Think back to the the work you did in the first real job you had in your field.
The unglamorous, repetitive, slightly humbling work nobody brags about. Building the deck someone else will present. Reconciling the report. Sitting in the meeting you barely understood, taking notes, watching how the people in charge handled the moment when everything seemed to go sideways. You did not know it at the time, but that busywork was your apprenticeship. That was where you learned judgment, context, and craft, one ordinary task at a time.
Here is the question I keep coming back to. If artificial intelligence is now doing all of that work, the first-rung work, the practice-field work, then where will the next generation learn what you learned?
We are about to find out and I don't think we are ready for the answer.
The First Rung Is Disappearing
This is not a forecast. It is already happening, and the numbers are not subtle.
In November 2025, Erik Brynjolfsson and his colleagues at the Stanford Digital Economy Lab published a study appropriately titled Canaries in the Coal Mine. Using payroll records from ADP, the largest payroll provider in the country, they found a 13% relative decline in employment for workers ages 22 to 25 in the occupations most exposed to AI. Software developers. Customer service. The classic on-ramps that were the practice fields for the future leaders. Meanwhile, employment for older, more practiced and experienced workers in those same fields held steady or grew.
Read that again. The work is not vanishing for everyone. It is vanishing for the people on the practice field, at the very beginning of their careers.
You can argue about the exact figures. You cannot argue about the direction. Entry-level postings are down roughly a third since the start of 2023. More than four in ten companies told surveyors they intend to replace roles with AI, and the entry level positions are the first in line. The bottom rung of the ladder is being quietly sawed off, and most organizations are celebrating the cost savings without asking the obvious follow-up question.
If nobody is climbing onto the first rung, who is standing on the fifth rung in ten years?
The Middle Is Thinning Too
While the bottom rung is disappearing, the middle of the org chart is being pressed flat.
Between May 2022 and May 2025, manager headcount at public companies fell by 6.1%. Korn Ferry's 2025 workforce research found that 41% of employees say their company has reduced management layers. Google trimmed roughly a third of its small-team managers. Meta, Intel, Amazon, and others have publicly thinned their hierarchies. Gartner has predicted that through 2026, one in five organizations will use AI to eliminate more than half of their middle management positions.
The industry even has a name for it now. The Great Flattening.
Here is what makes this more than a cost story. Middle management was never just a coordination layer. It was the development layer. It was where a strong individual contributor got handed two people, then five, then a small team, and learned the hardest lesson in all of leadership: how to get work done through other talented human beings instead of doing it all yourself.
"Flatten the middle and you don't just remove managers. You remove the place where managers are made into leaders."
And the managers who survive the cuts? Their span of control keeps widening. The average team size per manager has crept up, and the junior work AI was supposed to absorb is quietly landing on their desks instead. So the people best positioned to coach the next generation have less time to coach than they have ever had. We are asking fewer leaders to develop more people in less time, and we are calling it efficiency.
Why This Should Worry You
For decades, the most reliable research we have on how leaders actually develop has pointed to the same place, and it is not the classroom.
The Center for Creative Leadership, building on the work of researchers like Morgan McCall, gave us the 70-20-10 model. Roughly 70% of leadership growth comes from challenging on-the-job experiences. About 20% comes from developmental relationships, the mentors and managers who let you see their thinking. Only 10% comes from formal training and coursework.
Pay close attention to those numbers
Ninety percent of how a leader gets built happens in the daily flow of real work and real relationships. The stretch assignment that was slightly over your head. The manager who walked you through the reasoning behind the call they made. The mistake you made on something small enough to not sink the company but that stung enough for you to never forget the lesson.
That is the machinery AI is now disrupting. When the tool produces the first draft, the junior employee never struggles through the messy middle where judgment gets forged. When the tool handles the coordination, the would-be manager never learns how to coordinate. Deloitte named this plainly in its 2025 Human Capital Trends research. It warns of a growing "experience gap," one that keeps widening unless organizations deliberately rebuild the developmental pathways AI is erasing.
The cruelest part is the timing. The savings are immediate and visible. They show up on this quarter's numbers and somebody gets a bonus. The cost is delayed and invisible. It shows up three to five years from now as a thin leadership bench, a stalled succession plan, and a room full of senior leaders asking why no one is ready to step up.
By then, the person who approved the cuts may well be gone.
The Honest Counterargument
I owe you the other side, because it makes some valid points.
The optimists argue that AI does not erase the apprenticeship. It raises it. Freed from the grunt work, a new hire can do judgment-based work far earlier than any of us did. Instead of spending two years creating slide decks and formatting reports, that person spends the time reviewing AI output, catching its mistakes, deciding which hard cases need a human, and absorbing business context at a speed the rest of us never had. The World Economic Forum has documented exactly this shift in entry-level work, away from execution and toward evaluation and judgment.
