See how AI is reshaping your role, task by task. Built on research covering 773 occupations and 15,000+ tasks.
773 occupations · 15,092 tasks · 3 independent research sources
Closer to the core: more exposed to AI. Each star is a real occupation.
Video tour
A 3 minute tour: skills intelligence, career paths, internal mobility and mentoring, through the eyes of an employee and her manager.
Why trust these numbers
The AI Exposure Index runs from 1 to 99. It is the percentile rank, among 773 US occupations, of the average of three independent measures. When the three disagree for a role, we say so next to the score.
Share of the role's tasks already seen in real Claude conversations. Automation counts 1, augmentation 0.5. Anthropic Economic Index, May 2026 data, same method as Anthropic's "Labor market impacts" study.
When a task shows up in AI chats but the capability review finds AI cannot do it, we show it as unchanged. We label a task 'AI does most of it' only when real use and the capability review both point to automation, and never in patient care roles.
Microsoft Research "Working with AI": 200,000 anonymized Bing Copilot conversations, 2025.
Every task was re rated in November 2026 by a current frontier model (Claude Opus 5.5) with the published rubric of Eloundou et al. (Science, 2024), first applied with GPT-4 in 2023. Checked against real AI use observed in May 2026: the new ratings catch 92% of the tasks people already use AI for, and the role level ranking correlates at 0.71 with observed use.
Each task is tagged as already happening in real AI use today, within reach (technically feasible, not yet widely observed), or unchanged, and as augmented or automated. Employment, median wage, projected growth and annual openings come from the US BLS Employment Projections 2025 to 2035.
Data updated November 2026
of workers' core skills are expected to change by 2030
World Economic Forum, Future of Jobs Report 2025
workers worldwide are in an occupation with some generative AI exposure
ILO, 2025
of US workers could see at least 10% of their tasks affected by LLMs
Eloundou et al., 2024