Research Library
The evidence behind Life in the AI Age, translated into plain language without hiding the uncertainty.
Labour-market evidence
Employment, hiring, earnings, hours, job postings, and career-entry patterns.
Task and usage evidence
What AI systems are used for, which tasks they touch, and where exposure is mistaken for elimination.
Productivity evidence
Controlled experiments and field studies—including results where AI helps, does little, or makes experts slower.
AI and Jobs — Evidence Baseline, 2023–2026
This public summary brings together research about occupational exposure, observed AI use, task productivity, organisational response, and labour-market outcomes. These are different questions. A study can answer one without answering the others.
Current scope: the evidence supports meaningful task automation, pressure on some entry-level and freelance work, and substantial workflow reorganisation. It does not establish economy-wide occupational disappearance.
What the evidence can show
- AI may affect particular tasks without replacing the whole occupation.
- Observed employment pressure is concentrated in some highly exposed entry-level and modular work.
- AI adoption can reorganise tasks even when short-term earnings and hours show little aggregate change.
- Productivity effects vary by task, worker experience, context, review burden, and workflow maturity.
What it does not show by itself
- An exposure percentage is not a jobs-lost forecast.
- A productivity gain does not automatically mean fewer workers.
- Observed use on one platform is not economy-wide adoption.
- A labour-market correlation does not automatically prove that AI caused the outcome.
Research status: living baseline; claims and source versions require periodic review. Last reviewed 2026-08-03.
Primary sources and further reading
- Eloundou et al., “GPTs are GPTs” — potential task exposure.
- Brynjolfsson, Chandar, and Chen, “Canaries in the Coal Mine?” — observed employment patterns.
- Humlum and Vestergaard, “Still Waters, Rapid Currents” — adoption and short-term labour outcomes.
- Handa et al., “Which Economic Tasks Are Performed with AI?” — observed task use.
- International Labour Organization, “Generative AI and Jobs” — global occupational exposure.
This page is a public summary, not a substitute for the underlying papers. It does not expose private project files, unpublished notes, or internal research paths.
Library status
The research library is active. Public summaries will be added as evidence packages are verified, with source links, limitations, and review dates recorded alongside each topic.