Executive Summary

Key Figures

Source: PwC 2026 Global AI Jobs Barometer

The four numbers below tell one story. The skills demanded of junior roles moved up to senior level 7× more often; more than half of the skills newly appearing in entry-level ads were ones once asked of experienced hires. Over the same stretch, ordinary junior openings fell while "seniorized" ones rose, and the going rate for someone who can work with AI more than doubled in two years.

For two years, the labour-market news has run in one direction. Stanford's Digital Economy Lab analyzed payroll data covering one in six U.S. workers and reported that entry-level hiring in highly AI-exposed roles fell 13% relative to less-exposed roles. Employment of software developers aged 22–25 dropped roughly 20% from its late-2022 peak. The headlines race to a familiar conclusion: AI comes for the newcomer's seat first.

But there's another photograph of the same market, taken from a different angle. When PwC combined more than a billion job ads with company financial and role data, it found that the firms most exposed to AI grew their headcount faster, not slower. Headcount at the most AI-exposed companies rose 52% versus 2018 — well ahead of the 36% at the least-exposed firms. Productivity tracked the same way: labour productivity at the top 20% of companies grew several times faster than the overall average over the period.

Set the two photographs side by side and they look contradictory. On one side, juniors are vanishing; on the other, total employment is climbing. The key isn't "volume" — it's "entry." Jobs didn't disappear wholesale; the doorway newcomers used to walk through to join a company got narrower. In the Stanford researchers' phrasing, technology doesn't erase the work so much as it erases the on-ramp.

PwC looked separately at 2.4 million U.S. entry-level jobs. It found that junior roles with high AI exposure were 7× more likely to demand "human-intensive" capabilities once expected of experienced hires (leadership, creativity, face-to-face communication) than the least-exposed roles. The same junior seat, with its entry bar bumped up a notch. Fortune summed up the phenomenon in a sentence: employers are now asking a 22-year-old for what a 35-year-old would have.

The numbers get more specific. In the most AI-exposed occupations, 52% of the skills newly appearing in junior ads were ones once asked of experienced hires. In the least-exposed occupations, that share was just 7%. As a result, "seniorized" junior postings rose 35% since 2019, while ordinary junior postings fell 10%. Entry-level hiring didn't vanish; the definition of what an entry-level worker must be able to do was pulled upward.

The mechanism is surprisingly simple. Traditionally, newcomers learned the job by doing the simple, repetitive work — a kind of apprenticeship. When AI took exactly that simple work, what remained for newcomers shifted toward judging, generating fresh ideas, and dealing with people. PwC's Pete Brown notes that the change removes the first rung of the ladder that newcomers used to climb to build a career.

And here an uncomfortable question follows. If newcomers no longer grow by working through the simple tasks, where do the seniors of five and ten years from now come from? A decision to narrow the on-ramp for the sake of today's efficiency can, over time, dry up the very supply of skilled people. The seniorization AI creates is both an individual employment problem and an organizational talent-pipeline problem.

When the bar rises, the price of clearing it rises too. The AI-skills wage premium PwC measured shows that repricing directly. The average wage premium for people with AI skills jumped fast: 25% in 2024, 56% in 2025, 62% in 2026. For the same work, the going rate is widening every year between those who can work with AI and those who can't.

PwC calls this a "two-track labour market." On one track are "professionalised" roles, where AI automates the routine and human judgment and expertise are emphasized all the more. Think of a radiologist or a recruiter: the more AI handles the first pass, the more the value of the person making the final call grows. These roles saw jobs grow twice as fast, and wages climb roughly 42% more steeply than other roles. On the other track are "democratised" roles, where AI makes the work doable even by non-specialists. Entry got easier, but differentiation got harder, and they've stagnated relatively.

This polarisation shows up in Korea too. According to a Korea Chamber of Commerce analysis, of roughly 140,000 job ads in the first half of 2025, only 2.6% were for entry-level-only positions, while 97.4% asked for experience. The gap a young person who has never worked can slip through has narrowed that much. The share of so-called "experienced newcomers" — people who pass through a smaller firm to look experienced, then reapply as juniors — rose from 25.7% in 2023 to 28.9% in 2024. The narrowed on-ramp is being filled by people who already have experience.

Of course, it would be hasty to pin all of this on AI. Some economists point out that the decline in junior hiring is tangled up with rate hikes, a correction to pandemic-era over-hiring, and a shakeout in tech. In aggregate unemployment figures, a clear rise among AI-exposed occupations hasn't yet shown up sharply. Change is flat in the middle and moves first only at the edges. But the fact that the edge in question is precisely the newcomer is what makes this data hard to wave away.

That's the picture PwC paints. For a Pebblous reader, there's one more step of translation left. If this is a market where human judgment keeps getting more expensive, then what determines the quality of that judgment becomes the competitive edge. And every judgment feeds on its inputs. The input to judgment is data.

The new work people take on where AI took the simple tasks, PwC's Dan Priest describes as directing AI, doubting it, and applying it to real problems. All three work only on one premise. To direct, you need grounding data to decide what to ask for; to doubt, you need trustworthy reference data to verify the output against; to apply, you need data that carries the context of the actual situation. The human work that got more expensive in the AI era ends up overlapping with the work of discerning and handling good data.

Half of what a seniorized newcomer must bring lives right here. It's no longer the ability to repeat a fixed procedure, but the eye to tell which data to trust and which to doubt. The same question lands at the organizational level. When human judgment becomes the most expensive resource, the side better equipped with the data that keeps that judgment from wobbling is the one that ultimately goes further.

Whether AI destroys jobs may be the wrong question. The sharper one is this: in a market where the definition of capability has been pulled upward, who holds — first, and properly — the data to support that elevated judgment? The direction PwC's billion data points to is clear. The bar has already risen, and what holds up the person working above it is, in the end, the quality of the data.

Primary Reports

• 1.PwC. (2026). 2026 Global AI Jobs Barometer. PricewaterhouseCoopers. pwc.com/gx/en/services/ai/ai-jobs-barometer
• 2.Brynjolfsson, E. et al. (2025). Canaries in the Coal Mine. Stanford Digital Economy Lab.

Korea Data

• 3.Korea Chamber of Commerce and Industry (대한상공회의소). (2025). 2025년 상반기 채용시장 분석 [H1 2025 Hiring Market Analysis]. korcham.net

Industry & Press

• 4.Fortune. (June 18, 2026). Entry-level work didn't disappear — AI 'seniorized' it instead, PwC finds. Fortune.
• 5.CNBC. (August 28, 2025). AI isn't killing jobs, it's killing entry-level careers for young workers. CNBC.