·2025

AI and the Future of Work: A Transition Framework

Internal benchmark analysis; capability assessment against standardised professional tasks.

augmentationdisplacementskills
View original report ↗

60–80%

White-collar tasks where GPT-4 class models perform at/above median human level

Junior colleague

Best mental model for AI: requires oversight, correction, direction

What this report says in 60 seconds

  • GPT-4 class models perform at or above median human level on 60–80% of white-collar tasks
  • Complex, multi-step reasoning with real-world stakes: human superiority remains significant
  • Tasks requiring novel judgment, relationships, and physical presence remain human-dominated
  • AI best understood as a 'junior colleague' requiring oversight, correction, and direction
  • Transition support framework: identify which tasks to delegate to AI vs. retain for humans
  • Workers who treat AI as a collaborator (not a threat) are most productive and most satisfied

What this means for you

Career Navigator

This is me →

The 'junior colleague' framing is the most useful mental model. Your role evolves to manager and director of AI output — which requires your domain expertise, judgment, and communication skills, not coding.

Career Starter

This is me →

Understanding what AI can and cannot do is the foundational skill. Knowing where AI is unreliable in your sector means knowing where you are irreplaceable — and where to position your career.

Career Reinventor

This is me →

The tasks where AI falls short — novel judgment, real-world stakes, relationships, complex multi-step reasoning — are the tasks that 20–30 years of experience builds. This is your differentiated offer.

Roles this research highlights

Community

Discuss what this means for your career

Bring your questions to the community. Members from every background are working through the same research.

Free to join. No credit card required.

Related research