To save entry-level jobs from AI, look to the medical residency model

At the Davos World Economic Forum this week, the CEOs of two leading artificial intelligence (AI) companies issued a joint warning: Entry-level workers are about to feel AI’s impact. Demis Hassabis of Google DeepMind said he expects AI to begin to impact junior-level jobs and internships this year, while Dario Amodei of Anthropic reaffirmed his prediction that 50% of entry-level jobs could disappear within five years.

If they’re right, the traditional model of developing young talent in knowledge sectors—hiring junior workers to perform routine tasks while they gain expertise over time—won’t survive when AI handles those tasks instead. I’ve been warning about this risk for over a year; now, the people building the technology are putting timelines on it. While labor market evidence does not conclusively show that AI is already claiming entry-level jobs, we should prepare solutions now.

I propose we reimagine the career ladder, and to do that, we can look to medical professions. For decades, hospitals have used residencies to quickly develop expertise—structured, mentored programs where learning is the job itself. White collar professions facing AI disruption could adopt their own version of the residency model.

While AI excels at knowledge tasks performed on computers (researching, writing, calculating, coding, etc.), it cannot replace a trial lawyer who reads a jury and delivers a persuasive closing argument, or a manager navigating a sensitive termination. These responsibilities require judgment, intuition, and embodied presence—qualities that aren’t digitized, but developed through years of practice.

Here’s the problem: The entry-level jobs that once provided that practice—through drafting documents, building decks, running analyses, etc.—are built around the exact tasks AI is learning to do. If employers stop offering those roles, they sever the pathway to senior expertise and their future managers and leaders. After all, entry-level work serves two purposes: getting routine tasks done and preparing workers for advanced roles.

Medical residencies take the opposite approach. Residents are doctors in training, but they’re also practicing physicians. From the start, residents examine patients, propose treatment plans, and make real health care decisions under supervision. Senior physicians coach them through cases, explain their reasoning, and gradually grant autonomy. Crucially, learning is not a side perk of the residency—it is the job.

White collar professions could adopt the same model. This would mean that in law, junior associates wouldn’t spend years reviewing documents and drafting contracts—work that software increasingly handles. Instead, they’d shadow negotiations, practice courtroom arguments, and progressively lead cases, with mentors debriefing their choices. The goal would be skill development, not billable hours. In consulting, residents could join client presentations from day one, watching how senior consultants read the room and handle objections. With coaching, they would gradually take the lead.

This kind of training is expensive. In medicine, U.S. taxpayers subsidize hospitals’ training costs through Medicare, because society benefits from competent doctors. While taxpayers shouldn’t bankroll white collar training, companies that benefit most from AI-driven productivity gains should help sustain the talent pipeline those gains depend on.

One way to address this is an AI workforce reinvestment fund: Firms that automate away entry-level roles would contribute to pooled resources underwriting residency programs across industries. This isn’t a penalty for innovation, but a mechanism to reinvest some efficiency gains in the next generation of talent. The United Kingdom already uses a similar “use it or lose it” model for apprenticeships, requiring larger employers to pay a modest payroll levy that can only be reclaimed through approved training programs. Unlike universal basic income or reskilling programs that assumes those jobs are forever lost, a reinvestment fund keeps the pipeline open for the professions young people have worked to enter.

Philanthropy can also play a role. Foundations concerned about workforce disruption could fund the infrastructure that residencies require, including standardized curricula and intermediaries to help firms offer training they couldn’t build alone. Government could also adapt youth service programs to the residency model, placing young, aspiring professionals in legal aid organizations, public agencies, or nonprofit tech teams to gain real professional skills and mentorship while serving the public good.

If employers, philanthropy, or government don’t step in, the cost will fall on young people themselves. Graduates may face a difficult choice: pay for additional degrees, shell out for pricey bootcamps and training of questionable quality, or compete for shadow internships landed through family connections. AI’s productivity gains would flow to shareholders and senior employees while young people foot the bill for their own career development.

AI may soon build every slide deck and stress-test every model. What it still can’t build is judgment: the ability to make high-stakes decisions amid ambiguity, to earn trust, and to lead. And judgment isn’t downloaded—it’s developed. If we want future leaders, we have to invest in the messy, mentored work that creates them. Policymakers, employers, and philanthropists should start building the infrastructure to make a new career ladder possible—one that prioritizes the learning and the skills that AI can’t replicate.