After seeing a colleague's AI agents turn project data into useful insights, Sarah built three of her own in Claude Code over breakfast. She started with fewer agents than she expected to need, gave each one a single job and wrote down rules, including that nothing goes out without her approval. The article introduces the three (a wellbeing guardian, a blog writer and a per-project assistant), shows what sits under the hood, and explains why handing preparation work to agents leaves more room for judgment and relationships. It is episode 7 of her Hop on series.

Read more on LinkedIn