Back to Help Home
Chatbots & AI

Training AI with your business data

5 minutes read time Difficulty: intermediate

Training AI with your business data

The AI chatbot is only as good as the knowledge you give it. Here's how to train it with your business information.

What is the AI knowledge base?

The knowledge base (called "AI Contexts" in AIsoule) is a collection of documents and information that the AI uses to answer customer questions. Without it, the AI can only give generic responses.

Types of knowledge you can add

1. Website URLs

The AI crawls your website pages and learns from the content.

  1. Go to Chatbot → AI Contexts
  2. Click "New Context"
  3. Select type: Website URL
  4. Enter the URL (e.g., https://yoursite.com/faq)
  5. The AI crawls and indexes the page content

Tips:

  • Add your FAQ page, product pages, pricing page
  • The AI re-crawls periodically to stay updated
  • Add multiple URLs for comprehensive coverage

2. PDF Documents

Upload product manuals, policy documents, or any PDF.

  1. Click "New Context"
  2. Select type: Document
  3. Upload your PDF file
  4. The AI extracts and indexes all text content

Good documents to upload:

  • Product catalogs
  • Return/refund policies
  • Shipping information
  • User manuals
  • Company policies

3. FAQ Pairs

Add specific question-answer pairs for precise responses.

  1. Click "New Context"
  2. Select type: FAQ
  3. Enter:
    • Question: "What are your business hours?"
    • Answer: "We're open Monday to Saturday, 9 AM to 6 PM IST."
  4. Add as many pairs as needed

Tips:

  • Use the exact questions customers ask (check your chat history)
  • Keep answers concise (2-3 sentences)
  • Add variations of the same question

4. Custom Text

Paste any text content directly.

  1. Click "New Context"
  2. Select type: Text
  3. Paste your content (product descriptions, policies, etc.)

How the AI uses knowledge

When a customer asks a question:

  1. AI searches your knowledge base for relevant information
  2. Finds the most relevant context
  3. Generates a natural response based on that context
  4. If no relevant context found → either gives a generic response or transfers to agent

Best practices for training

Start with your top 50 questions

  1. Review your last 100 support conversations
  2. Identify the 50 most common questions
  3. Add them as FAQ pairs with clear answers
  4. This alone can resolve 60-80% of queries

Organize by topic

Create separate contexts for:

  • Product information
  • Pricing and plans
  • Shipping and delivery
  • Returns and refunds
  • Account and billing
  • Technical support

Keep it updated

  • Add new FAQs when you notice repeated questions
  • Update product info when prices/features change
  • Remove outdated information
  • Review AI responses weekly for accuracy

Testing your knowledge base

  1. Go to Chatbot → AI Settings
  2. Use the "Test" feature
  3. Type questions as a customer would
  4. Verify the AI gives correct answers
  5. If wrong → add/update the relevant context

Measuring effectiveness

Track in Analytics:

  • Resolution rate — % of queries answered without human help
  • Handoff rate — % transferred to agents (lower is better)
  • Response accuracy — Review AI responses for correctness

Limits

  • Maximum 50 contexts per organization
  • PDF size: max 10 MB per file
  • URL crawl: max 50 pages per URL
  • Text content: max 50,000 characters per context

Was this guide helpful?

Your feedback helps us make these guides better for everyone.