Background
During interactions between the AI and customers, you can view conversation records in the Inbox.
When you find that an AI response does not meet expectations, you can tag specific messages and make targeted improvements. This helps continuously optimize AI performance and improve overall operational efficiency.
Operation Steps
1. View Chat Records
- Click Inbox → Inbox to enter the page, where you can view conversation records between the AI and customers across different channels.
2.Add New Tagging Fields
- When you identify an unsatisfactory AI response, go to Open Platform → 3Chat Domain Configuration → 3Chat Messages to create new tagging fields.
-
Click Add, and select a Multiple Choice field type.
-
Note: You must choose a Multiple Choice field. Otherwise, it cannot be used for tagging.
-
You can enter a specific field name and add options. By editing the option names, you can describe the specific issues in the AI response. For example:
-
Field name: Response Style
-
Option 1: Overly exaggerated tone
-
Option 2: Response is too verbose
-
Option 3: Response is overly structured
- Click Save. The newly created multiple-choice field will appear at the end of the field list.
3.Tag Conversations in the Inbox
- After creating the tagging fields, go to the Inbox conversation interface.
Hover your mouse over the target message, and a tagging entry will appear in the bottom-right corner of the message.
- Click it to open the tagging pop-up. Select the field and the corresponding option to describe the issue with the response, for example:
Response Style – Response is too verbose.
- Click Confirm, and the message will be tagged, making it easier to review and optimize later.
- If you want to remove a tag from a message, hover over the tag and click the delete button.
4. Search Tagged Messages
-
After tagging a series of messages, you can analyze messages with the same tags in bulk to precisely optimize AI responses.
-
Click Inbox → Message History to enter the message list.
- Click Filter, select the field name, relation, and option, then click Apply to filter conversations with the same tag.
- After filtering, click Details to view the context of the tagged messages.
- If you remember specific keywords from the conversation, you can select Message Content → Contains → Keyword in the filter, then click Apply to quickly locate the relevant message.
5. Message Optimization
- For unsatisfactory responses, you can directly optimize the AI’s reply.
Go to the Inbox, hover over the message, and click to enter the optimization entry.
- We provide the following optimization options:
5.1Add FAQs
- If you want to directly optimize the response content, click Manually Add Reply, then enter the user question and the expected response.
- After clicking Add, the FAQ will be added to the knowledge base. The next time the AI encounters the same question, it will reference the knowledge base response.
- If you want the AI to strictly output a predefined response, click Fixed Script to jump to the fixed-script Q&A creation page.
5.2Optimize Knowledge Base
- If the AI response was generated from the knowledge base, you can view the specific knowledge snippets used under Optimize Knowledge Base.
Click View Details to jump to the corresponding knowledge file, where you can add, edit, or delete content.
5.3Optimize AI Tasks
- If the AI response was generated using an AI Task, you can view the specific trigger conditions and execution actions under Optimize AI Tasks.
Click Go to Optimize to jump to the task details page, where you can add, edit, or delete tasks.
5.4Optimize Fixed Scripts
- During responses, aside from AI Tasks, the AI either uses knowledge base content or fixed scripts.When a response uses a fixed script, the Optimize Knowledge Base option will be replaced with Optimize Fixed Scripts.
- Here, you can view the specific fixed script used. Click View Details to jump to the script details page, where you can add, edit, or delete scripts.
