AI is reshaping careers.
Here are the numbers.
Data sourced from the Anthropic Economic Index, Bureau of Labor Statistics, Stanford Digital Economy Lab, and our own research. Updated quarterly.
Which roles are most exposed?
Percentage of occupational tasks where AI can meaningfully assist or automate, based on real Claude usage patterns.
The pipeline is drying up
Entry-level job postings indexed to 2019 baseline (100). The traditional path from graduation to employment is narrowing across industries.
Stanford Digital Economy Lab
Nov 2025
A common objection: isn't this just ZIRP reversing? Stanford researchers controlled for interest rates and COVID effects and found the entry-level contraction in knowledge work still tracks AI adoption timelines — not rate hikes. The decline started before rates peaked and hit hardest in roles most exposed to AI.
Read “Canaries in the Coal Mine” →Augmentation leads, automation follows
Most AI usage today enhances human work rather than replacing it — but automated workflows are growing fastest in API traffic.
What AI is actually doing at work
The most common occupational tasks performed with Claude, ranked by share of total conversations.
| Task | Share | Trend |
| Modify software to correct errors | 10.2% | ↑ |
| Write technical documentation | 4.8% | ↑ |
| Analyze data to identify trends | 3.9% | ↑ |
| Draft business correspondence | 3.1% | → |
| Review financial statements | 2.7% | ↑ |
| Create marketing content | 2.4% | → |
| Translate documents | 2.1% | ↓ |
| Prepare legal briefs | 1.9% | ↑ |
The data tells one story.
Our guests tell another.
Hear from the people living these numbers.
Listen to EpisodesAI is redrawing the map.
We'll help you read it.
dear [CC] is a nonpartisan New Work Foundation initiative helping the next generation navigate an AI-transformed economy. Your donation goes toward producing content, building tools, and keeping everything free for the people who need it most.