AI + Careers Briefing — September 23, 2026

Two AI-labor stories landed the same week, from opposite ends of the ladder. New surveys and US labor data say the entry-level rung that used to teach the job is shrinking because AI already does the grunt work. And the top of the pipeline — where the elite researchers building these systems come from — is shifting east. Neither shifts a decision this week. Both argue for the same next move: bet on AI-adjacent roles that scale with adoption, not on the narrowing paths at the very top or the very bottom.

Is AI already eroding the entry-level rung for new grads?

Yes — new UK survey data and US employment analysis both point that way. In Open University's 2026 Business Barometer, which surveyed 1,500 UK business leaders, 51 percent said AI is changing how they hire, and 19 percent said they had already reduced entry-level recruitment — with 42 percent of that group naming AI adoption as the reason, according to CNBC. A separate Klarus survey of 500 UK and Irish mid-market leaders found 45 percent said AI helps junior employees work faster and better — so the disruption cuts both ways.

But several people CNBC interviewed argued the "grunt work" junior staff used to do wasn't just work — it was how they learned the job. "If AI removes the grunt work, then the grunt work is no longer relevant," Tobias Green, founder of London ad agency Catalyst, told CNBC. Lucy Beaumont, global SVP of product at SHL, put it more bluntly: "Those entry-level roles, they're your training ground — that's where you get your training wheels." The pattern shows up in US labor data too. A Stanford Institute for Economic Policy Research review of employment since ChatGPT's November 2022 launch found young workers in AI-exposed occupations, including customer service and software development, have seen employment declines while older workers in the same fields held steady or grew. SIEPR notes that isolating AI's exact share of that shift, versus other factors, remains an open and active area of research.

For new grads, the skill employers pay for now is catching AI's mistakes, not producing first drafts. Cheney Hamilton, a director at research firm Bloor, told CNBC that "AI raises the floor of what a junior can produce, but it doesn't give them the judgement to know when the output is wrong." Build a portfolio that shows judgment and review — internships, class projects, freelance work where you caught an error or improved on an AI draft. Read up on skills-based hiring, since employers increasingly screen for demonstrated capability over titles when the ladder rung isn't there.

Source: CNBC; Stanford SIEPR

Is China now training more top AI researchers than the US, and what does it mean for a US-based job seeker?

By one closely watched measure, yes. The share of the world's top AI researchers who earned their undergraduate degree in China rose from 29% in 2019 to 47% in 2022, according to MacroPolo's Global AI Talent Tracker — a much bigger slice of the elite research pipeline than the same measure showed just three years earlier.

That doesn't mean China is keeping that talent at home. A separate benchmark, Stanford HAI's 2026 AI Index, found the US still houses more AI talent overall than any other country. But the same report flags a slowing pipeline into the US: the number of AI researchers and developers moving to the United States has dropped 89% since 2017. Stanford HAI also notes the US still produces more notable AI models than any other country, at 59 in 2025 against China's 35. Read together, the two data points describe less of a talent-quality gap than a training-origin shift, plus a narrowing inflow of top researchers to the US.

Neither figure means AI jobs are drying up domestically — but if the elite research pipeline is training more people in China while fewer top researchers overall are choosing to relocate to the US, competing head-on for the narrowest pure-research roles gets harder for a US-based candidate every year. The more durable move is lateral: aim at AI-adjacent roles that scale with adoption rather than with the size of the research talent pool — deployment engineering, AI product management, integration work, and governance roles that every company rolling out AI tools needs regardless of where the underlying models were built. Our guide to AI skills for tech jobs in 2026 breaks down which of those adjacent skills are seeing the most hiring demand right now.

Source: MacroPolo; Stanford HAI

What to watch

• Whether the 2026 Business Barometer's "reduced entry-level recruitment" cohort keeps growing when the 2027 edition lands — if the 19-percent share widens, the ladder erosion is not a one-cycle anomaly.
• SIEPR and other US labor economists on isolating AI's specific share of the young-worker employment decline; they've flagged it as "open and active," which means the current picture will sharpen fast.
• Follow-on data from MacroPolo and Stanford HAI on whether the 89-percent drop in top-researcher immigration to the US reverses — it's the shortest lever US policy has for changing the "training origin" trajectory.