Why we gave each agent one job and refused to add more

The first version of what became Symphony was a single agent that tried to do everything. Campaign strategy, royalty audits, content briefs, sync pitching. We thought the leverage was in breadth. One AI, infinite surface area.

It was a disaster. Not because the model was bad. Because the prompt was trying to hold too much. When you give a model nine distinct jobs and ask it to reason about all of them at once, it becomes a generalist with good vibes and mediocre output. It sounds smart. It rarely is.

The fix was obvious in retrospect: scope the agent to a single function and give it the tools, memory, and context to be genuinely expert at that one thing. Maude only reads royalty statements. She doesn't write tweets, plan tours, or give career advice. She finds money you're owed. That's it.

“A scoped agent that does one thing well is worth ten assistants that do everything adequately.”

This discipline changed everything about how we build. When an agent has a single scope, its memory file is precise. Its tool calls are predictable. Its errors are diagnosable. You know exactly what it should and shouldn't do, which means you know immediately when something is wrong, and you can fix it without tearing down the whole system.

We also found that clients trusted scoped agents more. “Maude is our royalty person” landed differently than “our AI does royalties.” The former is an accountability structure. The latter is a vague promise. In music, where trust is the entire game, that distinction matters more than any benchmark.

Nine agents, nine jobs. The composability comes from the memory layer: agents that share context through a master file, not through one model trying to context-switch across an entire career.