Platform
What an agent is made of
Four parts: how an agent is built, how its behaviour is checked, where its knowledge comes from, and how people work alongside it. Every page says what the thing is before it says why it matters.
The core of the platform
Agents
The core of the platform. An agent reads a request, checks it against your rules, acts in one of your systems, and reports what it did.
- Overview
Start here for agents.
Agentic tools
The specific actions an agent may take in your systems, granted one at a time rather than as blanket access.
Voice AI
Agents that handle spoken conversations, with the same boundary and the same escalation path as everything else.
Workflow logic
The written rules that decide what an agent does in each situation, including the situations nobody thought of.
Evaluating agent behaviour
Governance
How agent behaviour gets evaluated: what it is measured against, how it is attacked on purpose, and what the record shows.
- Overview
Start here for governance.
Boundaries
The written statement of what an agent may decide alone, what it prepares for approval, and what it never touches.
Adversarial testing
Deliberately trying to make an agent behave badly, before a customer manages it by accident.
Audits and logs
The record of every action an agent took, why it took it, and what came back. Enough to reconstruct any decision.
What agents learn from and generate
Context
Where an agent gets its knowledge: your documentation, your records, and the data its own work produces.
How people work with agents
Interfaces
How people work with an agent: approving what it prepared, taking over what it escalated, and changing what it does.
Seeing it beats reading about it
Half an hour, no slide deck: an agent running against a process of the same shape as yours, with the boundary visible.