Skip to main content

Knowledge Base Overview

What is the Knowledge Base?​

The Knowledge Base is where AI employees access factual information about a business. It ensures that answers are accurate and grounded in real data.

Key Principle

Do not add behavioral instructions to the Knowledge Base. Use capabilities for mandatory behavior instructions and use knowledge base for facts that can optionally be looked up when needed.

Why is the Knowledge Base Important?​

AI employees rely on the Knowledge Base to answer questions consistently and accurately. Without it, responses may be incomplete or outdated.

How the Knowledge Base works​

When a customer asks a question, your AI employee:

  1. Generates a search term based on the customer's message (e.g., "What time do you close on weekends?" becomes "hours weekend")
  2. Searches all assigned knowledge sources to find the most relevant information
  3. Retrieves the top 10 most similar results from your knowledge base
  4. Includes this information in its response to provide accurate, grounded answers

This happens automatically for every customer interaction, ensuring your AI always has access to the right information at the right time.

Technical Note

Behind the scenes, the knowledge base uses vector embeddings and semantic search to find relevant information based on meaning, not just exact word matches. This means the AI can find answers even when customers phrase questions differently than your documentation.

Types of Knowledge Sources​

You can add or update knowledge in several ways:

Business Profile​

  • Available for all accounts in Business App.
  • Automatically added to AI employees.
  • Shares the following fields with AI:
    • Business name
    • Address and/or service areas
    • Website
    • Email
    • Phone number
    • Categories
    • Hours
    • Services offered
    • Short description
    • Long description
    • Social media URLs
    • Booking URL
  • Updated by editing the Business Profile.

Text​

  • Add FAQs or custom information manually.
  • Best for short, business-specific details.

Website​

  • Crawl the business website to retrieve content.
  • Ideal for keeping product, service, or policy information up to date.
  • Can be refreshed manually or on an automatic schedule.
  • See Website Source for details on keeping website knowledge current.

File Upload​

  • Upload documents (PDFs, manuals, policies).
  • Allows the AI to reference structured content not published on the website.

Supported file types:

  • Documents: PDF, PowerPoint, Word docs, text files, Markdown
  • Structured data: CSV, Excel spreadsheets (especially useful for pricing, services, specifications)
  • Code/Data files: JSON, HTML (great for structured data)
  • Images: Common image formats (JPG, PNG, etc.)

Knowledge, Capabilities, and Tools: What's the Difference?​

Knowledge base vs. capabilities vs. tools​

The knowledge base is one of three key components that make AI employees effective:

  • Knowledge Base provides factual information the AI can reference (what you're reading about now)
  • Capabilities define how the AI behaves and when to take actions
  • Tools enable the AI to perform actions in external systems via APIs
Understanding the Difference

The knowledge base contains information to reference (reactive, fact-based). In contrast, capabilities contain instructions on behavior (proactive, process-based), and tools provide the ability to take actions in other systems (integration-based).

What it's for:

  • Business hours, location, contact information
  • Pricing, product specifications, service details
  • FAQs, policies, procedures
  • Any factual information customers might ask about

Capabilities​

Instructions that shape behavior

Capabilities tell your AI employee how to act and when to take specific actions in conversations. Think of these as process instructions you'd give to a human employee.

Example scenario: Booking appointments

  • Your capability says: "Whenever someone asks a question, ask if they'd like to book an appointment"
  • This makes your AI proactive, not just reactive
  • The AI now consistently offers to book appointments, even when not explicitly asked

What it's for:

  • Conversation flows and processes (when to capture leads, how to qualify customers)
  • Tone and communication style
  • Conditional logic (if customer asks X, do Y)
  • Business rules and priorities
Knowledge vs. Capabilities
  • Knowledge = "Here's the booking link" (reactive, information-based)
  • Capability = "After answering questions, always offer to book an appointment" (proactive, behavior-based)

Best practices for Knowledge Base optimization​

Start with What You Already Have​

Begin with existing resources rather than creating new content from scratch:

  • Price lists or service catalogs (CSV or spreadsheet format works great)
  • Product data sheets (PDFs)
  • Your existing website content
  • Training manuals or employee handbooks
  • FAQ documents you've already created

Example: A pricing spreadsheet with columns for Service, Description, and Price is better than a paragraph describing all your services.

Less is Often More​

While you can add lots of knowledge, quality matters more than quantity:

  • Be selective: Only add information that's actually relevant to customer questions
  • Avoid redundancy: If the AI already knows general information, focus on what's specific to your business
  • Remove outdated content: Less accurate information is worse than no information

Example: If you have 18 pages of blog articles about general topics, they might create noise. Focus on specific, business-relevant content instead.

Knowledge is for Reference, Not Instruction​

Don't try to control AI behavior through the knowledge base:

  • ❌ Wrong: Adding "Always ask for email before phone number" to knowledge base
  • ✅ Right: Put that instruction in a capability instead
  • ❌ Wrong: Knowledge base entry saying "If customer asks about pricing, do X, Y, Z"
  • ✅ Right: Add pricing information to knowledge; put the "do X, Y, Z" process in a capability

Keep knowledge relatively static​

The knowledge base works best for information that doesn't change frequently:

  • ✅ Good: Service descriptions, product specifications, business policies
  • ✅ Good: Pricing (if it changes monthly or less frequently)
  • ❌ Poor: Daily specials or promotions (update via capabilities instead)
  • ❌ Poor: Live inventory (use a tool/API for real-time data)

Use website auto-refresh for dynamic content​

If you do have website content that changes regularly:

  • Enable the "Automatic refresh" option when adding website knowledge sources
  • The system will check monthly for updates automatically
  • This is great for blogs, menus, or service pages that evolve over time

Exclude irrelevant website sections​

When retrieving website content, use path exclusions to skip:

  • Blog archives that are more for SEO than customer service
  • Legal pages that aren't commonly asked about
  • Admin or login areas
  • Marketing pages not relevant to customer questions

Example: Exclude /blog/articles/ if those posts are general content not specific to your business.

Review what the AI actually uses​

Use the conversation explanation feature to see:

  • Which knowledge sources the AI referenced in responses
  • Whether it found the right information
  • If any knowledge sources are causing confusion

This helps you refine what to keep, update, or remove from your knowledge base.

Frequently asked questions (FAQs)​

How do I update the Business Profile knowledge source?

Edit the Business Profile in Business App. Changes automatically sync to AI employees.

When should I use text vs. website knowledge?

Use text for short, specific information not on the site. Use website for broader content that is already published online.

Does uploading a file automatically update if the file changes?

No. You'll need to upload a new version of the file to refresh its content.