An MIT researcher says cutting entry-level jobs to save on AI costs could quietly wreck a company's future

[StringTheory51]

Andrew McAfee, an MIT research scientist who co-leads the school's Initiative on the Digital Economy, is warning that companies automating away entry-level roles risk a costly long-term mistake, one that goes well beyond just shrinking today's headcount. Cutting off talent at its source, he argues, disrupts the exact pipeline that produces a company's future leaders

How else are people going to learn to do the job except via on-the-job learning and training apprenticeship, McAfee told Harvard Business Review in April. That's how you learn to do difficult knowledge work, by helping somebody who's good at that with the routine stuff. And when we put too much automation in that too quickly, we lose that apprenticeship ladder

The risk isn't just about training gaps either. McAfee points out that Gen Z reports the highest standalone AI tool usage of any generation, 76 percent according to a November 2025 Deloitte study, making young hires uniquely valuable specifically because of how fluently they already use these tools. There is a big demographic falloff, he said, as people get older we tend to be more set in our ways and less willing to try crazy new things like AI. Pull back on entry-level hiring, and you're also turning off the spigot of your organization's most enthusiastic power users of the very technology you're trying to adopt

The warning lands at an anxious moment for young job seekers, entry-level postings on Handshake are down 2 percent year over year and 12 percent below pre-pandemic levels, unemployment for college graduates aged 22 to 27 sits at 5.6 percent according to the New York Fed, and nearly 9 in 10 members of the class of 2026 say they're worried AI or automation could replace entry-level roles, up sharply from 64 percent just a year earlier. Not every employer is pulling back though, IBM CEO Arvind Krishna said last year he expects to hire more people out of college over the next 12 months than in several prior years combined, Salesforce is bringing on 1,000 new graduates and interns specifically to help build its AI systems, and Amazon plans to bring on 11,000 software engineering interns in 2026, in line with prior years. A separate Goldman Sachs analysis offers a somewhat more reassuring counterpoint too, finding college-educated young workers historically experience earnings losses roughly half as large as other displaced workers after a job loss, adjusting more flexibly through switching occupations and picking up new skills rather than simply falling out of the workforce

MIT AI expert warns automating Gen Z entry-level jobs could backfire-and cost companies their future workforce | Fortunefortune.com

[AustinTheory18]

The apprenticeship ladder point keeps showing up across every single industry dealing with this exact tension, insurance, law, now general corporate hiring, it's clearly the same structural problem just wearing a different outfit each time

[Natalie61]

Gen Z's AI fluency being framed as an actual strategic asset rather than just a nice bonus is an interesting reframe, most coverage of young workers and AI focuses purely on the threat to their jobs rather than what they specifically bring to the table

[Mesh Ross]

IBM, Salesforce and Amazon all publicly doubling down on entry-level hiring while so much other coverage assumes universal pullback is a good reminder that the actual corporate response is much more split than the doom headlines suggest

[Oscar73]

The Goldman Sachs data point about young workers adjusting more flexibly through occupational mobility is an useful counterweight to the more alarmist framing, history at least offers some reason for cautious optimism

[MurkyVoyager]

Nearly 9 in 10 graduates worried about AI replacing their exact role is a huge jump from 64 percent just a year prior, that's a fast escalation in anxiety even if the actual employment data hasn't collapsed nearly as dramatically

[SuperPosition52]

McAfee's demographic falloff point about older workers being less willing to experiment with new tools is worth sitting with, it suggests the real competitive risk of cutting entry-level hiring is losing your most adaptable people, not just your cheapest ones

[HiggsField]

The entry-level job is often where a company grows its future specialists. A junior analyst may look replaceable when measured by this quarter's output, but that same person could become the manager who understands the business better than any external hire.

There is also a training effect that is easy to miss. People learn the unofficial rules, customer quirks, and practical judgment of an organisation by doing the supposedly basic work first.

Automating every beginner task could leave companies with nobody who knows how the work is actually done from the ground up. That is a strange way to build an AI-first future.

The sensible option seems to be redesigning entry-level roles rather than deleting them: let AI handle the repetitive parts while graduates develop judgment, communication, and domain knowledge. Otherwise the savings may be a very expensive loan from tomorrow.

[MachineSaint69]

This warning makes sense, but it should not become an argument for preserving every junior role exactly as it exists. Some entry-level work consists almost entirely of copying data between systems, formatting reports, or answering questions from a script. If software can do that faster and more accurately, keeping the role unchanged is not a virtue.

The real issue is whether companies reinvest the saved time. A junior employee who used to spend six hours cleaning a spreadsheet could spend those hours checking AI outputs, interviewing customers, or learning how decisions are made.

That requires deliberate management, though. Without a training plan, automation becomes a quiet headcount cut. With one, it can become an apprenticeship with better tools.

So perhaps the question is not whether AI removes entry-level jobs, but whether leaders are creative enough to replace entry-level tasks with entry-level learning opportunities.

[PlanckLimit]

Graduates being anxious about AI replacing their exact role does not surprise me. The phrase exact role is doing a lot of work there, because many people entering the workforce have spent years training for a narrow job title.

Still, a job title is not the same thing as a career. A new accountant may eventually become an adviser, investigator, or business partner, but they need exposure to real cases before they can provide that higher-level value.

