The Most Durable Jobs in the AI Era — And No One’s Training for Them - Matt Sigelman
“We know which jobs are resistant to automation. We just haven’t figured out how to get people into them.”
Everyone’s asking which jobs AI will take. Fewer are asking which ones it won’t — and why we’ve made those the hardest to access.
Matt Sigelman is one of the few people studying that question with real data. He’s the President of the Burning Glass Institute and co-author of the widely cited report No Country for Young Grads. His work tracks how AI, demographics, and employer behavior are reshaping the labor market in real time.
Our conversation is, in part, about the other half of the automation story: Not what AI will do, but what it leaves behind — and why we’ve failed to build the pathways to get there.
We get into:
• Why applied jobs — in healthcare, skilled trades, and logistics — are AI-safe but structurally ignored
• How workforce training systems broke, and what a better version might look like
• Why the collapse of the early career ladder could push students toward more interesting, expansive lives beyond Big Tech and Big Banking
This transcript has been edited for clarity and length.
On The Early Career Ladder’s Silver Lining
ALLISON: What is a belief you hold now that you didn’t believe a year ago?
MATT: When companies like Google, Amazon, and Goldman Sachs stop hiring at the entry level due to AI, college students must broaden their view of opportunity and rethink what a meaningful career and life look like. I now see this as the silver lining to the early career labor cloud: young people may now make a broader impact than they could through narrow pipelines into Big Tech or Big Banking.
I do worry that high school and college students already on this track will be automated at the entry level. Colleges have been largely ineffective at providing solid career guidance, so students will need to supplement their learning. But how to do that remains an open question. I tell students to study what employers want: look at LinkedIn, read job ads, and build a skills inventory for top roles in their field. Then, work to gain those skills and learn to articulate how they show up in your education and experience.
ALLISON: What’s a belief you hold that most people would disagree with?
MATT: Much of my career has focused on how skills can close labor market gaps, so I strongly advocate helping students gain the right skills to launch their careers. But in emphasizing skills, many educators and policymakers lose sight of education’s deeper purpose: helping students discover what is beautiful and true in the world.
To do this, you can’t separate skill acquisition from content. They are not agnostic; they are intertwined. My wife and I founded a school 11 years ago based on this premise. We ask: how do we ensure kids read extraordinary books and explore the natural world’s mathematical beauty? This approach makes education harder, but far more rewarding.
ALLISON: That’s beautifully said, and a good reminder that education, especially in a democracy, serves the whole person, not just our economic selves. We’re also citizens and purposeful human beings. Education often gets reduced to skills in discussions of opportunity and mobility, simply because skills are easier to measure in relation to opportunity. Thank you for grounding the conversation in this broader context.
MATT: Kent McGuire describes human thriving as education’s core goal. Part of that thriving is economic: students and families consistently see education as a bridge to opportunity. For education to be truly student-centric, it must also be opportunity-centric. But we can’t reduce it to opportunity alone.
On AI’s Real Impact on the Early Career Crisis
ALLISON: Burning Glass Institute released one of the best reports I’ve seen on the early career ladder, No Country for Young Grads. It showed demand shifting from entry-level to senior-level talent. How is AI accelerating that collapse, and what concerns you most?
MATT: The keyword is acceleration. AI is just one of several forces reshaping the early career ladder, including a flood of college graduates, labor shortages in non-degree high-paying fields, and employers adopting low-employment, high-productivity models that rely less on people, partly due to the pandemic. AI adds fuel to that fire.
The biggest demand shifts toward senior talent are happening in fields where AI is deployed most. This relates to learning curves. My colleagues Joe Fuller, Mike Fenlon, and I studied these curves across occupations by mapping salary growth against productivity. Some jobs require a lot of early expertise but offer limited long-term growth — those tend to pay less. The best-paying roles, often degree-based, have long learning curves: you’re hired for potential, not mastery, and you build expertise through unglamorous early work. That early-stage work often overlaps with AI’s strengths, which explains part of what’s happening.
At the same time, in other fields, AI could expand access to well-paying jobs by helping people master knowledge faster and overcome language and even social capital barriers — but many of those roles don’t require degrees, which is tough news for college grads.
ALLISON: What top job categories might AI open up?
MATT: Technical, clerical, and customer service roles — areas with knowledge barriers that AI can help people overcome quickly.
ALLISON: Right. Not the jobs most college graduates want.
MATT: First, while that’s often true, there are in fact plenty of jobs with high wages and good career prospects for graduates. Think of roles like data warehousing specialists, network administrators, or rehab counselors. And, even if many professional jobs will see diminished access, there are steps we can take to ensure it’s not all doom and gloom. AI’s impact creates two imperatives. First, employers must define proficiency levels more clearly for new hires. Second, colleges must help students gain real experience before graduating. If students are expected to start careers mid-ladder, we need to build that learning earlier.
That’s difficult, especially since fewer entry-level jobs mean fewer work-based learning opportunities. Yet those experiences are crucial; they let students apply skills in real-world settings and demonstrate them to employers. It’s worth it for companies to partner with institutions to make these structures work.
