Is AI Disappearing Early Career Roles? Parsing the Data From Three High-Profile Studies

Headlines about AI wiping out Early Career jobs are everywhere. Some cite macroeconomic data showing graduate unemployment rising in AI-exposed jobs. Others point to companies refusing to approve headcount unless the role cannot be done by AI. Still others rely on CEOs’ sweeping declarations that AI will replace swaths of Early Career knowledge workers.

Yet when you talk to employers, many insist the opposite: they’re not pulling back on Early Career hiring; in fact, some are doubling down.

So what’s really happening? What is the true story of AI and Early Career roles?

To answer that question, we examine three high-profile studies that appear, at first glance, to contradict one another, but taken together, they reveal a story that is both stark and nuanced: AI is not eliminating Early Career work, but it is reshaping the foundations of how young professionals enter organizations – and forcing us to rethink how we prepare them for careers.

Three Studies, Three Different Stories

Stanford Says Young Workers in AI-Exposed Jobs are Losing Ground

A new study from Stanford researchers Brynjolfsson, Chandar, and Chen (2025) uses ADP payroll data on millions of workers between 2021 and 2025. They find that employment for 22- to 25-year-olds in highly AI-exposed jobs – such as software engineers, customer service representatives, marketing and sales managers – fell by 6% between late 2022 (when ChatGPT launched) and July 2025. The story is different for older workers: in the same occupations, employment for more experienced employees has held steady or even grown.

The authors point to the mechanism driving this divide:

• Decline happens in roles where AI automates junior-level tasks.
• Growth occurs in roles where AI augments those tasks.

 

Why the divide? AI is strongest at tasks that rely on codified knowledge – the kind of information learned from textbooks or coursework. It is far weaker at tasks that require tacit knowledge– the kind of judgment, intuition, and context one can only develop on the job. Since new graduates enter the workforce primarily with codified knowledge, they are disproportionately vulnerable in AI-exposed occupations where automation dominates.

Employ America Says the Problem Predates AI

Employ America looks at the same issue as Stanford but lands in a very different place: AI isn’t the root cause of rising graduate unemployment, but a convenient scapegoat for a trend that began years earlier.

Drawing on data from the Bureau of Labor Statistics (BLS), they show that the “recent graduate advantage”– that is, the long-standing advantage young degree-holders once had in the job market – has been eroding since the 2010s. By 2018, four years before ChatGPT’s release, unemployment among 22- to 27-year-olds had already surpassed the national average.

They also challenge the idea that AI-exposed majors are uniquely at risk. Yes, computer science graduates have faced higher unemployment in recent years. But other high-exposure majors – like math, accounting, and business analytics – have not shown the same pattern. In some cases, unemployment rates in these fields have improved.

Instead, Employ America points to more traditional explanations:

• Graduate oversupply: Colleges are producing more degree-holders than ever, especially in popular fields like computer science. This means employers have a larger supply to choose from, which reduces the likelihood of an individual being hired.
• Weaker overall hiring: The overall hiring rate has slowed, which could be making it more difficult for recent graduates to enter the workforce.

 

In their telling, AI isn’t rewiring the labor market – at least not yet – but rather, is amplifying pressures that have been building for over a decade.

Burning Glass Says the Story is More Nuanced

In its landmark report No Country for Young Grads, the Burning Glass Institute confirms rising unemployment among recent degree-holders, but frames it as the product of several converging forces rather than AI alone.

They identify four key drivers behind the breakdown of the college-to-career pipeline:

1. AI erasing “Growth Roles.” These are jobs where junior workers handle simple tasks while senior workers tackle more complex ones. When AI absorbs the simple work, it eliminates a natural entry point for young talent.
2. Post-pandemic leanness. Many companies discovered they could maintain revenue without rebuilding headcount after COVID. Leaner staffing has stuck.
3. AI as accelerant. Far from being the sole cause, AI is speeding up efficiency trends that were already underway, giving firms the tools to do more with less.
4. Graduate oversupply. More degree-holders are chasing fewer jobs that require degrees. One year after graduating, 52% of the Class of 2023 were in roles that didn’t require a degree at all.

