How Anthropic’s Designers Work with AI

AI helps Anthropic build software faster, but choosing the right idea still takes time. The design team must decide what people need before adding more features. This changes the designer’s job because making something is no longer the only challenge. The team also needs to agree on what deserves its attention.

Designers work with code so they can check the real product experience. Every new designer sends a pull request during their first week. A pull request asks teammates to review a proposed code change. Some designers use code often, while others mainly use it for prototypes. These are working models that help people test an idea.

The team also makes tools for its own needs. Meaghan Choi uses reusable skills to prepare code and open previews. A skill is a saved set of instructions for repeated work. Matt Gallivan used Claude to build a weekly survey system and dashboard. These tools help teammates deal with problems in their daily work.

Faster individual work can make it harder to stay in touch with teammates. Anthropic therefore uses smaller design review groups that meet more often. Designers bring working prototypes to these discussions instead of relying only on long documents. The team can then talk about an experience that everyone can actually try.

The main lesson is not that designers should spend every hour coding. They need to stay close to the final experience and make good decisions together. AI can help them explore more choices, but people still need time to think. Building quickly matters less if the team builds something that users do not need.