Research Director: Philip Tomei

The Problem

Since the release of ChatGPT in November 2022, entry-level hiring in AI-exposed occupations has fallen drastically (Brynjolfsson et al., 2025; Hosseini & Lichtinger, 2025; Klein Teeselink, 2025). Junior roles have historically functioned as paid training grounds: document review taught legal reasoning, data entry helps ingrain business processes, photocopying taught office navigation. LLMs can now perform some of these tasks better and cheaper than junior employees, which diminishes the incentives for firms to hire early career individuals. This contraction risks hollowing out the economy’s expertise pipeline, as fewer junior roles today mean fewer mid-career experts in a few years, which means lower human capital, lower productivity, and fewer people to build, audit, and govern AI systems when they become increasingly advanced.

The Market Failure

One may ask the question: If firms require experts in the future, why would they not simply hire young people and train them to become experts, even if some of the useful tasks they used to perform are now done by AI? The reason is that early-career training in portable skills generates a poaching externality: firms cannot capture the full return on training investments because workers may subsequently be hired by competitors, leading to systematic underinvestment in worker development. Because LLMs have reduced firms’ private return to hiring juniors, less hiring will take place, even if the social return of developing professionals who can audit AI outputs, identify hallucinations, and build domain-specific applications remains high or has even increased.

This creates a social inefficiency: today's hiring and training opportunities determine tomorrow's stock of experts. Hence, the economy risks losing productive capacity, which lowers the capacity to govern increasingly sophisticated AI systems in the future. The welfare loss compounds over time as each cohort of excluded workers represents permanently foregone human capital formation.

Solution

To address this market failure, we propose establishing a Human Capital Investment Sharing (HCIS) program that realigns incentives for early-career hiring. Under this system, firms that hire workers within their first five years of labour force participation would receive a modest share (initially set at 1%) of those workers' gross earnings throughout their careers, regardless of where they subsequently work. The 1% contribution would be deducted from workers' pay checks throughout their careers, similar to student loan repayments, with payments flowing directly to their early-career employers.

This mechanism transforms early-career hiring from a risky investment into an internalization of positive spillovers. The 1% rate is deliberately modest. For example, for a worker earning £40,000 annually, this represents £400 per year total across all their early-career employers. Yet, £400 per year over 25 years, at a discount rate of 3.5% would imply a value of £6.6k to the firm, which constitutes a meaningful incentive for hiring junior employees, and investing in their personal development.

The fact that firms are entitled to the payment regardless of where the employee will work afterwards solves the job-hopping problem that traditionally discourages firms from investing in early-career workers. Rather than losing their entire investment when a worker leaves, firms retain permanent financial interest proportional to their contribution to that worker's development. This creates natural incentives for good management, as firms benefit when their early-career employees succeed and earn higher wages over time.