The agent gave a wrong answer
The agent gave a wrong answer
Open the reply in the Inbox to see which Knowledge, Skills and tools it used. Fix the source it relied on, for example an out of date Knowledge page, rather than correcting the single reply. See How Conduit agents work.
The agent is using information for the wrong property
The agent is using information for the wrong property
Check that the Knowledge is scoped to the right listing or group, and that the conversation is linked to the right reservation. Knowledge scoped to a group is only used for listings in that group. See Listings and groups.
The agent is not replying at all
The agent is not replying at all
Check the agent is attached to the inbox the conversation is in, that it is within the agent’s working hours, and whether it is on copilot, where replies wait as drafts for your team to approve. A conversation tagged as urgent also prevents the AI from acting. See Agents and Conversation tags.
My changes to the agent are not taking effect
My changes to the agent are not taking effect
Changes to an agent are staged and only go live when you publish them. Try them in the Sandbox first, then publish. See Testing agents.
The agent escalates more than I expected
The agent escalates more than I expected
Look at the escalation reasons on the AI Automation Dashboard. Escalations often point to a missing Knowledge page or a situation with no Skill. Filling those gaps is usually the quickest way to raise your automation rate. Operator can help you find them.
I am not sure whether something belongs in Knowledge, a Skill or Persona
I am not sure whether something belongs in Knowledge, a Skill or Persona
Facts go in Knowledge, processes in Skills, tone in Persona and hard limits in Guardrails. See What goes where.