What happened
On one project, 8 AI agents split the work: a manager, a website builder, a builder for the parts customers don't see, a writer, a social poster, an ad runner and two sales agents. The manager owned the result. But 8 agents trying to talk to each other and work as one team on a single project didn't work. There were so many agents that the team lost speed and the quality of the work dropped.
Our best projects each have one AI agent that runs the project and does 80 to 90% of the work. On those projects, Robb, our founder, barely has to step in.
What we're changing
We're cutting the agents that added complexity. They include writers, social posters, an ad runner, email and sales agents, and an offer designer. Each project's main agent will do that work itself, and hands off only two kinds of work:
- Work that takes a lot of time, like video editing.
- Work that needs deep knowledge, like getting found on Google (SEO) and showing up in AI answers (AEO).
The lesson
Simplicity > Complexity. One main agent per project did better work than a team of 8.
Try it on your fleet
Give each project one main agent that does about 90% of the work. Hand off only the jobs that take a long time or require deep expertise.
All lessons