Skip to main content

Knowledge Base for VoiceAgents

Teach your VoiceAgent to answer custom questions by uploading information to the Knowledge Base.

V
Written by Valeriia Volobrinskaia

In this article, you’ll learn how the Knowledge Base feature allows you to equip your VoiceAgent with accurate, business-specific information to confidently answer customer questions.

User Level:

  • Admin


Knowledge Base for VoiceAgents

The Knowledge Base feature lets you give your VoiceAgent the right information to answer specific customer questions with accuracy and confidence. It's especially useful for handling details that are unique to your business and aren't covered in general documentation or public pages.

Let’s say you run a logistics company. Your VoiceAgent might already handle basic queries like tracking packages. But if someone asks about the specific customs documentation needed for importing to Switzerland, it won’t know unless you’ve uploaded that information. Or maybe you operate a pet clinic chain, and you want the VoiceAgent to list your weekend vaccination packages. That, too, needs to be uploaded in a document.

📖 The Knowledge Base serves as the primary source of information for your VoiceAgent.


How to Upload a Document

Documents can be uploaded to the VoiceAgent Knowledge Base in two different places within the CloudTalk Dashboard.

Uploading document during VoiceAgent creation

You can add documents while creating a new VoiceAgent or when editing an existing one.

  1. Navigate to VoiceAgents → Agents in Cloudtalk Dashboard.

  2. Create a new VoiceAgent or open an existing one for editing.

  3. In the Knowledge step of the VoiceAgent Wizard, click + Add files.

  4. Upload a new document or select an existing the knowledge base to be associated with respective VoiceAgent.

    The uploaded documents will be available for that VoiceAgent to use when answering questions during calls.

Uploading document through Knowledge Base section

  1. Navigate to the VoiceAgents > Knowledge Base section in your CloudTalk Dashboard.

  2. Click Add document to upload the files.

  3. Select a type of upload you would like to perform and follow general instructions for and upload:

    1. File upload

    2. URL upload

    3. Manual upload

  4. Link you document to existing VoiceAgent or finish upload without linking for now.

Once the upload is finished, you can see your newly added Knowledgebase file at the top of the list.

File upload

The Document Upload method allows you to provide the VoiceAgent with specific, high-quality internal files that aren't necessarily published on your website. This is ideal for detailed product manuals, internal SOPs, or complex pricing sheets.

Once a file is uploaded, our system processes the data to make it searchable for the AI in real-time.

  • Unlike the daily sync of URLs, uploaded documents are processed immediately. Once the upload is complete, the VoiceAgent can start referencing the content in the very next call.

  • The system scans the entire document, creating a semantic map of the information. When a caller asks a question, the agent retrieves the most relevant "chunk" of text to formulate its answer.

To ensure a successful upload, keep your files within these limits:

  • Max Size: 21MB per file.

  • File Types: .pdf, .txt, .docx, .html, .epub.

⚠️ For best results, use files that are primarily text-based. The system cannot read images or complex charts within these documents.

URL Upload

Integrating your documentation via an Upload URL is the most efficient way to keep your VoiceAgent updated with your latest website content, FAQs, or public help centers. Instead of manual uploads, the system "crawls" your site to build its intelligence.

The URL upload method functions as an automated web crawler designed to digest structured information. Here is the technical breakdown of the process:

  • Recursive Crawling (Depth): When you provide a root URL, the system doesn't just look at that single page. It parses the content and follows internal links up to 2 levels deep.

    • Example:

      domain.com/help

      domain.com/help/shipping (Level 1)

      domain.com/help/shipping/returns (Level 2).

  • To ensure the VoiceAgent remains focused on conversational data, we extract text content only. All non-textual elements, such as images, videos, and interactive links, are ignored during the ingestion process.

  • Your knowledge base is never static. The system automatically re-syncs with the provided URL once every 24 hours, ensuring the agent’s responses reflect any recent updates to your live documentation.

⚠️ The URL must be publicly accessible. Password-protected pages, internal intranets, or content behind a paywall cannot be reached by our crawler.

Manual Entry

Manual Entry is the simplest and most direct way to add information to AIVA’s knowledge base. It provides a straightforward text editor where you can type or paste specific details you want the agent to know.

Use the text box to write out instructions, facts, or FAQs. Quickly move information from emails, Slack messages, or internal notes directly into the system. Unlike URLs that sync daily, Manual Entry updates are applied the moment you save them.

Best Practices for Uploading Documents

To get the most out of your VoiceAgent’s Knowledge Base, keep the following in mind:

  • Use clear, descriptive filenames. A simple name like returns_policy_ecommerce.pdf helps your VoiceAgent understand and retrieve the file faster, and makes it easier for your team to manage documents. Avoid cryptic or overly complex filenames like final_doc_12233_k9v2025-update_v4_NEW.pdf.

  • Keep content focused and relevant. Upload only the information your VoiceAgent needs to answer questions accurately. Skip lengthy manuals or documents packed with unrelated details. The more concise and specific the file, the better your VoiceAgent will perform.


FAQs

The VoiceAgent gave the caller completely wrong information.

This usually happens when the knowledge base is outdated, unclear, or the AI is allowed to infer beyond the provided content. Keep the knowledge base accurate and instruct the VoiceAgent to answer only from approved information.

The VoiceAgent is making up answers that are not in our FAQ

Add a clear instruction into Goal section telling the AI to only answer from the knowledge base and never guess. Use Answer Questions skill to define, what should happen when information is not part of knowledge base.

The AI keeps saying “I don’t know” to every question

This often means the knowledge base is empty, too short, or does not contain the information callers are asking about. Review the uploaded documents and make sure they cover common inquiries.

The caller asked something not in our FAQ

Configure the fallback behavior in Answer Questions. You can choose whether the VoiceAgent should admit it does not know, guide the conversation elsewhere, or transfer the caller to a human.


Need help or have a question? Just reach out through our Support portal — we’re here for you.

Did this answer your question?