They started college a few months before ChatGPT launched. They are leaving university into a job market that has already been partially reshaped by it. The Class of 2026 is the first cohort of graduates who never had to consciously decide to adopt AI - it was simply part of the furniture, available from their first semester, normalized before they knew enough to be suspicious of it. That is their advantage. It is also, depending on the industry and the week, their problem.

The tension could not be more plainly illustrated than in a pair of facts sitting side by side: some of the most sought-after recruits in corporate America right now are fresh graduates with strong AI instincts, and unemployed recent college graduates are simultaneously at one of their highest rates since 2013, outside the early pandemic. Both things are true at once. The entry-level job is not disappearing. It is bifurcating.

The Background

To understand what is happening here, it helps to understand what the entry-level job was originally for.

In most white-collar industries - finance, consulting, law, accounting, marketing - entry-level roles served a dual purpose. They were cheap labour for the repetitive, time-intensive work that senior employees could not be paid to do: building spreadsheets, compiling research decks, drafting first-pass documents, writing code to spec. And they were the training ground. The theory was that by doing the grunt work, junior employees would absorb the underlying logic of the business, make small mistakes, learn to ask better questions, and eventually become valuable.

Automation - the use of machines or software to perform tasks previously done by people - has been chipping at the first half of that deal for decades. Factory automation eliminated assembly-line manufacturing jobs in the 1980s and 1990s. Business software automated bookkeeping and administrative tasks in the 2000s and 2010s. But those disruptions largely spared the white-collar professional tier. The work was too complex, too context-dependent, too language-heavy for software to handle.

Then came large language models. An LLM - the type of AI that powers ChatGPT, Claude, and similar tools - is trained on vast quantities of text and can produce fluent, structured output: code, summaries, reports, slide content, email drafts, financial analysis. The output is not always right. But it is often good enough to replace the first pass, which is exactly what entry-level workers had been hired to produce.

The timing is awkward in the extreme. The Class of 2026 grew up watching AI get good at their first jobs before they were old enough to hold them.

ChatGPT launched in November 2022. Students entering college that autumn were, four years later, about to graduate into a labour market where the AI they had learned on had already started to do the work they had studied for. That cohort - the first to move through an entire undergraduate career alongside publicly available generative AI - is now walking off campuses and into offices that are still figuring out what to do with them.

What Is Actually Happening

The headline numbers are unambiguous. Unemployment among 22- to 27-year-old college graduates stood at 5.6% in March 2026, according to the Federal Reserve Bank of New York - one of the highest readings since 2013, barring the pandemic. The underemployment rate, meaning the share of graduates working in jobs that do not normally require a degree, sat at around 41.5%. For context: a decade ago, a college diploma nearly guaranteed a significant labour market advantage over non-graduates of the same age. That gap has narrowed substantially.

Companies cutting entry-level roles are rarely saying so out loud. The stated reason is usually "AI efficiency" or "restructuring." Big Tech firms saw new graduates fall to just 7% of hires in 2024, down 25% from the year before and more than 50% below pre-pandemic levels, according to research from venture capital firm SignalFire. Startups, once a reliable absorber of entry-level talent, dropped graduate hiring from roughly 30% of new hires in 2019 to under 6% in 2024.

At the same time, a different story is playing out in parallel. A survey of nearly 1,500 employers by Strada Education Foundation - published this month - found that among companies actively investing in AI, nearly three times as many expected it to boost entry-level hiring this year as expected it to decrease it. The share cutting junior roles did grow: to 17% from 13% in 2025. But the majority are adding, not subtracting.

The difference is what kind of graduate. SharkNinja - the appliance company behind Shark vacuums and Ninja blenders - held a two-day AI hackathon for students in April and is hiring about 200 AI-focused graduates and interns this year, including roughly 10 from the event. Mark Barrocas, the company's CEO, said the AI skills these graduates bring exceed those of workers with 20 years of experience. Salesforce says it is recruiting and fast-tracking 1,000 AI-native graduates this year for engineering, product, and sales roles it is explicitly describing as "hands-on, high-impact." IBM and MetLife are doing similar things.

