Forty-three percent of Fortune 500 CEOs say they’re shifting hiring away from junior roles over the next two years. Last year that number was seventeen percent. The Oliver Wyman Forum and the New York Stock Exchange polled four hundred fifteen of them. The trajectory isn’t subtle: corporate America is intentionally amputating its own talent pipeline.
In the same breath, Heidrick & Struggles’ 2026 Route to the Top report says that more than a third of major U.S. companies do not have a CEO with the capability to lead the organization through the next two to three years. Forty-six percent of directors flag a real gap between what their CEO can do and what the business actually needs. The external hire rate jumped from 18 percent to 27 percent because boards are scrambling for ready-made talent — the internal bench is dry.
Read those two numbers next to each other. The people who can’t lead their companies through the AI transition are the same people firing the kids who would have replaced them.
This is the pipeline. And it’s broken.
How they got the job in the first place
I spent a few hours this week working through the demographic and structural research on Fortune 500 CEOs. The picture is consistent across decades:
76 percent of CEOs are internal promotions.
The average sitting CEO has 24 years of professional experience.
Internal candidates averaged 12 to 14 years inside the company before getting the seat.
41 percent came up through the COO/President role. Another 10 percent came up through the CFO chair (a decade high). Roughly 12 percent came up through running a division.
Only 4.2 percent are founders. The Zuckerbergs and Musks are a rounding error.
Every one of those promotion paths is an apprenticeship path. The COO learned the business by running operations under a CEO. The CFO learned strategy by sitting next to a CEO who treated him like a co-pilot. The Divisional CEO ran a contained P&L because someone gave a 32-year-old a chance to break a small thing before they were trusted with a big one.
The Harvard CEO Genome study tracked 17,000 C-suite executives over ten years. It identified the people who reached the corner office faster than the 24-year average. Sixty percent of these “sprinters” got there by taking smaller jobs — voluntary downward or lateral moves into divisions where they could own everything. A third got there by accepting roles they were arguably underprepared for. A big chunk got there by inheriting a mess and fixing it.
In other words: every modern CEO is the product of someone deciding to put an underprepared junior person into a position that exceeded their reach, and then betting on them.
That’s the system. That’s how it has always worked.
The current generation of CEOs is dismantling it.
What they’re actually cutting
The 43 percent number is the headline. The full picture is uglier:
74 percent of CEOs are freezing or reducing headcount, up from 67 percent.
Entry-level job listings are down 30 percent since 2022.
Middle management postings are down 42 percent.
Corporate America eliminated 1.17 million jobs in 2025 — most of it positioned to investors as “funding AI.”
Stanford’s November 2025 study found young workers are 16 percent more likely to lose their jobs in the most AI-exposed sectors.
Klarna went from 7,000 employees to roughly 3,000, with a public target of under 2,000 by 2030.
Salesforce cut customer support from 9,000 heads to about 5,000. Marc Benioff said it on a podcast: “I need less heads.”
Amazon laid off 14,000 corporate roles in October 2025, then another 16,000 in early 2026.
UPS announced 20,000 cuts, the largest in its 116-year history.
The reasoning the executives give is interchangeable. Andy Jassy wants Amazon to run “like the world’s biggest startup.” Luis von Ahn declared Duolingo “AI-first,” including for performance reviews and hiring decisions. Sebastian Siemiatkowski says the quiet part out loud about Klarna’s automation deals.
There is a contrarian counterpoint. IBM is tripling U.S. entry-level hiring this year. McKinsey is planning a 12 percent increase in North American hiring. They are explicitly betting that whoever is going to manage an agentic workforce in 2030 has to be apprenticed today. They are the exception, not the rule.
The capability gap nobody wants to name
If the AI bets were working, you could maybe argue the cuts were rational.
They aren’t working.
The same Oliver Wyman survey reports that only 27 percent of CEOs say AI ROI is meeting or exceeding expectations. Last year that number was 38. Two-thirds are still at the planning or pilot stage. Nearly a quarter say AI has had no measurable impact on revenue at all. They are firing people now to pay for benefits that have not materialized and that, in most cases, they don’t yet know how to extract.
That is not a strategy. That is panic dressed up in a board deck.
Now layer in the Heidrick & Struggles data. Only 25 percent of business leaders consider CEO succession planning a top priority. A third of companies have zero KPIs measuring the effectiveness of their succession pipelines. One in four corporate leaders openly admit their current process is missing high-impact leaders inside the company. Half of organizations that claim alignment actually don’t have a comprehensive plan.
The pipeline is broken because nobody is being paid to fix it. CEOs are compensated on quarterly performance and stock price. Headcount cuts boost both in the short term. The damage to the leadership bench shows up in five to ten years — long after this batch of executives has collected their exit packages.
Technical weakness: they don’t understand the thing they’re betting the company on
Here is the part nobody on a board wants to say out loud.
The CEOs cutting juniors to “fund AI” are, in most cases, the least technically literate people in the building. They went to business school. They came up through finance or operations. Their understanding of a large language model is roughly: “It writes good emails and the consulting deck said it would 10x productivity.”
I have been writing C# since before some of these CEOs were running their first P&L. I have shipped agentic systems on the Microsoft stack, built MCP servers in production, debugged tool-call mismatches between qwen and gemma in a multi-agent pipeline. I know what a vector store costs at scale, what context-window degradation looks like at turn thirty, and why a junior who can read a stack trace is worth ten McKinsey slides about “AI transformation.”
The CEOs do not know any of this. They are buying narratives, not capabilities. They are setting headcount strategy on the assumption that an agent can replace a junior developer, when in reality the agent will produce code that needs a senior to review, a mid to deploy, and a junior — the one they just fired — to actually understand the codebase well enough to spot the hallucinated function call.