I want this to be true. In some of the best organizations, it already is.
But notice the quiet assumption tucked inside the optimism. It assumes someone is deliberately teaching that judgment to the apprentice. It assumes a manager is sitting beside the 23-year-old, explaining how the decision gets made, why this AI output is wrong, what context the machine is missing. That coaching does not happen by accident. It used to come built into the simple act of reviewing a junior person's work. Now it has to be designed on purpose, by the same managers we just spent a whole section watching get overstretched and cut.
A team at IMD framed it well: the hard part is no longer doing the work, it is building judgment while AI does the work. The opportunity is real, it's just not automatic. And anything that is not automatic, in a busy organization under cost pressure, is the first thing to get skipped.
What Leaders Worth Following Do Now
I refuse to name a problem that has no solution. A design flaw is not the same as a death sentence, and this one can be redesigned. Here is where to start.
Put development on the books. If 70% of growth comes from challenging experience, stretch assignments are too important to leave to chance in a flattened org. Name them. Assign them. Track them the way you track revenue. Ask of every junior role: where, specifically, is this person being stretched past their current ability? If you can't answer, you have found your first gap.
Pick one early-career person on your team. Write down the single hardest, most growth-producing thing they will do in the next 90 days. If the honest answer is "nothing AI is not already doing," you have just diagnosed your future leader pipeline.
Coach the thinking, not the task. When AI writes the first draft, the old way of coaching by marking up someone's work disappears. Replace it with something better. Narrate your decisions out loud. Tell your people not just what you decided but how you got there, and why. Let them watch you wrestle with an AI output and overrule it. Judgment is caught more than it is taught, and people can't catch what they can't see.
Protect the managers who mentor the future leaders. You cannot flatten the org, widen a manager's span of control, hand them the displaced junior work, and still expect them to mentor anyone. If developing people matters, give your remaining leaders the space, the time, and the mandate to actually do it. Then reward them for the leadership pipeline they build, not only the targets they hit.
Count the bench, not just the budget. The savings from cutting entry-level roles are easy to measure. The damage to your leadership pipeline is easy to ignore, right up until it becomes a crisis. Put both in front of the decision-makers, on the same page. Make the delayed cost visible before anyone signs off on the immediate savings.
The Choice In Front of Us
Here is what I believe, and I believe it strongly.
AI is not the thing that breaks the leadership pipeline. Neglect is. AI just removed the accidentally built in version of development we have always relied on, the apprenticeship disguised as the ordinary entry level work. Now that the safety net is gone, whatever replaces it will have to be built on purpose by leaders who realize the flaw in the new design.
Whatever we allow today becomes the culture that shapes tomorrows leaders.
So now that you recognize the canaries in the coalmine, do not let your organization ignore the impending danger and optimize its way into a leadership crisis before it is too late to fix. Look at the people on the bottom rung of your ladder, the ones AI could most easily replace, and ask a better question than the one your spreadsheet is asking. Not can we automate this? but how are we going to develop the one that used to do that work?
The next generation of leaders is sitting in your organization right now, doing work a machine could do. They are waiting to find out whether anyone is going to automate their growth into the succession plan.
That somebody is you.
It's time to get off the bench...and lead.
Sources
- Brynjolfsson, Chandar & Chen, Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence, Stanford Digital Economy Lab / SIEPR, November 2025. siepr.stanford.edu
- Yale Insights, The Real Job Destruction from AI Is Hitting Before Careers Can Start. insights.som.yale.edu
- CNBC, AI is not just ending entry-level jobs. It's the end of the career ladder as we know it, Sept 2025. cnbc.com
- HR Dive, Over one-third of companies plan to replace entry roles with AI, survey says. hrdive.com
- Axios, Middle managers in decline as "flattening" spreads, AI advances, July 2025. axios.com
- Korn Ferry / Lepaya, The Great Flattening: Middle Management Cuts in the AI Era. lepaya.com
- Center for Creative Leadership, The 70-20-10 Rule for Leadership Development. ccl.org
- Deloitte, 2025 Global Human Capital Trends (the experience gap), via RewardsDNA analysis. rewardsdna.com
- eMarketer, AI helped cut costs in 2025 but may hurt future leadership pipelines and increase workloads. emarketer.com
- World Economic Forum, How AI is changing the nature of entry level work, 2026. weforum.org
- IMD, How to build judgment when AI does the work. imd.org
- Gartner middle-management prediction, via People Managing People, The Great Flattening. peoplemanagingpeople.com