Labor economists say AI is not what’s driving the hiring freeze

The freeze isn't being driven by AI

The most striking reframe of the evening came from Daniel Zhao, Chief Economist at Glassdoor, on what is actually responsible for the current labor market freeze. Despite the constant AI layoff coverage, Zhao argued that the technology isn't the primary force shaping the slowdown. "The current state of the business cycle is not something that's being driven by AI or by the labor market," Zhao said. "It's things like policy uncertainty, energy prices."

Glassdoor's data backs the diagnosis, as the company's Employee Confidence Index hit a record low in April 2026, with only 43.8% of employees reporting a positive six-month business outlook – the weakest reading since the index was created in 2016. Glassdoor's analysis attributes a significant portion of the drop to the current geopolitical climate and the resulting spike in energy prices.

If AI were the primary driver, the unfreezing would depend on technological adoption patterns. However, because the freeze is being driven by macro uncertainty, the unlock depends on entirely different variables, which is why employers waiting for AI clarity may be watching the wrong indicator.

What makes the current moment feel different from past tech-driven shifts isn't the scale of the change but the speed, Thomas argued. "Adoption of this kind of technology is a lot faster," she said. Drawing a comparison to past economic shocks like trade, she added that "a lot of the negative impacts happened because of speed." In her framing, disruption that would have eventually happened from a slower technological shift is being compressed into a window short enough to register as a shock.

The Class of 2026 is being underestimated

Randy Tarnowski, Head of Economic Research at Handshake, presented data showing that Class of 2026 resumes mention AI skills more than nine times as often as Class of 2022 resumes did at the same career stage. Two-thirds of those mentions come from students outside computer science majors, according to a Handshake report on the cohort.

Thomas explained that employer demand is catching up to the supply, but not quickly. The share of active job descriptions mentioning AI-related skills has more than doubled since January 2023, from 2.3% to 5.6%. For roles targeting 18-to-24-year-olds, the share has nearly tripled.

The result is a cohort that has built deep AI fluency through the entirety of its college career – they were freshmen when ChatGPT launched in November 2022 – walking into a market still calibrated around AI as a senior-level differentiator rather than a baseline entry-level skill. "Freshmen and CEOs have access to the same tools," Tarnowski said. He described entry-level workers as "at the tip of the spear with these tools," and noted that "early talent is often overlooked as skilled AI users."

The paradox is that this is also the cohort feeling worst about its prospects. Glassdoor's confidence data shows entry-level workers tracking below the overall index, and Thomas's research shows under-25 job hopping has dropped by more than half – the steepest decline of any age group. The most AI-prepared cohort to ever graduate is simultaneously the most stuck.

The hiring funnel broke on both sides

The application process itself was the longest sustained conversation of the dinner, with both economists and reporters circling the same diagnosis: it has structurally failed both candidates and employers. Zhao framed it as a coordination problem rather than anyone's specific fault. "Everyone's behaving irrationally," he said. "Once you lower the barriers so much to submitting applications, that is kind of a natural consequence."

The math justifies the flood. If 100 applications yield roughly one callback, and that callback doesn't yet lead to a job, sending 500 becomes rational behavior. From the employer side, the same volume forces hiring managers into 15-second resume scans because spending five minutes per application doesn't scale. Neither side can afford thoughtfulness because the volume has made thoughtfulness uneconomic.

The downstream effect is visible in how companies are now sourcing. Tarnowski described a deliberate return to face-to-face hiring – campus visits, in-person job fairs, and direct outreach – as employers try to remove AI from the sourcing process entirely. The internship pipeline has also lengthened: companies are recruiting interns up to 18 months in advance, meaning high school sophomores are now applying for internships they will take during college.

The interview side has degraded for a related reason. Zhao noted that hiring managers are increasingly seeing near-identical answers from candidates to the same questions – the signature of AI-generated responses – with the candidates who stand out being the ones who customize or build on the prompts they use. The result is that employers are putting more weight on signals AI can't easily fake, such as referrals, in-person observation, and demonstrated work.

Thomas's advice for job seekers reflects the new reality. She tells people not to apply to 200 companies at a time and to lean on connections and networks for referrals. "You'll get a much better return on investment," she said. She compared the U.S. college application system, where students can submit to as many schools as they want, to the U.K. system's five-school cap, where targeted applications produce healthier acceptance rates because applicants signal real interest in each one.

A middle-management cliff is forming under the freeze

The most forward-looking framing of the evening came from Guy Berger, Workforce Economist in Residence at Guild. While companies remain frozen by uncertainty, he argued, they are quietly engineering a hiring crisis for themselves a decade out. "Who's going to be managing all these agents?" Berger asked. "We didn't hire anybody. And you didn't hire anybody who can actually use this technology… At some point, you do have to start hiring young people because those are your future middle managers."

Tarnowski's data on entry-level workers as the first true AI natives sharpens the stakes of Berger's point. The companies pulling back hardest from early-career hiring right now are the ones most exposed when AI competence becomes a baseline expectation across the org.

Layered on top of the freeze is a demographic dynamic that, in Berger's framing, didn't apply in past slowdowns. "Demographics are very different [from what] they were back then," he said, referring to past slumps. He warned that in 10 years, companies could find themselves struggling to find young workers at all – if current demographic patterns hold.

Tarnowski's research adds another layer to the argument. In Handshake's data, the number-one indicator of optimism about the job market, and stronger than any demographic variable, was AI usage. AI users were more optimistic than non-users. Combined with the AI-fluency data on Class of 2026 graduates, it points to the same group at the center of both findings: early-career workers who already understand the tools companies will need them to deploy, but who are feeling the chill of a market still calibrated to the old playbook.

The market behind the headlines

The dominant narrative isn't wrong, exactly. AI is reshaping hiring, and the entry-level market is brutal. But the four findings that Deel’s event surfaced complicate the headline story in important ways. The freeze has drivers that aren't AI. The cohort everyone assumes will be hit hardest by AI is the one most fluent in it. The application process is collapsing under its own weight, and companies are responding by returning to the oldest sourcing tools they have. And the freeze itself is staging a future scarcity that companies have barely begun to think about.

Taken together, the four threads don't add up to a single thesis. What they add up to is a more careful read of what's actually happening in the labor market right now, and a clearer view of where the headlines have gotten it half-right.