Our go-to AI assistant
Claude
An addition to the mix. We use Claude to think problems through, draft, and build and check code. This website was built with it, and people still make the calls.
Claude at a glance
- What it is
- An AI assistant from Anthropic that you talk to in plain English.
- What we recommend it for
- Thinking problems through, drafting words and plans, and building and checking code.
- How often you will use it
- Every day
- The most popular choice?
- No, but a good fit for the right job
- Do we always recommend it?
- Yes, for every team we work with
Why we recommend it
Claude is an AI assistant you can talk to in everyday language. It can help you think through a problem, write and work with information. Give it useful context and a clear task, and it can become a practical partner in the work rather than just a question-and-answer box.
Notion remains our main context window, and Claude brings additional capabilities when we need them, including coding work.
For much of our development work, we use Claude Code - a way to work with Claude directly on a codebase. It can inspect files, help make changes and run checks, within the access it has been given. We used Claude while building this website, including its layouts and the process that brings content across from Notion.
Claude Code can also use specialist subagents to work on separate parts of a task, and skills - reusable instructions and supporting resources - to give it a more consistent way to handle particular jobs. That can make a capable development workflow easier to organise; it doesn’t remove the need to test the result.
People remain responsible for the decisions, permissions, review and what goes live. Check important answers and be careful about the information you share with any AI tool.
How we use it
- Thinking things through. Talking a messy problem out until the real question appears.
- Drafting. First drafts of words, plans and documentation, which people then shape.
- Building. This website, its Notion sync and its tests were built with Claude.
What we’ve found
- Give it context. The more it knows about your work, the more useful it gets.
- Keep people in charge. The AI helps; people make the calls on what’s right and what goes live.
- Start with one real, annoying task rather than a grand AI strategy.