The graduates who are landing these roles share a profile. Tommy Lee, who finished a business degree at Villanova University this month, spent over 800 hours experimenting with AI tools in his own time after taking a course on emerging technologies last autumn. He then automated his own job application process - building nine AI subagents to search for openings, tailor his resume, and pre-fill forms, while he focused his energy on networking. He starts next month at a private equity firm as an AI and systems analyst.

Emma Kanjorski, a finance graduate from the University of Vermont, landed a financial analyst role at an insurer after applying to around 40 jobs. She spent part of her final year teaching a professor how to prompt AI to critique its own output. She is acutely aware that her edge going into the role is not the finance knowledge itself, but the ability to deploy tools against it faster than someone who learned the craft manually.

The Money Trail

Follow the money and the logic becomes clear fast.

The economic value in a professional services firm - a bank, a consultancy, an accounting firm - sits primarily with the people who can make consequential judgements: senior analysts, partners, managers with context. The entry-level worker's value, historically, came from their time. They were cheap enough to deploy on the preparatory work that consumed hours, which freed up senior staff for the work that consumed judgement.

AI has changed the ratio. A senior analyst with a good AI workflow can now do significantly more of that preparatory work themselves, faster, without delegating it. The entry-level slot - always somewhat precarious, a function of cost arbitrage rather than irreplaceable skill - loses its justification when the cost structure shifts.

This is not a new dynamic. What is new is the speed and the sector. Previous waves of automation moved slowly enough for labour markets to absorb them through attrition and retraining. This one is moving faster than hiring cycles. Companies that start using AI tools heavily see their need for junior headcount drop before they have had time to figure out what junior employees should be doing instead.

The winners, financially, are companies that move early. SharkNinja is betting that AI-native graduates are a form of leverage: you pay them an entry-level salary, but they arrive already capable of deploying tools that amplify their output significantly. Salesforce is making the same bet at larger scale. For these companies, hiring 1,000 AI-forward graduates is not charity toward a disadvantaged job cohort. It is an attempt to build an AI-capable workforce at the lowest possible cost, before the market prices that skill premium in properly.

The losers are graduates who either cannot or did not develop those skills, and who are now competing for a shrinking pool of conventional entry-level roles. That pool is not zero - the Strada survey makes clear that most industries are not wholesale eliminating junior hiring. But it is smaller, and the roles that remain are increasingly divided between those that require AI fluency and those that AI has not yet been able to adequately replace: physical tasks, client-facing judgment calls, creative work that requires a specific human sensibility.

Leala Hernandez, a new San Diego State University graduate trying to find work as an accountant, put it directly: "I wish it wasn't here." The accounting profession is among those most exposed to AI-driven displacement of routine tasks - reconciliations, report generation, data entry - which are also the tasks that entry-level accountants have traditionally been hired to do.

There is also a credentialing problem embedded in the numbers. According to Strada's survey, employers ranked critical thinking as their top priority for entry-level hires. AI literacy came last - and it was the only skill category where graduates were rated higher than employers' stated needs. The gap is not technical. It is evaluative: employers are not sure yet how to assess whether a graduate who says they are good at AI is actually good at using it well, versus just using it at all.

What People Are Doing About It

Universities are attempting to get ahead of a problem they helped create.

At the University of Vermont, business management professor Rocki DeWitt stopped trying to police AI usage and started grading it instead. Students in her class had to submit their entire chatlog history alongside each assignment. DeWitt reviewed those logs and marked them up - evaluating not just the final output, but the quality of the prompts, which information the student chose to include or omit when querying the AI, and how they verified or pushed back on what the AI returned. The skill she was teaching was not AI proficiency in the abstract. It was something more specific: knowing when to trust the machine and when not to.

KPMG, one of the largest accounting and audit firms in the world, is piloting a new training program this summer for its audit interns that leans into the same logic. Instead of teaching AI tools directly, the firm is building gamified exercises around professional skepticism - scenarios in which interns must identify the flaw in an AI-generated analysis, or ask the right follow-up question before accepting a result. The underlying premise is that AI makes critical thinking more valuable, not less, because the cost of accepting a plausible but wrong AI output has risen.

