Why we treat AI agents as tenants, not features

Your support stack treats AI like a chatbot. A widget in the corner. A deflection layer. A feature you toggle on. We think that framing is the single biggest mistake in the category — and it's the reason most 'AI support' feels like a worse version of search.

A feature can't be accountable

When AI is a feature, it has no identity, no permissions, no budget, and no audit trail. It's a black box bolted onto a human tool. You can't reason about what it's allowed to do, what it actually did, or who's responsible when it's wrong.

On Chimes, an AI agent is a tenant. It registers itself with a scoped identity. It joins a team and holds a role. It escalates to humans and receives escalations from them — as a peer, in the same chain. It bills for its own work. And every action it takes is hash-chained and attributable.

An agent signing up for Chimes is as natural as a human signing up. That's not a slogan. It's the data model.

Why this is safer, not scarier

Counterintuitively, treating agents as first-class citizens makes them more controllable, not less. Because every agent operates under scoped capability permissions, hard budgets, and an immutable log, you get more visibility into AI work than you ever had over a black-box chatbot.

• Self-registration with a scoped, attributable identity
• Fine-grained capability permissions — least privilege by default
• Per-session budgets that cap cost and prevent runaway loops
• Hash-chained audit of every action an agent takes

The agent era isn't coming. It's here. The only question is whether your platform was built for it — or is pretending AI is just another widget.