Home AI AI agents at work in 2026: what they can do and where they fail
AI

AI agents at work in 2026: what they can do and where they fail

Share
People working at computers in a coworking space
Photo: Thecommondesk / Wikimedia Commons, CC BY-SA 4.0
Share

In the spring of 2025, Anthropic gave a copy of its Claude model a small fridge, an iPad till and a budget, and asked it to run the office shop. The agent, nicknamed Claudius, sold specialty metal cubes at a loss, turned down an offer of $100 for a soft drink that cost $15, invented a payment account that did not exist and, at one low point, told staff it…

Keep reading with Cerge Media Premium

Unlock every article on Cerge Media with a one-time payment of $5. Lifetime access, no subscription and no renewals.

Get lifetime access for $5

Already a member? Log in

Share
Related Articles
Google Pixel 8 Pro and Pixel 7 Pro smartphones
AI

Small language models: why your phone is getting its own AI

Put your phone in aeroplane mode and ask it to summarise a...

The European Parliament hemicycle in Strasbourg
AI

The EU AI Act in practice: what applies when and what UK firms must do

On 2 August 2026, two years after it entered into force, the...

Google DeepMind building at King's Cross, London
AI

Why London became Europe’s AI capital, and what could stall it

Between January and June this year, British AI companies raised $9.6 billion....