Most public discourse about AI starts with the wrong question. People ask, “What jobs will disappear?” as if we’re waiting for some dramatic wave of layoffs that will clearly signal the turning point. That makes for great headlines, but it misses something far more important and far more subtle.
What I’ve been paying attention to lately isn’t job loss. It’s the quiet disappearance of entry-level jobs.
Every profession I’ve ever worked around had an on-ramp. Junior associates reviewed documents. Entry-level accountants reconciled accounts. Young analysts built models that were later checked by someone else. Newly minted programmers debugged code written by more experienced developers. Even physicians learned through repetition and supervised diagnostics before being trusted with full autonomy.
It wasn’t glamorous work. It was often repetitive. But historically, it was how judgment formed. It was how you learned to see patterns, make tradeoffs, and understand consequences. That repetition wasn’t inefficiency. It was apprenticeship.
What AI is beginning to compress isn’t just output. It’s the bottom rung of the job ladder. And when the bottom rung disappears, the ladder eventually collapses.
Why This Wave Feels Different
I’ve lived through automation before. In the 1990s, I worked with industrial robotics. Machines replaced human hands in tightly controlled environments. The work changed. Some roles disappeared. Others evolved. But the disruption was bounded. It was local. It moved at the speed of physical installations.
What’s happening now is different because cognitive work, which AI’s large language models do really well, doesn’t require physical replacement. Once a task can be performed in software, it can be replicated everywhere at once.
The same AI system that drafts a contract can also summarize research, generate marketing copy, analyze financial statements, and write code. It compresses entire categories of entry-level cognitive work across multiple industries simultaneously.
What is different this time is that there is no adjacent sector ready to absorb the labor displaced by AI systems. That’s new.
The Apprenticeship Problem No One Is Talking About
Let’s look at just one example, the law profession. For decades, junior associates earned their stripes doing the kind of work that most people never see. Research. Document review. Drafting early versions of briefs. It was time-consuming and expensive. It was also how they developed the judgment required to eventually lead cases.
Today, AI can do much of that entry-level work faster and far cheaper. From a pure margin standpoint, that looks like progress. If you’re running a firm, it’s hard to ignore.
But here’s the question I keep coming back to: if you stop hiring junior associates because AI can do their work, where do your future senior partners come from?
So, this isn’t about layoffs. It’s about the quiet decision not to hire. And that distinction is important to understand because a missing entry doesn’t show up in employment statistics. It just slowly hollows out the profession from the inside.
I see similar patterns emerging in accounting, consulting, software development, and even in diagnostic roles in healthcare. The early-stage work that once trained people is increasingly handled by systems rather than humans. The organization looks more efficient in the short term, but its ability to reproduce expertise over time quietly erodes.
Why Small Business Owners Should Care
You might be thinking, “That’s a Big Law problem,” or “That’s for large enterprises to figure out.” I don’t think so.
If you run a small company and rely heavily on AI to draft proposals, analyze data, write content, generate reports, or even diagnose operational issues, you’re making rational decisions. You’re lowering costs and increasing output. In the early innings, that feels like smart leverage.
But if no one on your team is learning by doing, your organization becomes thinner than it appears.
You may still have output, but you don’t have depth. You have results, but not necessarily internal capability. And over time, that means greater dependence on the platforms that generate those results for you.
Dependency on AI shifts the leverage upstream to the AI systems and away from the companies that use them.
Education Just Became a Riskier Bet
For decades, we told young people to specialize. Get the degree. Earn the certification. Acquire the credential. The market will reward you. That advice assumed something critical: that entry-level work would be available to absorb you after graduation.
But when AI begins absorbing diagnostic, analytical, and drafting work at the junior level, that path narrows. The more specialized and narrow the training, the harder it becomes to pivot if the core task is compressed by software.
This doesn’t mean education is worthless. It means the traditional contract between education and employment is shifting underneath us.
And when capable, motivated people find that the on-ramp has quietly disappeared, the frustration doesn’t just remain economic. It becomes social. Institutional trust erodes. Confidence in the educational system weakens. Not because people are lazy or entitled, but because the structure they were promised has changed.
Institutions rarely break all at once. They decay quietly.
What concerns me most short-term isn’t mass unemployment, although that is undoubtedly a third-order effect that is coming. It’s skill atrophy.
If fewer people are allowed to struggle through early-stage work, fewer people develop real judgment. Firms stop training internally. Apprenticeship ladders disappear. Dependence on the AI platform increases. The platforms capture more of the value while labor captures less.
Over time, that shifts the balance of economic power in ways that are easy to miss when you’re focused only on quarterly efficiency.
And because each individual firm is making a rational choice in isolation, no one feels responsible for the collective outcome. That’s how structural change hardens and becomes irreversible.
So What Should We Do?
The answer is not to reject AI. That would be naive and counterproductive. The answer, in my view, is to add some intentional friction into the process.
If AI drafts something, have someone at the company rewrite the results in their own words. If AI produces an analysis, have a junior employee explain the logic and assumptions behind it. If AI generates a diagnosis, make someone defend it before acting.
Use AI as a tool that accelerates learning, not as a replacement for learning. Because the first casualty of relentless optimization isn’t jobs, it’s judgment. When people stop struggling through the early reps, they never develop the instincts that experience builds. And judgment is infrastructure. You don’t notice its absence until something breaks.
This is one of the structural shifts I explore more deeply in “The Quiet Disruption,” where I examine how AI is reshaping work, power, and trust in ways that are subtle, cumulative, and easy to overlook. The book expands this idea beyond hiring and into what happens to institutions when apprenticeship ladders quietly disappear.
If you’d prefer a shorter overview, I recorded “The Learning Ladder Collapses,” a brief video that walks through this specific vector and explains why the collapse of entry-level cognitive work may be the most underappreciated consequence of AI adoption. It’s a quick synthesis of the core idea and a useful starting point for discussion.
Whether you read the book or watch the short video, the goal is the same: to slow the conversation down long enough to ask better questions before efficiency hardens into fragility.
How will you ensure AI strengthens the learning ladder inside your business rather than quietly removing it?
