so teams build systems that focus on developing players over time. They take raw potential and refine it until it becomes something exceptional. It’s no accident that many of today’s top NBA stars, including multiple MVPs, come from those development cultures.

AI pushes companies into the same choice. You can churn talent and hope the tools fill in the rest, or you can build a culture that develops people into powerful directors. If the leverage is equal, the differentiator becomes talent density. Not how many people you have, but how many people know how to think.

This forces founders and operators to rethink what “junior” even means. You’re not hiring someone to do the grunt work anymore. The tools do the grunt work. You’re hiring someone with the curiosity and appetite to understand value, ask sharper questions, and learn how to express their thinking in a way the tools can actually use. That’s an entrepreneurial mindset, and it can be trained if you design for it. And it has to be trained, because the speed of AI adoption is faster than any other technological shift we’ve lived through. Companies that cut junior development today will wake up in a few years realizing they have nobody left who can operate the systems at the level the market now demands.

And for young people trying to figure out where they fit in all this, you’re not being left behind. You’re entering at the exact moment the entire ladder is being rebuilt. You have access to the most powerful apprenticeship engine we’ve ever had. If you take it seriously, if you use AI to practice, explore, and understand the pieces of the craft that sit before and after your role, you can move faster than any generation before you. The reps still matter. Judgment still matters. But you can build them differently now.

This is the part that rarely gets said. AI doesn’t remove the need for talent. It shifts where talent creates value. And the companies that learn to develop that value, and the individuals who learn how to direct it, are the ones who will define the next era of work.

‍