ALLISON: What’s the one thing you wish higher education leaders understood about preparing learners for the future of work?
MATT: Public education isn’t used to pivoting quickly. Higher education leaders must recognize that AI is already reshaping work, and they need to respond accordingly, now.
With that in mind, I hope they’re redefining how teaching and learning should evolve with AI. Much of the discussion focuses on personalizing learning or preventing cheating, but not enough addresses what students need to learn differently now that this technology exists. We need a framework for identifying which skills will become more important and which must be taught in new ways across domains.
ALLISON: Last question about the early career ladder. How soon will we see the impact on entry-level jobs, and how significant will it be?
MATT: Not all entry-level positions will evaporate, and that won’t happen overnight. So far, we have seen a 9% decline in entry-level hiring compared to expert hires. That is obviously not the entire early career market, but still a meaningful shift. Recent graduates face much more competition than before. It’s not that no one is getting hired, but they have to work much harder to land those roles.
On What It Takes to Get More People into AI-Safe Jobs
ALLISON: Let’s talk about applied jobs. Jobs that require skill of both body and mind are more resistant to AI automation: healthcare, skilled trades, high-tech manufacturing, etc. Yet we just don’t see skilling programs that have succeeded in scaling and moving the appropriate number of skilled workers into these applied jobs. Why have we failed to innovate here?
MATT: First, thanks for spotlighting these jobs. As mechanical automation replaces frontline workers and generative AI replaces many professional roles, skilled technical jobs that rely on human judgment and are hard for machines to replicate will be a “safe center”.
To answer your question, we haven’t been training people for these roles, not from a lack of awareness but really a lack of investment. We spend 54 times more on college degrees through Pell Grants than on workforce training through the Workforce Innovation and Opportunity Act.
It’s easy to say, “Let’s spend more,” and we should — but we also need to recognize the sober reality that existing training programs often don’t work. The TAACCT grant program showed no evidence that its graduates found well-paying jobs. Studies of WIOA training found little success, either. Programs rarely align with local demand, we lack solid data systems to measure outcomes and scale what works, and employers often fail to anticipate labor shifts or prepare their workforce for new roles.
One promising approach is understanding and matching skill adjacencies. For example, the large home health aide workforce faces low pay and high turnover, but clearly has interest and transferable skills for better healthcare jobs. We should better match people to community opportunities that build on what they already know — training only where new skills are needed.
Home health aides illustrate this perfectly. They have broad foundational skills that can be specialized in many directions. Their work corresponds closely to higher-paying clinical and technical roles. My friend Van Ton-Quinlivan at Futuro Health has done excellent work helping low-wage healthcare workers advance — often without leaving their current employer. Hospitals, too, have entire teams who could reskill into clinical or tech roles if clear pathways existed. Futuro is building those pathways.
On What Parents Should Know To Guide Their Children’s Learning
ALLISON: What do you tell parents who ask what their kids should learn to thrive in the next 10–15 years?
MATT: I encourage a twofold approach.
First, students need a strong grounding in the core liberal arts. Foundational skills – critical thinking, analysis of texts and cultural artifacts, scientific experimentation, civic and historical literacy – map across nearly every high-value domain today.
Second, students need data, digital, and core business skills like project management. Both areas will require higher levels of proficiency.
Thankfully, AI can help build that proficiency. We recently modeled which workplace skills would be automated versus augmented by AI, and many skills appeared on both lists. That’s no accident — AI can make people both more efficient and more effective simultaneously. Writing illustrates this well: AI can draft text, but it cannot replace argumentation, rhetorical craft, or finding words that resonate. Right now, students often outsource their voice rather than develop it; we need better instruction to retain and coach this human work.
Finally, some say skills like coding or graphic design won’t be needed because AI handles them so well. There may be less demand for coders and graphic designers, but that means that all of us are going to need to be able to do those things precisely because the skills are more accessible, in the same way that Excel made data analysis accessible. I wouldn’t necessarily push my child to become a software developer, but I would ensure they develop strong computational thinking.
One Small Signal
ALLISON: What’s a small signal in the world right now that you think will become more important over time?
MATT: We may be seeing a reversal of our “causal” labor market, as jobs become more complex, and expertise is increasingly prized. Employers can’t just hire on the spot for roles in this labor market.
This is important because it puts pressure on companies to invest in training and workforce development.
In addition to more investment in training, I expect employers will put more effort into the economic mobility of their workforce, both as a way of building the value of their talent over time and as a retention measure for their highly-skilled, well-trained employees.
ALLISON: That’s a beautiful view of the future. It’s certainly one that I would love to live in. And AI may be the tailwind that creates this reality.
Explore more conversations on the early career ladder:
AI at Work: What Zapier’s Chief People Officer Thinks Comes Next - Brandon Sammut
Jailbreaking the Degree; Keeping the Pipes: David Blake’s Redesign for Higher Education