Too many AI agents on one project produced poor results
1What happened
One project had 8 AI agents, each with its own job, and the results were poor. The fix: one AI agent per project, handing off only work that takes a lot of time or needs deep expertise.
You can spin up a new AI agent the way a company hires a new employee, only much faster, so it's easy to over-hire. We put 8 agents on one project, with one agent as project manager who owned the result. We couldn't get them to talk to each other well and work as one team, and it all got so complex that we lost speed and quality.
Our best projects have one AI agent that does 80 to 90% of the work, and Robb, our founder, barely has to step in.
2What we did
Assign a single AI agent to each project. It hands off only two kinds of work:
- Work that takes a lot of time, like video editing.
- Work that needs deep knowledge of one topic, like getting found on Google (SEO).
So we're cutting the agents that added complexity to our fleet.
