Anthropic Launches Economic Index to Track AI’s Effects on Labor Markets
Using AI to bolster economic progress has been a topic of conversation for the past few decades. We are only starting to see the potential of massive efficiency and augmentation improvements across all industries with new widely accessible tools like Claude ai. Anthropic is analyzing millions of anonymized conversations on Claude.ai, revealing the clearest picture yet of how AI is being incorporated into real-world tasks across the modern economy. This a unique approach because never before have we been able to analyze such a large dataset of human-AI conversations across a wide spectrum of working environments.
Usage of Anthropic’s models are currently concentrated in the software development and technical writing tasks. AI is leaning more towards augmentation (57%) compared to automation (43%) of tasks.
AI use is most prevalent for tasks associated with mid-to-high wage occupations like computer programmers and data scientists, but is lower for both lowest and highest-paid roles. The research study was done in partnership with Clio to collect and anonymize the data.
What makes this dataset really interesting is that it’s not based on survey data, it is direct data on how Anthropic models are actually being used.
One thing that Anthropic found was that sometimes it makes sense to focus on occupational tasks rather than the occupation itself, for example, a designer, a radiologist, a photographer, or a security screener may all perform similar tasks.
The vast outlier of the job roles was ‘Computer and mathematical’, which only takes up 3.4% of American jobs, but 37.2% of the use. This makes sense however, as these job types are already familiar with coding and tools similar to AI, and perhaps are more aware of the use cases.
The Dataset
The data set is completely open source on Hugging Face, and contains the SOC structure dataset which is the Standard Occupational Classification dataset which lays out the different job roles from the U.S. Department of labor, along with job wages, automation vs augmentation data, and metadata on conversation type.
Implications
AI is augmenting and automating jobs at an increasing pace, however, we are not yet at the point of complete job displacement. One study in 2023 posited that half of economic tasks could be done with LLM-powered software.
This dataset, although imperfect, will be able to track some higher level trends on how AI is affecting work occupations and tasks. Anthropic plans to run follow-ups every six months to track changes in AI use over time.
Learning and soft skill development will be at a premium for all levels of employment. Deeply understanding the problem you are trying to solve will become increasingly valuable. As AI starts to automate the more bespoke tasks, dealing with clients and co-workers effectively, along with the ability to learn to leverage the tools will be paramount in benefitting the most from these emerging technologies.
Next Steps
The index currently only looks at free and pro Claude conversations, no API calls, it also cannot track where the content from the conversation is going outside of Claude. What i’m interested in seeing in the future is how specific API’s are being used, for example, Cursor.ai, a coding platform that leverages AI is one of the fastest companies to reach 100M Annual Recurring Revenue (ARR) and is a highly Claude platform.
Conclusion
As AI continues to reshape the economic landscape, Anthropic’s Economic Index provides one of the clearest views into the evolving relationship between humans and AI. The research suggest that AI will continue to be more of an augmenting technology, rather than by automation. Anthropic is positioned to offer insights into AI productivity and the shifting economic landscape with deeper clarity than we have seen before.
Thank you for reading this piece. If you enjoyed, please follow me on LinkedIn. Or subscribe to my newsletter here.
