AI Agents: Prioritize Knowledge Sources by Topic or Keyword
F
Farah Hicham Hammoud
Business Problem:
Right now, users lack direct control over which Knowledge Source an AI Agent should prioritize when answering contacts. The AI Agent searches across all available Knowledge Sources and picks what seems most relevant. This can cause problems when multiple sources contain similar information.
For example, if a workspace has both a current Pricing Guide and an old marketing PDF that mentions prices, the AI Agent may use the wrong source. This can lead to incorrect or outdated answers.
There is no direct setting for admins to tell the AI Agent which Knowledge Source should be preferred for a certain topic, keyword, or type of question.
Desired Outcome:
Let admins influence which Knowledge Source the AI Agent should check first when answering contacts.
Admins should be able to prioritize Knowledge Sources based on signals such as:
- Topic of the question, such as billing, pricing, refunds, or API support
- Keywords in the contact’s message, such as “refund” or “pricing”
- Source priority when multiple sources contain similar information
If no priority rule or matching signal applies, the AI Agent should still search for the answer across all Knowledge Sources.
Use Cases:
- A customer asks about prices. The AI Agent uses the current Pricing Guide instead of an old marketing PDF.
- A customer asks about refunds. The AI Agent checks the Refund Policy before general help articles.
Current Workaround:
Update the AI Agent Instructions to search for specific keywords when handling specific types of questions. Example: “For pricing questions, search for the keyword ‘Pricing Options 2026’."
This may work for one or two topics, but it is not manageable when there are many topics.