Re:amaze gives you visibility into how your AI agent is responding to customers directly from within a conversation. You can see which knowledge sources were used to generate a response and, if needed, submit a correction to improve the accuracy of future responses. These tools are designed to help you get the most out of your AI agent over time.
Viewing Knowledge Sources used in an AI Response
For any conversation where your AI agent has participated, you can review exactly which knowledge sources it drew from when generating a response. Reviewing an AI response's knowledge sources is a useful way to audit your AI agent's behavior. If a response seems incomplete or off-target, tracing it back to the source can help you identify whether an FAQ article or Response Template needs to be updated with more relevant information.
Finding the sources in a response
Open the conversation and locate the AI response you want to review. You'll see a source count in the bottom-left corner of the response (example: 3 sources). The number will reflect however many sources the agent referenced for that particular reply.
Select the source count to expand the full list of sources. Each source will be labeled with its type, such as Article or Product, so you can see exactly where the AI agent retrieved the information from.
What sources are shown
Depending on which knowledge sources you have enabled for your AI agent, the list may include any of the following:
- FAQ articles
- External URLs
- Response Templates
- Shopify products and orders
- Custom AI context
Improving an AI Agent response
Alongside the source count, you'll also see an Improve Response button. If you find that an AI response is inaccurate or doesn't reflect the best possible answer for the customer, you can use this to provide a better response.
Selecting Improve Response will open a window displaying the original response, along with a free-form field where you can write the corrected version. Once you're satisfied with your correction, select Save Correction. This will be used to further train your AI agent, helping it provide more accurate and helpful responses to customers.