Skip to main content
AlonChat

Q&A

Add an approved answer and the ways customers ask for it, then train before it can be used

Q&A#

Q&A is an approved answer for a specific customer request. Add the question, the other ways people ask it, and the answer you want used. Q&A is not a general document, a dated promotion, or a communication style.

Q&AText or filesLearned Style
HoldsThe answer you wrote for listed questionsReference material the agent can useHow replies sound
Also includeAlternate questions, and photos or video the reply should showThe document or note itselfNothing factual
Becomes availableAfter you train itAfter you train itAfter you select it in Guidance

Past chats can suggest a Q&A. Importing Messenger or Instagram history does not publish that answer. Review it here, then train it. Style learned from those chats stays in Learned Style.

Add a Q&A#

  1. Open Agent Studio → Data sources.
  2. Choose Add source, then Q&A.
  3. Enter a title, at least one question, and the answer.
  4. Choose Add question for other phrasings of the same request.
  5. Optionally add context. Context is background for the agent, not the reply, and it is not a place for passwords or secrets.
  6. Optionally attach images or video that should appear with this answer.
  7. Choose Add Q&A.
  8. When the row says Ready to train, choose Train agent.
  9. Test the questions you wrote in Playground.

Write the answer as a complete reply. "Check the website" is not enough when the customer asked for the fact itself.

Q&A status#

StatusWhat to do
Ready to trainThe answer is saved and still needs training.
ProcessingThe answer or its media is still being prepared. Wait.
TrainingIt is being added to the agent's knowledge.
TrainedTest the main question and the alternate phrasings.
Needs attentionOpen it, fix the error, and use the offered retry.
RemovedIt is out of active knowledge. Restore it before training if you still want it used.

If a trained Q&A is missed, add the phrasing the customer actually used and train again. Missing answers and conflicts across sources are reviewed in Source Health.