AI engineer was the fastest growing job title in the United States in 2026, with postings up 143 percent year over year, according to LinkedIn's Jobs on the Rise report (2026-01-14). Four of the five fastest growing roles on that list trace back to AI. In the same year, 22 percent of chief human resources officers said at least one business leader in their organization had stopped hiring for entry-level roles specifically because of AI automation, based on Gartner's fourth quarter 2025 survey of 110 HR heads (2026-07-27). Those two numbers describe the same shift from opposite ends. Senior, AI-specific hiring is accelerating while the junior on-ramp narrows.
The barbell
Job postings for AI engineering roles rose 255 percent year over year, the highest growth rate of any tech role tracked, according to DataCamp's State of AI Careers 2026 report, which analyzed 2 million job postings across 85 regions from January 2024 to April 2026 (2026-08-18). That growth is not spread evenly across seniority. It concentrates in roles that require judgment about what a model produced, not roles that produce output from scratch.
Why the junior rung is shrinking, not disappearing
IBM offers the clearest counterexample to a straight cut. The company said it would triple entry-level hiring in the US in 2026, but CHRO Nickle LaMoreaux told Axios that IBM rewrote its entry-level software developer job descriptions so junior hires spend less time on routine coding, now handled by AI, and more time working directly with customers (2026-02-13). IBM did not eliminate the junior tier. It redefined what junior engineers are for.
That redefinition only makes sense next to how much code AI is already writing. At high-growth software companies, 36 percent of code is now written with AI assistance, up from 29 percent six months earlier, according to ICONIQ Capital's Q2 2026 survey of roughly 305 executives at software companies building AI products (2026-07-01). When more than a third of code is machine-assisted, the traditional junior job of writing code under supervision shrinks by definition. What's left is review, customer context, and calls a model can't make on its own.
The people already doing the work are stretched
The load is landing on engineers who already know how to use AI well. Terminal's seventh annual State of Remote Engineering Report, surveying more than 1,800 engineers worldwide, found 70.3 percent now qualify as "AI-enabled or AI-native" under its AI Fluency standard. 77 percent say AI has increased their capacity for higher-value work. But 42 percent also report more burnout than a year earlier, according to the same report (Terminal, via Business Wire, 2026-09-24). A leaner junior bench means more of the judgment work sits on fewer, more senior people, and the data shows they are already feeling it.
Three camps, and founders are picking sides in public
ICONIQ's January 2026 bi-annual State of AI survey found 78 percent of software companies are rethinking workforce planning because of AI. Of those, 45 percent expect a different mix of roles, 33 percent expect smaller teams overall, 42 percent report no significant headcount impact yet, and 15 percent are actively increasing headcount for AI-specific roles (ICONIQ Capital, 2026-01-01). Those categories are not mutually exclusive inside one company, which is part of why the public conversation is getting louder. OpenAI CEO Sam Altman told staff in a live-streamed town hall that the company plans to slow down how quickly it grows because it expects to do more with fewer people, choosing to hire more slowly but keep hiring rather than over-hire and later cut (Benzinga, 2026-01-27). By September the topic had moved from one company's town hall to a main-stage panel. TechCrunch Disrupt 2026 added a session titled "Hiring When AI Is a Co-Founder," featuring Gusto CEO Josh Reeves, Insight Partners SVP Michelle Johnson, and Leland CEO John Koelliker, built around how AI agents are absorbing work that used to go to a startup's first few hires (TechCrunch, 2026-09-16).
We are planning to dramatically slow down how quickly we grow because we think we'll be able to do so much more with fewer people.
What to do this week
Before you open or freeze a junior engineering req, run IBM's test on it. List what the role currently does. Mark what's routine code production a model already handles well. What's left, usually customer context, system judgment, and decisions with real consequences, is the actual job description. Write the req around that, the way IBM did, instead of defaulting to either hiring the same junior role as last year or cutting junior hiring altogether. And if your senior engineers are the ones absorbing the judgment work AI can't do yet, check whether their capacity gain is being offset by the burnout Terminal is already measuring at 42 percent. A team that looks leaner on the org chart is not actually cheaper if it is quietly running on fumes.