 

Together, these dynamics create what Burning Glass calls a “thriving economy that doesn’t need new graduates,” one which could pose an existential threat to traditional on-the-job learning pathways.

Why These Studies Don’t Actually Contradict Each Other

At first glance, the three studies seem to be at odds. But the tension dissipates upon realizing that they’re operating at very different levels of analysis.

• Employ America looks at broad sectors of the economy. For example, the “Information sector” lumps in everyone from software developers to journalists to telecom workers. When they report stable employment, it is because they are averaging across all of those jobs.
• Stanford, by contrast, follows specific people in specific occupations month over month. That allows them to spot finer-grained effects, such as a 22-year-old software developer who loses their job in March 2023, that would get washed out in sector-level data.

 

The same logic applies to timing.

• Employ America shows that graduate struggles began well before generative AI, with unemployment rising as early as 2017.
• Stanford shows those struggles deepened after ChatGPT’s release.

 

Both can be right. AI didn’t start the fire, but it poured fuel on it.

What the Evidence Actually Shows

Taken together, the three studies, plus supporting evidence, sketch out a clear timeline of how recent graduate employment has shifted:

• Pre-2018: An oversupply of college graduates meets changing skill demands. The bachelor’s degree slowly loses its premium.
• 2018-2022: Recent graduate unemployment surpasses the general population. COVID accelerates lean staffing models, and the Great Resignation makes companies more cautious about investing in junior talent.
• Late 2022-Present: Generative AI arrives. Certain occupations – especially those where AI can automate rather than augment Early Career tasks – experience sharp declines in junior hiring.

 

If we accept this timeline, then what exactly is the mechanism displacing Early Career talent?

A recent paper by Autor and Thompson (2025) out of MIT sheds light on this question. One of their major findings is that when routine tasks are automated, the leftover work is more complex. That raises the skill threshold of work, reducing the pool of qualified workers but also boosting wages for those who meet the bar.

This framework explains why AI can be especially damaging to Early Career roles. These jobs are often designed around routine tasks that junior workers perform while building up to expert work. If AI wipes away those stepping-stone tasks, what remains are higher-level responsibilities that new grads aren’t equipped to handle.

So, Is AI Disappearing Early Career Roles?

The short answer: yes – but under very specific conditions.

AI is eliminating Early Career jobs for young, college-educated workers when three factors line up:

1. The occupation is highly AI-exposed.
2. AI is being used to automate (not augment) core tasks.
3. The remaining tasks demand expertise beyond what new graduates bring from school.

 

That’s a narrow but important slice of the labor market. It helps explain why headlines about disappearing junior roles aren’t pure hype, while also clarifying why many employers continue to invest in Early Career hiring.

Outside of these conditions, we have little evidence (so far) that AI is causing broad, economy-wide job loss for new graduates.

What This Means for Recruiting Leaders

Not every Early Career role is equally threatened by AI. But the broader college-to-career pipeline is under real strain. For recruiting leaders, the challenge is twofold: in the short term, protecting Early Career roles, and in the long term, rethinking how you bring in Early Career talent.

Short-Term: Protecting Early Career Roles

Recruiting leaders can take a number of practical steps to protect and enhance their Early Career opportunities:

1. Audit AI exposure in Early Career roles.

Which roles are most vulnerable to automation? Break down the tasks within each position to see whether AI is being used in a substitutive way (replacing work) or an augmentative way (enhancing human output). This baseline will guide where intervention is needed.

2. Redesign tasks, don’t just remove them.

If AI is erasing too much of the “basic work” that once trained new hires, don’t leave a vacuum. Offload repetitive tasks to technology, but add responsibilities that build judgment and career capital – like data interpretation, client interaction, or cross-team collaboration.

