AGI and the Future of Work
How advanced general AI may reshape jobs, wages, and the structure of work — with benefits, risks, and what individuals and organisations can do today.
Executive summary
AGI is likely to automate, augment, and create work simultaneously. The most useful framing is not jobs lost vs jobs created but tasks redistributed. Workers, employers, and policymakers each have meaningful preparation moves available now.
What the evidence shows
Studies from OECD, ILO, and McKinsey (2023–2025) converge on three findings: a large share of tasks across many roles are exposed to current AI; exposure does not equal substitution; and the workers who benefit most are those who use AI as a tool rather than compete with it. The 2025 International AI Safety Report flagged labour-market disruption as a serious near-term concern requiring policy attention.
Plausible benefits
- Higher productivity in knowledge work, scientific research, and software.
- Lower cost of expert services (legal, medical, educational).
- New roles around AI design, oversight, evaluation, and integration.
- Better tools for people with disabilities to enter and stay in work.
Credible risks
- Rapid disruption that outpaces retraining systems.
- Wage compression for tasks easily automated.
- Concentration of value in firms that own frontier systems and the compute to run them.
- Erosion of entry-level roles that traditionally trained junior workers.
What helps
Continuous learning, transferable skills, comfort with AI tools, and union or institutional voice in how systems are deployed. Policy options include portable benefits, training entitlements, and competition policy on compute markets.
Key takeaways
- 01Tasks shift faster than jobs.
- 02AI-augmented workers are outperforming both unaided humans and AI alone.
- 03Entry-level roles need active redesign.
- 04Preparation is possible at individual, firm, and policy levels.
Frequently asked questions
Will AGI cause mass unemployment?
Possible but not inevitable. The pace of capability deployment relative to retraining systems matters more than capability itself.
Which jobs are safest?
Roles combining judgement, physical embodiment, social trust, and contextual knowledge tend to be more resilient — though no role is fully exempt.
Should I learn to use AI tools?
Yes. Across studies, workers fluent with AI tools earn more and adapt faster.