AI Is Letting Companies Cut Entry-Level Jobs. Here’s Why That Is a ‘Critical Strategic Mistake,’ According to an MIT Economist.

Frank Nagle spends his time watching how people actually work. 

As an MIT economist and the advising chief economist for the Linux Foundation, Nagle studies how AI is rewiring the modern workplace. He got his start in cybersecurity, then shifted his research to the intersection of technology and business decisions. 

“My research recently has been thinking about how AI not only improves productivity, but also how it changes the way that people work and the way that they spend their time,” Nagle tells Entrepreneur in a new interview.

He does not deny that AI is going to disrupt the workforce: “100% there’s going to be changes and job loss,” he says. He predicts that workers will “still have jobs that are different from those they have today” 15 to 20 years from now. What worries him is how people will manage that transition, so that workers can “feed their families and find meaning from their work” while society figures out how AI will reshape everything that we do. 

Why firing junior staff is a strategic error

Nagle says that companies that cut junior staff in the name of AI are making a “critical strategic mistake.” He frames the problem in two dimensions. First, if companies don’t hire junior people, they are missing out on choosing who will run the firm in a decade. Second, junior people benefit more from using AI and reorienting their workflows much more than senior people. “There’s a lot to be learned from the way that junior people are interacting with AI,” Nagle says.

He notes that from a purely competitive perspective, firms that keep hiring junior employees may have their pick of “more and possibly better people” if rivals stop hiring. He points to IBM, which has specifically advertised that it is going to hire three times more junior staff than it has in the past, as an example of a company leaning into junior talent and likely to “benefit from that in the long-term.”

Nagle says too many companies are thinking more about the short-term and not considering the long-term as they roll out AI. He suggests that headlines about AI destroying jobs often obscure more mundane budget and strategy decisions. AI is “an easy scapegoat” for companies that overhired and over‑expanded and are now right-sizing while blaming technology rather than poor managerial decisions, in Nagle’s view. 

Nagle is less interested in AI hype and more interested in workflows, tracking who does what, for how long and how that shifts when AI enters the picture. 

From that work, Nagle sees three broad types of jobs emerging in an AI-driven economy. Some jobs will be pushed close to full automation, while on the other end, others will barely be touched. A wide band in the middle will survive, but change in deep ways. 

Bucket 1: AI will fully automate these jobs

The first group contains jobs with core tasks that AI can handle on its own. Here, AI does not just speed up work; it replaces large chunks of it. 

Nagle points to translators as the clearest example. “AI can translate very well, because there are massive amounts of text and it has been well-trained on that,” he says. 

That doesn’t mean that every translation job will vanish. Nagle notes that there will still be people acting as fact-checkers for AI translations “to make sure that it’s not messing up stuff.” He also points out that translators working in less widely used languages, or spoken-only languages, may still be in demand because AI hasn’t been trained on enough material to replace them. 

However, as a category, translation is Nagle’s “go-to example” of a job that AI can do very well and thus heavily automate.

Bucket 2: AI will barely touch these jobs

The second group contains jobs that are largely physical and hands-on. They are hard to automate with software alone. 

Nagle gives a simple example: “They are building a house right outside my window,” he says. “AI is not going to automate most of the tasks that the folks building the house are doing. The version of AI that we have today is not going to automate much of the guy who is climbing up a ladder to put a roof on the house.”

Geoffrey Hinton, called the “Godfather of AI” due to his pioneering work in the field, agrees with Nagle. In an interview last year on the podcast Diary of a CEO, Hinton said that AI would replace “everybody” in white-collar jobs, but take a long time to become good at physical manipulation. “A good bet would be to be a plumber,” Hinton said on the podcast. 

These jobs are not immune forever. Robotics could eventually transform parts of plumbing, construction, logistics and other manual work — but we are not there yet at scale, Nagle says. 

Bucket 3: AI will change the nature of these jobs

The final group Nagle outlines is the largest. These jobs will not disappear, but the work will change.

Software developers are Nagle’s core example; he has studied how developers use AI tools on the job. He breaks a developer’s job description into two main tasks: coding and project management work related to coding. What he observed was that every developer started to do “more and more coding” and “less and less project management” as a result of AI. 

For the junior developers, that shift was dramatic. “For junior people, that’s three to five times more of that kind of shift than more senior people,” Nagle says. 

AI makes coding “faster, easier and cheaper” so people do more of it and less of the surrounding coordination work, Nagle explains. He calls this a classic economic pattern — when something becomes cheaper and easier, people use more of it. 

Nagle says that this is not about “vibe coding” systems generating code for entire products end-to-end. Nagle says most large companies are still wary of putting that kind of code straight into production. Instead, developers are using AI to “enhance” code and make prototypes, while humans still design the architecture and meet user needs. 

Skills and majors

Nagle advises students to major in AI or computer science, where he expects demand is “only going to increase,” and the humanities, which equips people to be the “thought layer” deciding “why are we doing things and how are we doing them” as machines handle more execution. He adds a pragmatic third field: “anything to do with healthcare,” where demographic trends are driving sustained demand despite AI’s growing role in the sector. 

For universities and training programs, Nagle says the priority is understanding that students are going to use AI tools in the real world while still teaching them the fundamentals of how to think outside of these tools. He likens AI to calculators: Schools still teach arithmetic before letting students rely on devices.

Nagle expects the best institutions will both encourage students to use AI tools while also ensuring that they understand core concepts deeply enough to get the most out of the tools. These schools will allow students to recognize limitations and hallucinating behavior in AI. The goal is a generation of workers who can spot when AI is wrong and correct it. 

“We have to understand that students are going to use these tools in the real world,” Nagle says. “Therefore, we should be teaching and encouraging them on how to use them.”