RAG stands for Retrieval-Augmented Generation, a technology that allows AI to answer questions by retrieving relevant information from specific databases, which allows the AI to provide richer and more sophisticated answers.
Currently, ALF can generate answers based on Channel Talk's document management feature, Documents. In the future, it will be able to generate answers based on various databases such as PDFs and user chats. (→ Learn more about Documents)
Train ALF on your documents and control what information to give them
Wouldn't it be great to be able to control how your AI agent ALF responds to customers? The channel team wanted to be able to control this. Not only can we base ALF's answers on documents, but we can also control what the customer sees by toggling ALF answers on and off per article within a document.
Customers can see the source of ALF answers and view the cited article directly.
ALF provides not only the 'source' but also the 'article link' when answering based on a document. Customers can view the entire article, not just the cited part, so they can answer their questions directly through the article.
Managers can check the cited article in the ALF answer and modify it immediately.
If they find that the article cited in the ALF answer needs to be updated, they can modify it in real time and update the information in the user chat screen without having to go to the document menu.
Since RAG allows ALF to answer based on the contents of registered articles in the document, you need to register articles first.
Activate ALF
ALF can only be activated at the final stage of the support bot, so you need to be on a paid plan and use the support bot to enable this setting.
You can configure ALF usage for each final stage within the support bot.
Set ALF Permissions on Articles
To allow ALF to reference an article for its answers, turn on 'Use ALF' at the top right of the article editing screen.
Both the space to which the article belongs and the article itself must be public to use this feature.
Enable Answers from Documents
Go to [Channel Settings] - [AI Agent] and turn on 'Answers from Documents.'
Can ALF read images within a document and refer to them in a response?
Yes, ALF can recognize images posted in document articles and respond to customers based on the image content. The image file extensions that ALF can respond to are png, jpg, webp, gif (non-animated).
However, if an image is only attached without any related explanation, ALF cannot refer to it and provide an answer.
For ALF to successfully refer to an image and answer, the image and its related explanation must be included together in the article. Please also include a text description that allows ALF to understand what situation and context the image is being used in.
For example,
If a document article includes an image and text description about how to invite a team member to a channel,
When a customer asks how to invite a team member, ALF can read the image and guide the customer through the steps.