You Don't Learn to Lead by Building Models. You Learn by Managing the Person Who Does.

The debate about AI and the junior layer has settled on one worry: that tomorrow's analysts won't get the reps but we see a bigger problem. The analyst pool was never only a training ground for analysts — it was the training ground for managers, and it is the only one most banks have ever had.

‍

The data on the junior layer is no longer in dispute. Stanford's Digital Economy Lab, working from payroll records covering millions of US workers, found employment for 22–25 year olds in the most AI-exposed occupations — accounting and financial analysis among them — fell 13% between late 2022 and mid-2025, while employment for 35–49 year olds in the same fields grew by more than 9%. Revelio Labs puts the decline in US entry-level postings at roughly 35% since January 2023. SignalFire found new-graduate starts down around half.

Behaviour across investment banks is seen to match. JPMorgan has proposed cutting its junior-banker-to-manager ratio from 6:1 to 4:1, with half the remaining junior seats moved to India and Argentina. PwC's UK practice cut 200 graduate roles, around 15%, with its head naming AI as a factor. KPMG trimmed graduate intake by 29%, EY by 11%. Goldman cut roughly 1,000 roles across 2025.

The most careful study of the mechanism comes from two Harvard economists, Hosseini Maasoum and Lichtinger, who tracked résumé data covering 65 million workers at more than 280,000 firms. Their finding: after a firm adopts generative AI, junior employment falls relative to non-adopters, and senior employment trends remain largely unchanged.

That last clause is where almost everyone stops reading, and where they draw the wrong conclusion. Senior headcount holding flat is not evidence that senior careers are unaffected. It is evidence that the damage hasn't reached them yet — and that when it does, it will arrive as a capability problem rather than a headcount one.

The apprenticeship runs in both directions

Here is the thing the pipeline argument gets backwards. Nobody in this industry became a leader by building a model. They became a leader by checking someone else's and by deciding whether a junior's output was good enough to put in front of a client, "managing the production team" by working out how to tell a bright twenty-three-year-old that their comps were wrong without breaking them, by carrying three people through a live process at two in the morning and discovering what they were actually made of.

The analyst's grunt work is a curriculum. So is the associate's supervision of it. And the second curriculum was always the more valuable of the two, because technical skill can be taught in a classroom and judgement about people (and how you manage them) cannot.

Management, in finance, has never been formally taught. It has been absorbed. Gallup's own position on this is blunt: most managers learn management from the managers they happened to have, or from whatever they imagine leadership looks like, rather than from any actual development. In investment banking and asset management that transmission ran almost entirely through the analyst pool. Thin the pool and you don't just lose the analyst's reps. You break the chain by which every layer above them learned to lead.

An associate who has never had a direct report becomes a VP who has never had a team. A VP who has never had a team becomes a director who has never built one. The gap doesn't appear in a headcount report. It appears in a client meeting six years from now.

The management layer is already under strain, and the numbers are stark

If this were purely speculative we'd say so. It isn't.

Gallup's State of the Global Workplace 2026 found manager engagement has fallen nine points in three years: 31% in 2022, 30% in 2023, 27% in 2024, 22% in 2025. Over the same period non-manager engagement moved from 20% to 19% — effectively flat. Gallup is explicit that lower engagement among managers accounts for most of the recent downturn in global engagement overall, which fell to 20% in 2025, its lowest since 2020 and, on their estimate, around $10 trillion of lost productivity.

Read that again. The managers are the ones coming apart, not the people they manage.

And Gallup themselves point at the cause. In their 2026 data, South Asia posted an eight-point drop in manager engagement, the largest of any region — and simultaneously, the share of the workforce in management roles fell. Their reading is that employers are cutting management roles and that organisational flattening is a factor. That is precisely the dynamic now being engineered into financial services teams on purpose.

Gallup's analysis puts 70% of the variance in team engagement down to the quality of the manager. Get management wrong and you don't get a slightly less pleasant team; you get a materially less productive one. Meanwhile only one in five US employees strongly agrees they've had a conversation with their manager in the past six months about how to reach their goals. Whatever development is meant to be happening in the absence of the old apprenticeship, it is largely not happening.

Where the other side has a point

Two counter objections.

First, the Harvard finding is that junior decline comes primarily through slower hiring rather than separations. That means this is gradual, and gradual means correctable. Nobody is facing a cliff in 2027.

Second, the "reps" argument can slide into nostalgia. Plenty of people spent three years formatting pitchbooks and learned nothing about leadership except how to survive someone else's bad version of it. A smaller team is not automatically a worse teacher — four analysts you actually know beats six you delegate to by email. There is a version of this where flatter teams produce closer contact and better managers.

The Gallup numbers suggest the opposite: as spans thin and layers compress, the quality of managerial contact is deteriorating, not improving. Flattening is being pursued as a cost exercise, and closer supervision is being assumed as a happy by-product rather than designed as an outcome.

What this means if you're hiring

Three things follow, and they're all actionable now.

Stop treating management experience as a reward for seniority. In most banks you are promoted to VP and then given people. That sequencing only worked when the analyst pool was deep enough that everyone got supervisory reps informally on the way up. It doesn't work in a 4:1 world. Management exposure has to become a criterion for promotion, not a consequence of it.

Give the second review to the junior, not the machine. If AI now produces the first draft and a senior signs off the final version, the reviewing step — the single most formative task in the entire apprenticeship — has been quietly deleted. The firms getting this right are inserting juniors as the checkers of AI output rather than the producers of first drafts and allowing them to build critical thinking skills.

Interview for it, because the CV no longer tells you. We are already seeing candidates at associate and VP level whose CVs imply team leadership and whose actual experience is coordination. Ask how many direct reports, for how long, what happened when one underperformed, and what they did about it. The answers separate the field faster than any technical screen.

And for candidates: in a market where AI can do the analysis, demonstrable experience of developing other people is becoming one of the few genuinely scarce assets you can build. Volunteer for it early. It will price well.

The bottom line

AI isn't going to leave financial services short of analysts. It is going to leave it short of people who have ever managed one.

The firms cutting junior ratios are booking a saving today against a capability they haven't noticed they were buying. Senior headcount will look fine the whole way through — right up to the point where the bench is full of people with impeccable technical records who have never been responsible for anyone but themselves.

Croft & Co is a Franco-British executive search firm specialising in senior appointments across investment banking and asset management in the UK and France. More from our team at croftandco.com/industry-insights

‍

Sources:

• Gallup: State of the Global Workplace 2026
• Gallup: 8 Behaviors of the World's Best Managers (70% of variance in team engagement)
• Hosseini Maasoum & Lichtinger, Harvard: Generative AI as Seniority-Biased Technological Change (SSRN, Aug 2025, rev. Jun 2026)
• Stanford Digital Economy Lab: Canaries in the Coal Mine
• CNBC: AI and entry-level hiring
• eFinancialCareers: JPMorgan junior bankers and AI
• Fortune: PwC UK cutting entry-level roles

‍