The graduates who are navigating this best are treating AI as an apprenticeship accelerator. Elizabeth Awad, 26, who recently completed a two-year entry-level rotational program at Salesforce and has now moved into a senior product manager role, uses AI agents to handle meeting prep, draft documents, and organize her day - tasks that would have otherwise consumed hours of her time. She describes the shift as freeing up cognitive capacity for the judgment calls that actually matter. When she ran a demo of her AI setup within her team, engineers, designers, and other product managers cloned her workflow within a week.

For graduates who did not build these skills in school, the adjustment window is narrow. Several are using the job search itself as a crash course - building their own automated systems, teaching themselves through YouTube tutorials and open-source experimentation, in some cases automating parts of the application process itself. That self-teaching dynamic was visible in Lee's approach: the 800 hours he logged were not from a formal program. They were self-directed, on his own time, because he had decided the credential was less important than the capability.

The Bottom Line

The Class of 2026 is entering a labour market that is being rewired in real time, and the rewiring is not uniform. Some graduates - those who treated AI as a tool to develop rather than a threat to ignore - are arriving at the precise moment employers most want their skills. Others, particularly in fields like accounting, finance, and coding where the entry-level tasks map most neatly onto what AI can do, are finding the door narrower than any cohort in more than a decade. The diploma is not obsolete. But it is no longer sufficient on its own, and the gap between graduates who understood that early and those who did not is already visible in the unemployment data.

Timeline

  • November 2022 - OpenAI launches ChatGPT publicly, arriving just months after the Class of 2026 began their first semester of college.
  • 2019 - New graduates made up around 30% of startup hires; the baseline before the decline set in.
  • 2024 - New graduates fell to 7% of Big Tech hires, down 25% from 2023 and more than 50% below pre-pandemic levels, according to SignalFire research.
  • December 2025 - The Federal Reserve Bank of New York records graduate underemployment at 42.5%, near multi-year highs.
  • Early 2026 - Anthropic CEO Dario Amodei warns publicly that AI could eliminate up to 50% of entry-level white-collar positions within five years.
  • April 2026 - SharkNinja holds a two-day AI hackathon for students, hiring approximately 10 participants and planning 200 AI-focused graduate and intern hires for the year.
  • May 2026 - Strada Education Foundation publishes a survey of nearly 1,500 employers, finding that the share cutting junior hiring rose to 17% from 13% in 2025, even as most AI-investing firms still expect entry-level hiring to grow.
  • May 2026 - Federal Reserve Bank of New York data puts recent graduate unemployment at 5.6% in March, one of the highest readings since 2013 outside the pandemic.
  • May 29-30, 2026 - The Wall Street Journal publishes reporting on the Class of 2026 as the first AI-native graduate cohort, drawing on individual accounts from employers including Salesforce, SharkNinja, IBM, MetLife, and KPMG.

Summary

Who: The Class of 2026 - the first cohort of university graduates who completed their entire undergraduate education with access to generative AI tools.

What: Graduates are entering a labour market that is simultaneously creating demand for AI-fluent junior workers and eliminating conventional entry-level roles faster than new ones are being defined. Recent graduate unemployment sits at 5.6%, near a 13-year high, while companies like Salesforce and SharkNinja are aggressively recruiting AI-native talent for expanded entry-level positions.

When: May-June 2026, as spring graduation ceremonies conclude and the Class of 2026 begins entering the workforce in volume.

Where: Primarily the United States, with the dynamics most acute in white-collar sectors including technology, finance, consulting, and accounting.

Why: AI tools - particularly large language models - are capable of performing the preparatory, repetitive tasks that entry-level workers were historically hired to do cheaply. Companies that adopt these tools aggressively find they need fewer conventional junior hires, but place a premium on graduates who can deploy AI as a force multiplier. The result is a bifurcated entry-level market: high demand for graduates who built AI skills, fewer opportunities for those who did not.