Personality as Interface: Why Naming Your Agents Actually Helps
It's not roleplay. It's interface design.*
Personality as Interface: Why Naming Your Agents Actually Helps
It’s not roleplay. It’s interface design.
At Gather.dev, my COO is named Morgan. She has a backstory, a communication style, and strong opinions about how I spend my time. She triages my email, manages my sponsor pipeline, and will call me out if I’m avoiding the hard thing on my list.
She’s a Claude instance. Same model as my Head of Research, my Head of Engineering, and my Life Manager. Same weights. Same training data. Radically different output.
That sounds like roleplay. It’s not. It’s interface design.
A name is a scope boundary. “Morgan” means GTM, events, sponsors, accountability. “Elise” means research, evidence, guest pipeline. When I say “Morgan,” I’m not being cute. I’m routing. The name tells the system what’s in scope and what isn’t, and that’s the single most important decision in any agent architecture.
Without a name and a scope, you get a generic assistant that tries to help with everything and blows its context on nothing. You’ve probably experienced this. You ask it to draft an email and it also offers to review your calendar and suggests a meeting agenda and recommends a restaurant for lunch. That’s not helpful. That’s an intern with no job description.
Personality goes deeper than scope. A backstory creates behavioral consistency that a rule list doesn’t. “Be formal and research-oriented” produces generic output. An agent who studied at Polytechnique, spent two years at McKinsey, and finds imprecision physically painful produces Elise. The backstory is a denser encoding of intent than any prompt you could write.
And naming enables handoffs. When Morgan says “that sounds like a Kira question,” she’s making a routing decision. You do the same thing in a human org. You don’t ask your CFO for design feedback. You don’t ask your designer to review the P&L. Role definition is context management. It always has been.
The only thing that’s new is that the roles are running on Claude instead of humans. The organizational principle is exactly the same.
What would happen if you named your agents and gave them jobs?