Most issues with the AI Agent fall into a small set of patterns. This article covers the common ones.
The AI Agent is escalating questions it should be able to answer
Most common cause: a knowledge gap. The AI Agent does not have a knowledge item covering the topic, or the item is not activated for this AI Agent.
Fix:
Look at the escalated conversation. What was the guest actually asking?
Check the AI Agent's active knowledge for that topic. Use the filter in the Knowledge toolbar to narrow the list by tag rather than scrolling.
If no item exists, write one. See Writing effective knowledge entries.
If the item exists in another AI Agent but not this one, copy it across: Import in the Knowledge toolbar, then Existing bot. See Importing knowledge into your AI Agent.
Click Activate to apply the change. Knowledge edits are held as drafts until you do.
The AI Agent is giving wrong or outdated answers
Most common cause: a stale knowledge item. The AI Agent confidently quotes whatever you have written.
Fix:
Find the relevant item. Filtering by tag is the quickest way if the knowledge base is large.
Update it.
Test with a sample message in the Preview panel.
Click Activate.
The AI Agent greets the guest again instead of answering
Before you change anything, check whether this is the greeting doing its job. A greeting at the top of the agent's first reply is expected behaviour when a greeting instruction is set on the General tab, and it appears again if the agent re-engages after stepping away from a conversation. See Setting a greeting and a sign-off.
If it is not that, the cause is usually no confident answer to match, so the agent opens rather than answers.
Fix: add a knowledge item that more clearly handles the question the guest is asking. Use the guest's actual phrasing.
The AI Agent does not respond to a multi-part question
Long messages that contain several distinct questions can cause the AI Agent to skip the conversation entirely.
Fix in the meantime: if you notice a guest asking multiple questions in one message, an operator can intervene. Longer term, this is on the product team's radar.
The AI Agent is responding when it should not
Most common cause: the assignment rule is too broad.
Fix: narrow the rule. Add a channel filter, a time filter, or a workflow type filter to scope the AI Agent to where you actually want it.
Check which rule the agent is connected to first. The Connected rule: summary on the agent's General tab tells you what it is currently picking up. See Setting assignment rules and channel filters.
When to escalate to Bookboost
If the issue is structural or persistent (the same problem after knowledge updates), raise a ticket. See Getting support for the AI Agent.
Need Help?
Please contact us through the 'Talk to Us' option on the left menu in the platform, or through the Bookboost Support email at support@bookboost.io if you have questions or need additional support.