AI agents now beat humans at computer tasks
Five charts to start your day
For $10 a month, or $100 a year, you support a simple mission: spread great data visualisation wherever it comes from. You help fund the work of finding, sourcing and explaining the charts that deserve a wider audience. And you back a publication built on generosity, transparency and the belief that better understanding makes a better world.
CHART 1 • AI agents now beat humans at computer tasks
This is quite incredible progress. Just a year ago, most leading AI agents would struggle with short tasks inside browsers, files and office software. They can now complete many of these jobs autonomously if you let them.
Look at this agentic AI test from OSWorld. This test puts AI agents inside a real computer and then asks them to complete certain tasks. The progress made is quite astounding. For instance, Claude Sonnet 4 delivered 42.2%, which has since been more than doubled by Claude Fable 5, which delivered 85%. For context, 72% is considered the human reference.
Source: a16z
What this means is that companies can hand agents more isolated computer steps without handing over the whole job. They don’t necessarily need agentic AI to carry out the whole task, but they can certainly improve and increase efficiency within the firm when it comes to managing workflow.
However, until AI agents can recover from changes, check linked decisions and verify the final result, the human remains responsible for the work rather than merely supervising a reliable digital colleague.
Paid subscribers get access to the other four charts: AI-written books, smartphone gaming, Apple’s iPhone revenues and Disney’s streaming profits. Together, they show where digital scale is creating lasting behaviour and where the old engine still pays the bills.
