If AI Eliminates Junior Work, How Will Professionals Become Capable?

In my previous essay, I explored the existential threat that has been weighing on my mind since Gabe (my business partner) and I sat in our internal strategic planning meeting in the Jericho Room at RVYC.

“We’re out of business. AI will take our jobs.”

I’ll admit that few people have accused me of optimism when it comes to my own circumstances. Gabe, ever the brave thinking partner, spent the next couple of hours arguing with me (it’s my preferred collaboration style, sigh). Exasperated, he finally said, “okay, it may be all over, but it’s not all over today so let’s strategize.”

My early catastrophizing aside, the technology has developed quickly. I’m not even at an apprentice level in its application, but my concerns have shifted. I, personally, do not believe I am going to lose my job. My experience is such that I can integrate AI and provide even greater value to the clients I serve. And I’m on the latter half of my career. I only need to hold out for so long.

But what of Bellrock? How will we develop junior consultants? And what about the other professions? Accountants, lawyers, doctors, engineers?

Last post, I used the example of what was lost when creating job descriptions became easy. The point was not that job descriptions themselves have become worse. In many cases, they have become better. Clearer. Faster to produce. More complete. The loss is elsewhere. The struggle of creating them had been producing people who understood how organizations work.

That realization raises a more uncomfortable question. If AI removes the economic value of junior work, how will the next generation develop the judgment that work once created?

For most of modern professional history, junior work served two functions. It produced output, and it produced capability. Clients paid for the output, but the individuals doing the work were simultaneously learning how systems functioned. They were building pattern recognition. They were developing judgment. The work was inefficient, but it was formative.

If AI can now produce the output more efficiently than a junior human, the economic justification for assigning that work to a person disappears. But the developmental necessity does not. Organizations still need people who understand how work flows, where accountability breaks down, and how decisions ripple through systems. Those capabilities do not emerge fully formed. They are built through exposure.

I have spent a great deal of time thinking about what might replace that formative experience.

Pay to Play

I think this is most likely. Individuals will begin to pay for experience directly. Not in the form of traditional education, but in the form of access to real environments where they can perform the work that no longer has immediate economic value. In effect, they would be paying for apprenticeship. This has historical precedent and humans tend to follow previous patterns.

Many professions required extended periods of poorly compensated or even unpaid work in exchange for access to learning. The compensation for their work was not financial. It was developmental. If organizations no longer need to pay juniors to perform certain tasks, the cost of capability formation may shift back to the individual. The hubbub around unpaid internships ain’t seen nuthin’ yet. Wait ‘til you pay for them.

AI as Training Ground

Another possibility is that AI itself becomes the training ground. The same systems that can perform complex work can also generate realistic scenarios, present incomplete information, and evaluate responses. My MBA was a bit like this. We used the case study method to learn different areas of business – cases were stories of real-life scenarios, where you then pretended you were the decider and decided what to do. Even there, they recommend three levels of learning. Solve the case yourself, take that solution to your small group (learning team) to solve again, and take that solution to the classroom where the professor explores the merits of the different solutions and reveals what happened in the real scenario.

With AI, individuals could be exposed to hundreds of simulated organizational problems, attempting to diagnose and resolve them. In this model, AI would become both the generator of the work and the evaluator of the learner. It would be the learning team, the professor, and even the case writer. It would compress years of exposure into a much shorter time period. Whether simulation can fully replace the consequences and emotional weight of real decisions remains an open question, but it would undoubtedly accelerate exposure.

Those were my two ideas, and then I became curious whether I was missing something. I asked AI directly what alternate solutions might emerge.

Organizations Pay to Train

The first additional solution was that organizations themselves may need to deliberately preserve certain forms of junior work, even when AI could perform them more efficiently. Not because the work cannot be automated, but because the developmental function of performing it remains essential. Junior professionals might still conduct interviews, still document workflows, still produce first drafts of analysis, even if AI later refines those drafts. In this model, AI becomes a reviewer rather than the originator. The inefficiency is intentional. It exists not to produce output, but to produce capability. This would require discipline from leadership, because it would mean accepting short-term inefficiency in exchange for long-term competence.

Rise of the Guilds

The second solution was the emergence of organizations whose explicit purpose includes capability formation, not merely output production. Many professional firms have historically served this function implicitly. Junior consultants, analysts, and associates learned by participating in real client work under the supervision of experienced professionals. In the future, this developmental role may become more explicit. Clients may select firms not only based on price or deliverables but based on their ability to produce professionals who genuinely understand complex systems. Capability formation itself becomes part of the service offering.

Any of these solutions could be the result of the same problem statement. What do we do when the economics of the developmental process change?

Capability cannot be downloaded. It cannot be generated instantly. It is formed gradually, through exposure to complexity, ambiguity, and consequence. AI can accelerate access to information. It can accelerate execution. But it does not automagically create understanding.

For decades, the inefficiency of junior work was tolerated because it produced both output and capability. Now that AI can produce the output more efficiently, the developmental function must be preserved deliberately, or it will disappear.

The organizations that recognize this early will have an advantage. They will not merely use AI to move faster. They will redesign how capability itself is formed. They will ensure that the next generation still has a path to developing judgment, even if that path looks different than it once did.

AI may eliminate the need for certain kinds of junior work but it will not eliminate the need for capable humans. And capability, inconveniently, still must be earned.