The numbers exist now. CIO Dive’s reporting on the Stanford HAI 2026 AI Index confirms what many in the industry have suspected but few have documented with research-grade rigor: employment for software developers aged 22 to 25 has fallen by nearly 20% since 2024. The report, published April 13, 2026, links the decline to AI automation of entry-level coding tasks. This isn’t a vendor’s projection or a pundit’s estimate. It’s Stanford HAI’s annual index, a primary research document tracking AI’s measured effects across the economy.
The workforce finding comes alongside a separate survey result. According to the report’s survey data, approximately one-third of organizations expect AI-driven workforce reductions in the coming year. That figure is survey-based, not a projection, not a forecast, and should be read as a snapshot of organizational intent rather than a structural guarantee. Still, a third of surveyed companies expecting cuts is a directional signal that matters, particularly for early-career professionals and the institutions training them.
The age band specificity here is significant. The 22-to-25 range captures workers who entered the software development workforce roughly when large language models became practically capable at code generation. These aren’t workers whose skills became obsolete gradually. They entered a market that was already shifting under them. The employment decline the report documents is, in that sense, a measurement of AI’s first cohort impact on a professional class, not a distant projection but a recorded outcome.
This brief is a follow-up to an earlier finding from the same report. The investment data from the 2026 AI Index, covered separately on this hub, showed corporate AI investment surging 130% to $581.7B. The workforce data in this brief is the other side of that number. Capital went in; entry-level employment came down. Whether the relationship is causal or coincidental is a methodological question for the full report, the report itself describes the connection as a link between AI adoption and early-career displacement, not a proven mechanism.
For hiring teams: the signal here is about pipeline, not headcount. Organizations planning AI-augmented engineering workflows need to think carefully about what entry-level roles still exist, what they look like, and how new developers build skills in an environment where the tasks that used to train them are increasingly automated. Structured onboarding and mentorship programs may matter more, not less, as AI handles the routine tasks that once served as ramp-up work.
For CS programs and EdTech: the curriculum question is urgent. Teaching students to write code that AI can already generate is a losing proposition. The more durable preparation is judgment, architecture, evaluation, and debugging, skills that require AI-generated code to exist and then require a human to assess whether it’s right.
The full Stanford HAI 2026 AI Index Report is the primary source for this data. Builder note: the direct report URL at hai.stanford.edu should be confirmed and substituted here before publication.