- An analysis of 100,000 real conversations on Claude.ai shows that the average time to complete a task without AI is 90 minutes, and AI helps reduce that time by 80%.
- When converted using ONET and BLS wage data, an average task is valued at $55 USD. Claude is often used for complex tasks that typically take 1.4 hours to complete.
- Time savings vary sharply across occupations: legal and management tasks take nearly 2 hours, while food preparation takes only 30 minutes. Medical assistance saves 90% of time, but hardware troubleshooting saves only 56%.
- Specific cases: developing curriculum reduced from 4.5 hours to 11 minutes; writing invoices and notes saves 87%; financial analysis saves 80% for a task valued at $31 USD.
- The median time saved across the entire dataset reached 84%, with the majority ranging from 50–95%.
- When simulated for the entire US economy and assuming widespread AI adoption within 10 years, labor productivity could increase by 1.8% annually — double the average growth rate since 2019. This is equivalent to a TFP (Total Factor Productivity) increase of 1.1%/year.
- The largest productivity contributions come from programming (19%), followed by management (6%), marketing (5%), customer service (4%), and high school teachers (3%).
- Some sectors, such as restaurants, field healthcare, construction, and retail, contributed little due to a lack of tasks suitable for generative AI in the sample set.
- Tasks heavily accelerated by AI will cause non-accelerated tasks to become bottlenecks, for example, teachers still having to supervise the class or engineers still having to manually install systems.
- Key limitations: time predictions are still noisy, lack of empirical data, simplified ONET task model, not accounting for corporate restructuring and technological innovation, and data sample only from Claude users.
📌 Generative AI like Claude can reduce the time required for an average task valued at $54–55 USD by 80% and could help increase US labor productivity by 1.8% per year over the next decade. When simulated for the entire US economy, labor productivity could increase by 1.8% annually, equivalent to a TFP increase of 1.1%/year. The largest productivity contributions come from programming (19%), followed by management (6%), marketing (5%), customer service (4%), and high school teachers (3%).