3. Pilot augmentative use cases.

Look for ways AI can amplify, not replace, junior employees. Tools that speed up research or draft-building, for example, can free up time for higher-order skills like analysis and presentation.

4. Make learning value explicit.

Ensure job descriptions and manager expectations emphasize the developmental aspects of Early Career roles. If a position is changing, spell out the skills and experiences new hires will gain. That clarity helps preserve the role’s value, both to the employee and to the organization.

Long-Term: Rethinking How We Bring in Early Career Talent

The old model was simple: hire promising graduates, assign them routine work while they shadow senior colleagues, and gradually increase their responsibilities. That model worked because there were enough basic tasks to justify the hire. Once AI takes on those routine tasks, the logic breaks. But alternative models are emerging that could rebuild the pipeline in new ways:

1. Hybrid education–employment partnerships.

Some universities already integrate extended work placements into degree programs. Northeastern University’s co-op program, for example, alternates semesters of full-time study with semesters of full-time, paid employment. Students graduate with up to 18 months of professional experience. Scaled further, such models could allow graduates to spend part of the week embedded in real roles while still completing formal education – sharing costs and stretching out the pipeline in sustainable ways.

2. Near-peer mentorship ladders.

Senior leaders are often stretched too thin to provide regular hands-on training. But professionals just a few years into their careers are well-positioned to coach those entering the workforce. Structured near-peer mentoring systems create a ladder effect: each cohort teaches and reinforces knowledge for the group below them, while deepening their own expertise in the process.

3. “Tour of duty” employment contracts.

Coined by LinkedIn founder Reid Hoffman, this model reframes Early Career roles as defined missions lasting two to four years. The company provides structured development and a credential at the end; the employee commits to delivering results during that period. Unlike traditional open-ended roles, a tour of duty makes the temporary, learning-oriented nature explicit.

4. Rotation programs with shared investment.

Sustaining true apprenticeships in knowledge work is expensive for any single employer. Consortium models – where multiple companies and sometimes government agencies share the cost – could make it feasible. Germany’s dual vocational education system offers a template: students split their week between classroom instruction and paid on-the-job training, with government, employers, and unions all sharing responsibility.

What’s Temporary, and What Change is Here to Stay?

For anyone who lived through the dot-com bubble, it’s fair to ask: how much of what we’re seeing now is a temporary shock, and how much represents a lasting shift? It is likely a mix:

Probably Temporary:

• Post-pandemic leanness. Companies could eventually rebuild headcount as economic confidence returns.
• Great Resignation risk aversion. Employers may regain confidence in investing in Early Career talent as fears of young job-hoppers subside.

 

Probably Permanent:

• Graduate oversupply. Demographics and education trends suggest this will continue for the next decade at least.
• The expertise-bar problem. Once AI is brought on to do junior-level tasks, there is no economic reason for companies to go backward.

 

Unknown:

• How far AI automation goes. Will capabilities plateau at current levels or keep climbing the ladder of expertise?
• Whether new Early Career roles emerge. Historically, technological disruption creates new jobs in addition to displacing some. But will these new roles be accessible to new grads without prior experience?

 

This mix of change suggests we will not return to the “old normal,” but we may reach a new equilibrium. The open question is whether that equilibrium still includes viable pathways for young professionals to develop expertise.

The Uncomfortable Reality

The three studies here tell the same story from different angles. Stanford traces the sharp edges in specific occupations. Employ America identifies the longer-term context. Burning Glass connects the dots.

Together, they point to a sobering conclusion: the traditional pathway of knowledge work – starting with simple tasks, then progressing to complex ones – is breaking down. For the first time, we may be facing a generation of workers expected to arrive as “experts” in roles that once served to build that expertise.

The good news is that companies committed to Early Career talent still have options. By redesigning Early Career roles, experimenting with new models, and making deliberate investments, they can both rebuild the on-ramps to expertise and prevent a generational rift in professional readiness that could haunt organizations for decades.