AI now drafts the first version of the contract, the code, the commercial offer, the market analysis, … These were precisely the tasks that juniors used to do.
But AI output still needs to be checked, corrected and validated, and that requires judgment that only experienced people have.
So seniors become reviewers. And the question follows naturally: how did those seniors acquire their judgment in the first place? Largely by doing, for years, the very tasks that AI now performs.
In the interviews Solvay Lifelong Learning conducted this year with executives across industries, this theme surfaced most spontaneously in the first conversations. Different sectors, different stages of AI maturity, the same concern: learning by repetition, and the critical judgment it builds, is fading, and nothing stable has yet replaced it.
The labour market evidence is accumulating. Using payroll data from the largest US payroll provider, Brynjolfsson, Chandar and Chen at Stanford's Digital Economy Lab track what they call the “canaries in the coal mine”. Their August 2026 update finds that employment of young workers (ages 22–25) in AI-exposed jobs now stands 19% below where it would otherwise be. Notably, employment declines are concentrated in occupations where AI is more likely to automate, rather than augment, human labour.
A separate study by Lichtinger and Hosseini at Harvard, based on nearly 62 million résumés, reaches a similar conclusion: from early 2023, junior employment in firms adopting generative AI declined sharply relative to non-adopters, while senior employment continued to rise. And the mechanism matters: the junior decline is driven primarily by slower hiring rather than increased separations. Juniors are not being fired; they are simply not being hired.
The picture is not settled, though. Some economists caution against attributing everything to AI: an analysis for the Economic Innovation Group points out that the sharp drop in job postings began in early 2022, which makes the entry-level cohort shrink mechanically, independently of AI. Causality is still debated. But for managers, the practical question doesn't wait for the academic one to be resolved.
The problem has a second side: the role of seniors. AI dramatically accelerates execution cycles, but decisions, validations and priorities still go through people. How can a senior professional keep up with a team (of juniors or agents) that now produces in hours what used to take days?
Several executives Solvay Lifelong Learning interviewed described exactly this: AI accelerates production, but if experienced people cannot review, decide and redistribute work fast enough, they slow the whole system down. And when seniors are absorbed by validation work, the time they once spent passing on their know-how is the first casualty.
This is where the two sides of the dilemma meet. Matt Beane's research, synthesised in The Skill Code (2024), shows that expertise grows out of the expert-novice bond, built on challenge, complexity and human connection. Fewer juniors doing formative work, and seniors with less time to guide them: the bond that turns one into the other weakens from both ends.
Together, this creates a set of dilemmas with no textbook answer.
Should we keep hiring juniors? Reducing entry-level hiring may be rational for this year's budget. But who will be the senior reviewers of 2035? And if every firm in a sector makes the same short-term choice, where will experienced talent come from?
What does a junior actually do on day one? If AI produces the first draft, is reviewing that draft a learning experience, or does it require exactly the judgment a junior doesn't yet have? Should some formative tasks be deliberately kept “manual”, even at a productivity cost?
How do we assess juniors now? When every CV, report and offer is “well written” thanks to AI, traditional signals of quality lose their discriminating power. What should we look at instead: reasoning, oral skills, the quality of questions asked?
How much authority should be delegated? What level of decision-making should be entrusted to more junior people or to AI agents to keep pace, and with what safeguards?
Who passes on know-how, and with what time? If seniors are absorbed by validation, when do they transmit their tacit knowledge? Should mentoring become an explicit part of senior roles, and how should their performance be assessed?
How do we keep a shared sense of priorities? When everyone can launch more initiatives than ever thanks to AI, how do seniors and managers keep the team focused?
Does team structure itself need to change? Smaller teams, different roles, new forms of coordination?
The answers will look very different in a law firm, an engineering company, a bank or a fast-growing tech scale-up. Some of the organisations we spoke to are experimenting with rotation, others with
reinforced internal mobility, others with redesigned onboarding. None claims to have found the model.
This is a question no single manager can solve on their own, and one where peer exchange is particularly valuable. Hearing how a CFO, a CIO and an HR director from different industries are approaching the same dilemma, with frameworks to structure the discussion and research to test assumptions, can help managers move from a vague concern to concrete experiments they can try in their own organisation.
This is one of the questions at the heart of Executive Programme in People Leadership in the Age of AI, a new programme starting in Spring 2027 at Solvay Lifelong Learning, designed for managers in organisations already transformed by AI. To help participants find answers that are pragmatic, contextual and actually implementable in their own organisation, the programme combines robust frameworks to read the situation, research to confirm or challenge intuitions, and structured exchange with peers facing similar dilemmas in different contexts, across four in-person modules and small-group online coaching sessions.
How is your organisation rethinking the roles of its juniors, and of the seniors who develop them? Have you found approaches that work, or that clearly don't?
Are you interested in joining our first cohort of the Executive Programme in People Leadership in the Age of AI? Find all the information on our programme page.