Answer: A2. Multiple agents are useful when a task has distinct specialist roles, separate tools or permissions, and clear handoffs that need coordination. A simple support bot can usually remain a single agent. A high volume of identical requests is primarily a scaling concern and does not by itself require a multi-agent design.
Answer: A1. A single agent is usually easier to test, observe, and operate when one set of instructions and tools can handle the task. Access to several tools does not automatically require several agents. Separate roles with different permissions and audit requirements may justify a multi-agent design.