Their technical model of AI is whatever the vendor sales deck told them. That is not a base on which to architect the workforce of a Fortune 500.
Soft-skill weakness: they outsourced their thinking
The behavioral research on CEOs is consistent. Russell Reynolds groups effective executive behavior into three pillars: forward thinking, intrepidness, and team building. Anticipating disruption. Tolerating ambiguity. Drawing strength out of other people.
You cannot build any of those by sitting in a chair where every input is a Claude or ChatGPT summary of someone else’s work.
MIT’s EEG research on heavy LLM use measured a roughly 50 percent drop in gamma-wave activity in regions associated with deep problem solving. Gerlich’s 2025 study found a direct negative correlation between AI usage and critical thinking scores. Multiple studies converge on the same finding: when humans outsource cognition to a model, the model’s reasoning structure becomes their reasoning structure. They stop generating arguments. They stop noticing gaps. They develop “cognitive debt” — confidence without competence.
Now ask yourself who in a corporation uses AI summaries the most. Who never sees raw data anymore. Who reads board decks generated by Copilot, gets briefings synthesized by an enterprise GPT, and signs off on strategy memos that the firm’s consultancy wrote with the help of a model that pulled from the same five reports every other Fortune 500 CEO is reading.
The CEO.
The CEO is the single most cognitively offloaded person in the building. They are the role most exposed to automation bias, most insulated from raw friction, most surrounded by yes-machines — both human and silicon. Their soft skills atrophy because the system removes every opportunity to exercise them. They are not building team capability; they are pruning it. They are not engaging ambiguity; they are buying tools that simulate certainty. They are not forward-thinking; they are pattern-matching to whatever Heidrick or Spencer Stuart told them their peers are doing.
This is what I mean when I say AI poisoned them. It did not make them dumb. It removed the friction that used to force them to be smart. And the same technology gave them a justification to fire every human in the building whose job it was to push back.
The poison spreads
The damage is not limited to the corner office. The World Economic Forum’s recent research found that the work AI was supposed to do at the junior level is instead being pushed upward. Middle management is absorbing it. They are overextended, burned out, disengaged. A Fortune piece earlier this year noted that by late 2026, 20 percent of companies expect to use AI to flatten their hierarchies, eliminating more than half of mid-tier roles.
The middle layer is connective tissue. It is where strategy gets translated, where talent gets coached, where exceptions get handled — the complex human problems algorithms do not touch. Remove that, and what is left is a CEO and a senior team trying to govern a stack of language models with no human signal between them and the actual work.
Decision latency climbs. Institutional knowledge evaporates. The juniors who would have become the next COOs, CFOs, and Divisional CEOs are never hired in the first place — or are hired and given no path because nobody mid-career has time to mentor them. In ten years, when boards go looking for an internal successor with 12 to 14 years of compounding experience inside the company, they will find an empty room.
This is how an industry strangles its own future on a forecast timeline.
The fix
I am not going to pretend this is a clean problem. But the fix is not exotic. It is a return to a handful of things the companies in the Fortune 500 used to know how to do.
1. Treat junior hiring as capital expenditure, not operational expense. IBM and McKinsey are right. The cost of a class of entry-level analysts in 2026 is the cost of an executive bench in 2036. Cut it now and you are not saving money — you are deferring a much larger bill to a future CEO who will inherit your mess. Boards should require junior hiring numbers in succession-planning KPIs, the same way they require capital reserve ratios.
2. Reinstate the apprenticeship path. The CEO Genome research is clear: every effective executive came up through end-to-end P&L responsibility, often on a small or messy unit. Stop killing the small divisions and turnaround assignments that produce real operators. The “inherit a big mess” path does not exist if there are no juniors left to throw at the mess.
3. Force friction back into the CEO’s day. The research on cognitive forcing functions is real and well-documented. AI tools should not be allowed to give a CEO a final answer without surfacing the disagreement, the dissent, the lower-confidence alternatives. If a board is letting its CEO consume only the polished, pre-digested, model-summarized version of reality, the board is paying for an executive who has stopped thinking.
4. Audit the AI ROI honestly. Seventy-three percent of CEOs are not getting the return they promised investors. That is not a marginal data point — that is most of them. Until the savings actually materialize, you cannot fire the humans on the assumption they will. Boards should require a 1:1 audit: every layoff billed as “funding AI” should be matched against a documented, deployed, measurable capability that replaces the work. If the capability is not there, the layoff was a cost cut wearing AI’s hat.
5. Make boards do their actual job on succession. A third of companies have no KPIs for succession effectiveness. That is governance malpractice. If only 25 percent of leaders treat CEO succession as a top priority, the other 75 percent are gambling with the enterprise. Compensation committees should tie a slice of CEO pay to the strength of the bench they leave behind.
6. Stop confusing leanness with leverage. A structurally leaner organization is not automatically a better one. It is a more brittle one. Klarna gets quoted a lot in these pieces. It is also a company that, when its AI deployments under-deliver, has nowhere to retreat — the institutional capacity to recover is gone.
Closing
The CEOs sitting inside Heidrick’s 2026 capability gap are the first generation of executives whose internal cognitive process is fundamentally entangled with a tool they don’t understand. They are governing companies where the talent pipeline they themselves climbed has been broken by their own headcount decisions. The boards above them are not measuring succession. The middle managers below them are drowning. The juniors they are not hiring this year are the executives nobody will be able to promote in 2036.
The thing that makes this story interesting, and not just sad, is that it is a system failure with a system fix. The pipeline did not break itself. CEOs broke it. They can also rebuild it — if their boards force them to, and if they can find the cognitive bandwidth to think about something more than five quarters out.
That is a real “if.”
-Gal