There is a persistent story that AI adoption means handing work over to a machine and walking away. The data says otherwise.
Anthropic publishes an open dataset called the Anthropic Economic Index, built from anonymized, aggregated Claude conversations matched to occupational tasks. In the latest release, 51% of Claude usage worldwide is augmentation, meaning a person stays actively involved while AI helps with the task. Automation, where AI is directed to complete the task on its own, accounts for the other 49%.
In other words, the most common way the world works with AI is side by side with it. That matches exactly how we build AI products at Rapptr Labs, so we charted the highlights on a new data page: How the World Uses AI.
What does the data actually show?
A few things stood out to us.
- The augmentation split. 51% of conversations keep the human in the loop. The world is not handing work to AI. It is working with AI.
- AI use is broad, not just technical. Software tasks lead at 23.8% of usage, but arts and media (13.6%), education (12.8%), and sales (9.1%) follow close behind. AI is showing up in every kind of work.
- The time compression is real. Tasks that a classifier estimates would take a person roughly 5 hours alone ran to roughly 40 minutes of conversation with Claude.
- Adoption is global. 121 countries have published usage data. Australia leads relative to population, and the United States ranks 12th.
One honest caveat: this data describes observed Claude usage matched to occupational tasks. It says nothing about jobs being replaced, and Anthropic is explicit about that. We like it because it shows how people choose to work with AI when nobody is forcing a workflow on them.
Why does this matter for your product?
Because the biggest AI product mistake we see is designing for full automation when users want a collaborator.
When we integrate AI into a product at Rapptr Labs, the first question is not "what can we automate?" It is "where does the user want control, and where do they want leverage?" The Economic Index suggests most people answer that question the same way: give me leverage, keep me in the loop.
Practically, that shapes decisions like these.
- Show your work. AI features earn trust when users can see and edit what the AI produced.
- Make reversal cheap. A human in the loop needs an easy way to say no.
- Automate the toil, not the judgment. The 5-hours-to-40-minutes number comes from AI compressing grunt work, not replacing decisions.
Explore the data yourself
We built an interactive page charting the collaboration split, the top job categories and request topics, and the country rankings. Take a look, and if you are thinking about how AI fits into your product, we would love to talk.
FAQ
What is the Anthropic Economic Index? It's an open dataset Anthropic publishes, built from anonymized and aggregated Claude conversations matched to occupational tasks. It's meant to track how AI is actually being used, not to measure or predict job losses.
What's the difference between "augmentation" and "automation" in AI usage? Augmentation means a person stays actively involved in a task while using AI, reviewing and directing the work as it happens. Automation means the person hands the task off and lets AI complete it with less ongoing involvement. Anthropic's data currently splits usage almost evenly between the two.
Is AI replacing software engineers? The data doesn't support that conclusion, and Anthropic is explicit that these figures describe observed usage, not employment or job displacement. Software development is a large share of how people use Claude, but that reflects engineers using AI as part of their process, not being replaced by it.
How is Rapptr's approach different from just using an AI coding tool? An AI tool alone gives you output. Rapptr pairs Rapptr AI, our internal platform, with real engineers who review every output before it ships. That's the augmentation model in the data, applied deliberately rather than left to chance.
Where does this data come from? Every figure on our AI Usage Index page is sourced directly from the Anthropic Economic Index, published under a CC BY 4.0 license, from a May 2026 snapshot.