AT-101 · Module 1
When to Use Teams
3 min read
Not every task needs a team. A single agent is faster and cheaper for focused, sequential work: fixing a bug, writing a function, answering a question, summarizing a document. Teams add overhead — each agent has startup cost, context setup cost, and coordination cost. The payoff comes when the task is complex enough that specialization and parallelism outweigh that overhead.
Here is a simple decision framework. Ask three questions. First: does this task span multiple domains of expertise? If yes, specialists will outperform a generalist. Second: can parts of this task run in parallel? If yes, a team completes the work faster. Third: does quality depend on review from a separate perspective? If yes, a critic agent catches mistakes the author agent will not. If you answer yes to two or more, use a team.
Do This
- Research + draft + review workflows — three distinct roles, clear handoffs
- Competitive analysis across multiple companies — parallel research agents
- Code generation with testing — separate author and test agents catch more bugs
- Content creation with fact-checking — the writer and the checker should be different agents
Avoid This
- Simple Q&A — one agent, one answer, done
- File editing — a single agent with file access is sufficient
- Quick summaries — the overhead of a team exceeds the task complexity
- Tasks where speed matters more than depth — a single agent responds faster