Last updated
30 Sept 2024
Tanaka Chamisa
You need to train your Cue AI agents to ensure they provide accurate and relevant responses to customer queries. This guide will walk you through navigating the “knowledge” section of your AI agent settings, including the process of uploading documents, managing them, and training your AI agents effectively.
Allowed document types
To train your AI agent, you need to upload documents that contain the information it will use to respond to queries. For now, Cue supports PDF, DOCX, MD, JSON and PPTX file types. However, the Cue team can cater to other document types upon request.
Make sure your documents are well-organised and contain clear, accurate information to maximise the effectiveness of the AI agent.
Uploading and indexing documents
Uploading and indexing documents in Cue is straightforward and critical for training your AI agent. Here’s how to do it once you’re in the AI section of your Cue workspace:
Navigate to the “AI agents” section and select the agent you want to train.
Click on the “Knowledge” tab within the settings page. This is where you manage all documents associated with your AI agent.
Uploading additional documents
You’re more than welcome to upload additional documents once your AI agent is already created and or published. Here’s how:
Click the “knowledge” tab in your AI agent settings section.
Drag and drop your documents or click on the “upload” text.
Select the documents you want to upload from your computer. You can upload multiple documents at once.
Once uploaded, Cue will automatically begin indexing the documents. Indexing is the process where we analyse the content, making it searchable and usable by the AI agent for responding to customer queries.
Searching and finding documents
After uploading, you can easily search for specific documents using keywords or phrases. The search function in the Knowledge section allows you to quickly locate any document based on its content or title.
Training your AI agent
Once your documents are uploaded and indexed, the next step is to train your AI agent. It’s safe to assume that if an untrained human reads the uploaded documentation and is able to answer customer questions, then the AI agent will be able to do the same.
AI agents can be trained on a variety of documents, including PDF, DOCX, PPTX, MD and JSON.
Training takes approximately 1–3 seconds per single-page document.
Once an AI agent has been trained on the content you've uploaded, you can test it using the live preview on the right of its settings page.
AI agents can only answer questions based on static training information. For now, they can’t perform actions such as update a customer's details or check the status of an order, etc.
In the knowledge section, review the documents you’ve uploaded. Ensure that they cover the necessary topics and information the AI agent needs to answer questions effectively.
You can add, remove, or update documents at any time, allowing you to continuously improve your AI agent’s performance.
In order to test the AI agent, return to the settings page of your AI agent and test its responses using the newly uploaded documents.
Ask the AI agent questions related to the content in the documents to ensure it’s referencing the correct information and providing accurate responses.
If the AI agent’s responses are not as expected, you can refine its training by adjusting the prompt, adding more documents, or removing irrelevant content.
You can use the test feature to preview how the AI agent will interact with customers, allowing you to make any necessary adjustments before going live.
Managing and updating documents
Regularly updating and managing your documents is essential for keeping your AI agent well-informed and effective.
Document deletion: If a document is outdated or no longer relevant, you can easily delete it from the knowledge section by clicking on the trash icon next to each document.
Document updates: When information changes, update the existing documents or upload new ones to ensure your AI agent always has the most current data.
Once you’ve trained your AI agent with the relevant documents, it’s ready to handle customer interactions. Continue to monitor its performance and make adjustments as needed to ensure it remains effective and aligned with your goals.