Bloomberg and the Indianapolis Business Journal reported on June 18, 2026 that Gallup analysis of a 23,000-person February survey shows US tech workers who use AI at least monthly have a predicted layoff probability of about 6%, compared with 18% for those who use it less often. The data included 660 respondents who said they were unemployed after their position was eliminated. Gallup ran a statistical model and reported that the AI-use signal held after controlling for age, education, sector, and tenure. Notably, only about 1% of laid-off workers attributed their job loss directly to AI — the most cited causes were restructuring, cost-cutting, and the macro economy.
The number that matters is the residual after controls, not the headline ratio. Plenty of confounders could in theory drive a raw 3× gap — more digitally native workers cluster in roles where AI tooling is permitted, and those roles may also be growing — but Gallup says the relationship survives a multivariate model. That suggests something closer to a real productivity signal employers are reading off individual workers. The mechanism is probably layered: AI users ship more, are easier for managers to keep when budgets tighten, and tend to interview better for the roles that survive a restructure.
Context matters here. Anthropic's Dario Amodei has been arguing publicly since 2024 that AI will displace half of white-collar entry-level jobs within five years; this Gallup data is the first US-wide quantitative signal that adoption-by-individual is now visible in layoff selection, not just headcount totals. It also lands while Black Duck's coding study put AI adoption at 97% of engineering teams but governance at 33%, and while OpenAI keeps reporting Codex weekly-active growth. The labor market is no longer treating AI fluency as a nice-to-have skill.
Takeaway for learners: "using AI at work" is now itself a measurable career safeguard, not a personal preference. The cheapest defensive move you can make this quarter is to push your monthly AI usage from incidental to embedded — pick one workflow you own and route it through a model end to end. The Gallup data does not say which tool to use; it says the people not using any are the ones being let go.