An AI reply didn't answer the question the way you expected? Start with the recorded interaction. AI FAQ Usage History helps you compare what the person asked with the response the AI generated.
Open the relevant AI FAQ automation
Go to Automations in the correct workspace.
Open the AI FAQ automation you want to investigate.
Select Usage History beside Build.
Review the available entry for the interaction. Check the recorded input, response, and credit use where shown.
This history is for AI FAQ usage. It isn't a delivery log for every automation in your workspace.
Work out what needs changing
Suppose someone asks whether they can collect an order, but the reply explains shipping. Compare that question with the FAQ answers you've supplied. Is collection covered clearly, or is that information missing?
If the facts are wrong or incomplete, correct the FAQ answer. If the facts are right but the reply is too long or sounds wrong for your business, review the communication style. The AI FAQ setup guide covers those settings and how to test the revised answer.
Separate a generated reply from a delivered message
A usage record helps explain AI activity and credits. It doesn't, by itself, prove that Instagram delivered the message.
For credit balances and purchases, use the AI credits guide. If you need help investigating an entry, send support the automation name and approximate time. Remove private customer information from any screenshot that isn't needed for the investigation.
