AI Agents
Systems that perceive, decide and act — with a human in the loop.
An agent is not a chat window. It is a loop: it takes in the world, remembers what matters, reasons about what to do, reaches into real tools, and comes back with the result. We have been building loops like that for a long time — they used to be called game AI.
How the technology works
Simple enough for a boardroom. Deep enough for an engineer.
Every intelligent thing we build runs the same loop. Walk around it — and go deeper when you want to know how.
The agent takes in what is happening — a message, a document, a signal, a scene.
Enterprise AI
Agents that do real business work, inside real systems.
Automation is a real capability here — agents that triage, draft, schedule, reconcile and route across the systems a company already runs. It is not the whole story. The larger idea is rethinking how people interact with intelligent technology at all; the business work is where that starts paying for itself.
Agents and characters
An agent gets things done. A character is someone you meet.
They overlap — a character can act, an agent can have a voice — but they are not the same idea. One is measured by what it accomplishes. The other by whether you'd come back.
The agent spectrum
- Tool
- Worker
- Character
- Collaborator
- Autonomous system
- reason
- make decisions
- use tools
- execute tasks
- adapt to context
- hold memory
- communicate
- accomplish goals