The Apprenticeship Inversion

What happens to mentorship when new hires can produce expert-looking work on day one, and how leaders rebuild the learning curve.

The premise

For generations, people learned a trade the same way. They started with the simple, repetitive work. They drafted the first version. They pulled the research. They formatted the report. A senior person corrected them, again and again.

That work was tedious. It was also the curriculum. Every draft that came back marked up taught something. Over a few years, the repetitions turned into judgment.

SI now does most of that work. A new hire can produce a polished first draft, a clean analysis, or a complete proposal in their first week. The output looks like an expert made it.

That's the inversion. The output used to show what someone had learned. Now it shows what the tool can do. The work looks advanced while the person behind it is still a beginner.

Why it matters

The early evidence points the same direction.

Stanford researchers Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen studied payroll data from the largest payroll provider in the United States. Since generative SI spread, workers aged 22 to 25 in the most SI-exposed jobs saw a 13 percent relative decline in employment. Experienced workers in the same jobs held steady or grew. The researchers point to tacit knowledge, the things experienced people learn on the job that never get written down.

Matt Beane at UC Santa Barbara saw an earlier version of this in robotic surgery. The robot let senior surgeons operate alone. Trainees were pushed to the side. Only a minority built real skill, mostly through unofficial workarounds he calls "shadow learning." The approved path to competence stopped working.

The pattern is the same in both. When technology makes experts efficient enough to skip the beginner, the beginner stops learning. Companies save time today and run short on experienced people later.

For a growing business, that's a real problem. Today's junior hire is the person who should be reviewing work, training others, and earning client trust in three years. If the learning curve disappears, so does that bench.

How leaders rebuild the curve

SI didn't remove the need to learn. It removed the structure that made learning automatic. Leaders have to put that structure back on purpose.

1. Separate production from practice

Not every task needs to be done the fastest way. Pick the tasks that build judgment and have juniors do some of them by hand, on purpose. Write the first draft before using SI. Build the analysis before asking SI to check it.

Treat this as training time, not lost time. It's the repetition the old grunt work used to provide.

2. Make the reasoning the deliverable

When the output is easy, the reasoning is what matters. Ask juniors to hand in a short explanation with their work:

• Why this approach?
• What did you reject?
• What are you unsure about?

You can't tell from a polished document whether someone understands it. You can tell from how they explain it.

3. Bring juniors into the real work

SI handles the simple work, so juniors need access to the complex work sooner. Let them sit in on client calls. Have them watch a senior make a hard decision and explain it out loud afterward.

Beane's research points here. The fix is pairing experts and beginners on challenging work, not keeping beginners on tasks a machine can do.

4. Turn checking into training

Reviewing SI output is one of the best learning jobs a junior can have. To catch an error, you have to understand what correct looks like.

Give juniors real verification work: confirming sources, checking facts against client records, catching claims that don't fit. Then have a senior review what they caught and what they missed. Both sides learn.

5. Count mentoring as part of senior work

If seniors are measured only on output, they'll do the work themselves. SI makes that easy. Juniors get left out.

Make developing people part of how seniors are evaluated. Beane suggests exactly this: judge experts partly on how well they develop the people under them. What gets measured gets time.

6. Expect slower early output, on purpose

A junior who is actually learning will be slower than one who's passing SI work through. That's the cost of building judgment.

Plan for it. Set expectations around growth in the first months, not raw volume. The payoff is someone who can stand behind their work, and eventually train the next person.

The takeaway

SI didn't flatten the learning curve. It hid it. New hires can produce expert-looking work long before they can explain it, defend it, or catch its errors.

The old apprenticeship happened by accident, through years of grunt work. The new one has to be designed. Keep some practice by hand. Review reasoning, not just output. Bring juniors into hard work early. Use verification as training. Reward seniors for developing people.

Businesses that do this will have experienced people when they need them. Businesses that don't will have plenty of output and no one ready to lead.

Sources

Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen, "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence," Stanford Digital Economy Lab, August 2025.

Matt Beane, The Skill Code: How to Save Human Ability in an Age of Intelligent Machines, HarperCollins, 2024.

Matt Beane, "Shadow Learning: Building Robotic Surgical Skill When Approved Means Fail," Administrative Science Quarterly.

Related guides

Get Started

Ready to grow your local visibility?

Get a free assessment of your current digital presence and a custom plan to improve it.