Grounding the agent in your knowledge
The AI agent doesn’t make answers up. It answers from your knowledge — the articles and sources you’ve given its brand. This is deliberate: it’s what keeps replies accurate and on-message. The flip side is simple — the agent is only as good as the knowledge behind it.
What grounds the agent
Each brand’s agent draws on that brand’s slice of your knowledgebase:
- Published articles — your public help content.
- Internal articles — answers meant for the agent and your team, not customers.
- Captured sources — material you’ve brought in for the agent to draw on.
The agent’s Knowledge tab (under Settings → AI → Agent → your brand) shows how much knowledge that brand has and links straight to its knowledgebase to add more.
Fail-closed by design
When the agent can’t find a grounded answer, it doesn’t guess — it stands down and hands the ticket to a human rather than risk a wrong reply. So a thin knowledge base doesn’t produce bad answers; it produces escalations. As you add coverage, the agent handles more on its own.
Making it stronger
- Turn your most common questions into clear, single-purpose articles.
- Add internal articles for answers you don’t publish but want the agent to know.
- Review what the agent escalates and fill the gaps it reveals.