Agent & knowledge
Configure safe AI behavior and give the Agent approved information.
Choose the right source type, wait for processing, and explicitly control whether the Agent may use it.
Add trusted files, websites, articles, and approved answers, then keep the Agent's sources current and conflict-free.
Define the Agent visitors meet, the company it represents, the audience it helps, and the boundaries it must respect.
The Agent answers from approved knowledge, uses the conversation context, and escalates when information or judgment is missing.
Configure Knowledge and the Agent in private, pass the launch checks, then deliberately make AI available to visitors.
Run realistic questions that test grounding, uncertainty, action confirmation, and handoff—not just the happy path.
Know when a source is ready, how to recover failed processing, and how unanswered questions become a content plan.
Use the readiness message to find and complete the exact requirement preventing public AI activation.