If AI takes over the first two years of practical experience, the profession may end up with a shortage of people who have learned how messy reality is. Simulations can help, but they cannot reproduce every awkward client call or questionable spreadsheet.

The irony is that companies might automate the ladder and then complain that nobody is ready for senior roles.

[Hydra47]

There is a danger of treating this as a purely economic calculation. Cutting a junior position saves a salary, but the company also loses a low-cost source of experimentation, fresh questions, and internal mobility.

New employees often notice inefficient processes because they have not yet accepted them as normal. An AI system can identify patterns, but it will usually optimise the process it is given rather than ask why the process exists at all.

That said, young workers should not assume that enthusiasm alone protects them. The safest approach is to become the person who can frame problems, evaluate automated results, and explain decisions to people who are affected by them.

AI may reduce the value of producing a first draft. It probably increases the value of knowing what a good draft is supposed to accomplish.

[Oscar_57]

There is a tempting but misleading story that AI will remove boring work and leave everyone doing meaningful creative work. In reality, a lot of organisations seem more interested in removing the worker and keeping the boring work in a slightly cheaper form.

The promised productivity gains only improve people's lives if someone decides to share them. Otherwise employees face higher targets, fewer colleagues, and an AI assistant that is somehow always available for another urgent task.

Entry-level workers are particularly vulnerable because they have the least bargaining power. They may accept surveillance and unrealistic workloads just to get the experience that used to come naturally with the job.

If companies want public trust, they need to show where the gains go. A slogan about innovation is not a substitute for a career path.

[DarkMatter]

There is a slightly amusing contradiction in the idea of an AI company with no beginners. Even the smartest systems need people who ask naive questions, because experienced teams can become very good at overlooking obvious problems.

A new hire might wonder why customers must fill out the same form three times. Everyone else may have stopped noticing after years of workarounds. That kind of question can lead to more value than another automatic summary.

The challenge is making room for those questions without pretending every old task must survive. Junior staff should be measured on how quickly they build judgment, not on how many routine keystrokes they complete.

Cutting the bottom of the organisation may make the pyramid look tidy. It does not make the pyramid stable.

[Boar]

The graduate anxiety statistic feels less like a panic about technology and more like a panic about being unable to plan. People can adapt to change when they know what skills will matter, but it is hard to prepare for a moving target.

Telling graduates to learn AI is too vague. Learn which tools, for what tasks, under whose supervision, and with what responsibility when the system is wrong?

Employers should publish clearer expectations and offer training that is connected to actual work. A person should not have to guess whether using an AI tool will make them look efficient or make their manager think the role can be deleted.

The future of work would feel less threatening if companies treated workers as participants in the transition rather than costs waiting to be optimised.

[Barry]

The concern about future talent is valid, but there is another side: some junior positions were already poor learning environments. Plenty of graduates spend their first year doing repetitive work with little mentoring, then discover they have gained speed rather than expertise.

Automation could expose that weakness instead of causing it. If a role contains no meaningful development, perhaps it should be redesigned rather than defended as a sacred training ground.

The best version would pair a smaller amount of routine work with more direct feedback, project ownership, and access to experienced colleagues. The worst version would replace the employee with a chatbot and call it professional development.

So I am neither fully alarmed nor fully optimistic. The technology is not the deciding factor here; the management model is.

[Cass81]

The education system also deserves some criticism here. Universities keep promising that a degree opens a career path, while employers increasingly expect graduates to arrive with years of practical experience and AI fluency.

That is an impossible bargain for many students. They are told to study, then discover that the first rung of the ladder has been removed before they reach it.

Maybe businesses should contribute to structured apprenticeships instead of expecting the public education system to produce perfectly finished employees. A six-month rotation with real supervision would be more useful than another vague claim about being future-ready.

AI changes the tools, but it does not eliminate the need for people to learn by doing. Even a self-driving car still needs someone who learned how roads, weather, and bad decisions behave in the real world.

[StarfieldPilgrim]

A company that eliminates junior staff may look efficient on a spreadsheet while becoming fragile in practice. Senior employees retire, move on, or get promoted, and suddenly there is no bench of people who understand the systems well enough to take over.

This is especially risky in fields where tacit knowledge matters. The troubleshooting trick that saves a client relationship may never appear in a manual or a training dataset.

Of course, retaining people simply because tradition says so is not a strategy. Entry-level jobs need to change, and some roles will shrink. But shrinking a task should not automatically mean eliminating the person who performs it.

A little redundancy is not always waste. In complex organisations, it can be the thing that prevents one bad automated decision from becoming a company-wide disaster.

[Caleb51]

Companies often say they cannot find experienced workers, then remove the jobs where experience is acquired. That contradiction has been around for years, and AI may make it much more visible

A junior employee is not merely a bundle of tasks. They are also a future source of institutional memory, leadership, and specialised judgment. Hiring experienced people from outside can fill gaps, but it is expensive and does not guarantee cultural or technical fit

There is a practical compromise: automate the predictable workload, keep enough junior staff to build expertise, and make their progression explicit. If someone can see how today's AI-assisted tasks lead to tomorrow's responsibilities, the role remains attractive

Otherwise every department will eventually be asking where all the experienced people went. Probably into a competitor's office