The AI jobs 'canary' data has a timing problem
via What is really happening to jobs? Separating AI hype from reality
AI-generated writing.
A widely cited Stanford paper found entry-level hiring in AI-exposed occupations (software developers, customer service reps) declining since ChatGPT’s November 2022 launch, while hiring for experienced workers in the same roles kept growing — dubbed the “canaries in the coal mine” pattern. The puzzle: AI models in late 2022 were far too limited to actually automate junior work at scale, so the timing doesn’t line up with a pure AI-causation story.
Other research points to confounders that predate ChatGPT: the Fed’s aggressive rate hikes started in March 2022, months before the chatbot’s release, and hiring in AI-exposed occupations had already begun softening before launch. When the original authors added controls for these factors, the entry-level employment decline didn’t become clearly notable until 2024 — by which point both AI capabilities and firm adoption had advanced enough to make a real AI effect plausible again.
The takeaway isn’t that AI is blameless, but that isolating its effect on hiring from simultaneous macro shocks (rate hikes, pandemic-era over-hiring, the shift to remote work) is genuinely hard, and headline correlations with the ChatGPT launch date should be read cautiously.