What the studies actually find
What the research actually says about AI and tech jobs
The serious research converges on a nuanced picture, not a slogan: AI is reshaping technical work faster than it is eliminating it on net, the impact falls unevenly (hardest on entry-level, lighter on experienced workers), and the people who adapt their skills capture most of the upside. "AI will take all the jobs" and "nothing will change" are both wrong.
Stanford AI Index 2026 — the uneven impact
The 2026 AI Index reports that employment for software developers aged 22–25 has fallen by nearly a fifth since 2024, as AI handles more of the well-specified, boilerplate work that defined entry-level coding. In the same AI-exposed roles, mid-career and senior workers have largely held steady or grown. Meanwhile coding-benchmark performance (SWE-bench Verified) jumped from around 60% to near 100% in a single year — which is why the routine end of the work is the part under pressure. About a third of surveyed organizations expect AI-related headcount reductions in the year ahead.
WEF Future of Jobs 2025 — churn, not collapse
The WEF projects roughly 170 million new roles created and 92 million displaced by 2030 — a net gain, but, as the report stresses, the displaced workers are not automatically the ones who fill the new roles. Nearly six in ten (59%) of the global workforce is projected to need reskilling by 2030, and workers who can demonstrate AI proficiency command a meaningful pay premium (reported around 56% on average by the PwC 2025 Global AI Jobs Barometer, the figure the WEF cites). The framing is shifting from bolt-on automation to systemic redesign of how work is done.
OECD — exposure is broad, outcomes depend on policy and skills
OECD analysis finds a substantial share of jobs across member economies are exposed to automation, with lower-skilled roles bearing more of the risk. The OECD's consistent message is that exposure is not destiny: outcomes hinge on reskilling, adoption patterns, and policy, not on the technology alone.
Reading it honestly
Two cautions, in both directions. First, these are projections and labor-market estimates; methodologies differ, headline numbers get repeated out of context, and nobody can forecast this precisely — treat the direction as robust and the exact figures as soft. Second, a net-positive market is cold comfort to an individual who doesn't adapt: aggregate optimism and personal risk coexist. The throughline that matters for a working engineer: the research keeps pointing to adaptation — adding AI-native skills to existing expertise — as what separates the people the transition helps from the people it hurts.
- Stanford HAI — 2026 AI Index Report.
- World Economic Forum — Future of Jobs Report 2025.
- PwC — 2025 Global AI Jobs Barometer (the ~56% AI-skills pay premium the WEF cites).
- OECD — Future of Work.
Figures are drawn from third-party summaries of these reports and are directional — read the primary report before quoting any number, and note that definitions of "exposure," "displacement," and "AI role" differ across studies. Corrections: hello@aiarch.dev.
Adaptation is the lever. Here's the path.
AI Architect Academy turns the research's headline advice — add AI-native skills to senior experience — into a concrete, mastery-based path across Anthropic, AWS, and Cloudflare.
Or get notified when new tracks ship.