What employers want, and what college delivers.

I was a History major in college, and loaded up on English and Government courses while avoiding science and math like the plague. And yet I quickly landed in tech after assessing my “History” job prospects.

I have always thought that all that time spent reading, writing and assembling arguments has served me well in my career. Storytelling, understanding diverse viewpoints and needs, even investigating root causes are day-to-day activities for project, program and product managers. While I added useful technical skills to my bag along the way, those are the skills that have endured.

My first job in tech was customer support. I learned how to talk to both non-technical users and very technical engineers at the same time. I leveraged that experience to become a first-time manager of a CS team. The people who saw me in that role thought my problem solving and communication skills could transfer to running a complex engineering program and I was off and running.

But over the last few decades, amidst the idolatry of the creators of our tech-driven world and a broad move towards coding and STEM courses, the “worthless” humanities degree has become a meme — a pipeline to employment as overeducated baristas.

While that underemployment for humanities grads shows up in the data, the cruel irony is those skills are very much in demand. But new graduates are both arriving in the workforce without them, and missing the opportunities to develop them.

Individual stories and anecdotes are one source of truth. But I like data, so I took a fresh look at what the data says about skills and the hiring market.

tldr;

—> Employers increasingly say they need durable soft skills. This is a no-brainer as AI collapses the boundaries between functional roles.

—> But the data continues to show that the lowest rungs of the career ladder in the most AI-exposed industries are disappearing.

I don’t think the answer is simply “more history majors.” We need a serious investment in developing trade skills at the same time there has to be a serious commitment to developing critical thinking and communication skills at a time when AI can answer any undergraduate test question.

Employers are hiring for durable skills.

76% of job postings now ask for at least one durable skill, up from 64% in 2019–20. Nearly half ask for three or more. Durable skills are the “soft” skills that transcend job categories and outlast automation and the adoption of new technologies. The jumps are steep in technical fields that have put a premium on hard skills — Computer & Mathematical rose 15 points, Finance & Insurance 16 — while fields that historically valued durable skills like Sales & Marketing and Management barely moved.

Even the AI jobs are hiring for human skills.

Communication alone appears in 26 million postings. Across the labor market as a whole, eight of the ten most-requested skills of any kind are durable. Inside AI-enabled roles specifically, it’s also eight of the top ten.

Students think they are developing these skills in college. Employers disagree.

Paired 2024 surveys by the National Association of Colleges and Employers (NACE) exposed this gap. 78% of students call themselves proficient communicators; 54% of employers agree. The gaps run 25 to 35 points on communication, critical thinking, professionalism and leadership. Teamwork is the only competency where the two sides agree. Technology use is the only area where students underrate themselves.

The push towards CS and other STEM degrees has left humanities in the dust.

In 2008 the country graduated roughly as many History majors as Computer Science majors. By 2023, CS was 5.4x larger — up 195% while History fell 38% and English fell 43%. The decades-long push towards CS degrees has left fewer students benefiting from the deep focus that humanities courses provide on communication and critical thinking, missing out on developing the skills that might best prepare workers for the AI economy.

Graduating with a History major still may not pay, but being a CS major has stopped being safe.

Data from 2026 shows that my fellow History grads aren’t faring too well salary-wise. While CS degrees still pay far better than history degrees ($87K early vs $47.5K) and History’s underemployment is 50% vs CS’s 19%, Computer Engineering and Computer Science now post the 2nd and 4th highest unemployment rates of 74 majors — 7.8% and 7.0%. History sits at 4.3%, right at the median. The trends will likely continue: As AI does more and more coding, CS majors could be chasing fewer higher-paying jobs that go to “stars.”

The career ladder rung where young employees develop durable skills is disappearing.

Younger workers are losing the opportunity to develop durable skills on the job. Employment for 22–25 year olds in the most AI-exposed occupations has fallen about 12% since late 2022, while their peers in less-exposed work grew. This cannot be directly attributed to AI adoption, however, as tech over-hiring in 2020 and 2021 and rising interest rates have softened the labor market in tech and other exposed occupations.

Coaching on the job represents an opportunity to close the gap, and employers are missing it.

Studies of professional coaching’s measured effect show it can move the needle on the gap: skill outcomes, g = 0.72. But only 27% of employees get a professional coach, up from 23% in 2019, and access is mostly reserved for senior staff. The coaching market grew 17% from 2023 to 2025, but the share of people it reaches barely moved.

Students are already adapting to the new reality.

New CS enrollments fell 12.9% this past academic year, and overall CS enrollment fell for the first time in two decades. But Higher Ed is notoriously slow to adapt. And with the rampant embrace of AI for research and writing (you said “cheating,” I didn’t), it’s foolish to assume that embracing humanities courses — as they have been constructed for decades — could actually address this gap. Curriculum and courses have to be re-thought to embrace AI literacy, classroom discussion, project-based work and written exams that can’t be done with LLMs.

Meanwhile employers are pulling back from the graduates they say they need.

Three quarters of employers now hire the same or fewer entry-level workers than a year earlier, and degree requirements for those roles went back up, from 55% to 71% — a middle finger to the notion of skills-based hiring. And managers who would have coached those hires are stretched thinner than ever: spans of control at small and mid-size firms have roughly doubled since 2019.

The trend might be correcting this year. The evidence is thin so far.

Employers say they are ready to hire again. NACE revised its Class of 2026 projection up to +5.6%, from +1.6% in the fall, with the gains concentrated at employers of 5,000 or more.

Payroll data is showing a marginal improvement over last year. Every June, employment for 22–25 year olds in the occupations Stanford ranks most AI-exposed bumps up as the graduating class starts work. It is the only month that reliably rises for this group: +1.1 points on average, against −0.1 for every other month. No older age band shows it, and it barely registers in less-exposed occupations.

For 2026, the pulse is still there, but far smaller than it was. June added 2.5 points in 2022 and 2.0 in 2023. Since then: 0.4, then 0.1, then 0.4 this year — the second-weakest June in the series. Graduates are still being hired into these jobs. There are just many fewer of them.

Sources

Lightcast / America Succeeds · NACE · NY Fed · IPEDS · Stanford Digital Economy Lab & ADP · Frontiers in Psychology · ICF / Human Capital Institute · CRA · Cengage · Gusto

Each chart carries its own sources-and-methods note with the caveats specific to that measure.