How to Create a Knowledge Base for Your WhatsApp Chatbot

Introduction

"The difference between a WhatsApp chatbot that frustrates customers and one that generates sales lies in a single factor: the quality of its knowledge base." A chatbot can only be as effective as the information it receives. If the knowledge base is incomplete, outdated, or poorly structured, the responses will be inaccurate.

This comprehensive guide covers what information to include, content formats to use, how to structure documents, keep the database updated, and verify the quality of responses.

What is a Chatbot Knowledge Base

A knowledge base is "the set of documents, files, and information that feeds your chatbot so that it can accurately answer your customers' questions."

The technology uses Retrieval-Augmented Generation (RAG). When a customer asks a question, the system searches the knowledge base for relevant fragments, and the AI model generates natural responses based on real business data.

"Think of the knowledge base as the brain of your chatbot."

What Information Should Your Knowledge Base Include

1. General Company Information

• Trade name and company name
• Physical location with references
• Opening hours (including holidays)
• Contact information
• Brief History
• Mission and values
• Branches or points of sale

2. Products and Services

• Name and detailed description
• Technical characteristics
• Main benefits
• Available variants
• Availability and stock
• Guarantees

3. Prices and Plans

• Individual prices
• Plans or packages
• Current discounts
• Accepted forms of payment
• Additional costs
• Comparison between plans

4. Policies and Conditions

• Return policy
• Shipping policy
• Cancellation policy
• Warranty policy
• Privacy Policy
• Terms and Conditions

5. Frequently Asked Questions (FAQ)

• Questions about the purchase process
• Questions about product usage
• Questions about after-sales service
• Comparative questions

6. Procedures and Processes

• How to place an order
• How to schedule an appointment
• How to request a quote
• How to report a problem
• How to activate a warranty

Content Types for Your Knowledge Base

PDF documents

PDF files are versatile for knowledge bases:

• Product Catalogs
• User manuals
• Price lists
• Commercial brochures
• Technical Data Sheets
• Procedure guides

Best practices:

• Use selectable text, not scanned images
• Includes clear headings
• Avoid overly complex tables
• Keep each PDF focused on a specific topic

Web Pages (URLs)

Your website contains valuable information:

• Products or services page
• Pricing page
• Blog and articles
• FAQ page
• About us" page
• Landing pages

Advantage: If you update the site, you can re-crawl URLs to automatically reflect changes.

Text Documents

For information without any other format:

• Adapted sales scripts
• Model responses
• Seasonal information
• Specific data

Structured Product Catalogs

For large inventories:

• Name, description and SKU
• Price, variants and availability
• Categories and subcategories
• Image URLs

How to Structure Your Content for Better Responses

Use Clear and Descriptive Headings

Bad example: "Section 3", "Additional Information"

Good example: "Aurora IA 2026 Plan Prices", "Return Policy for Electronic Products"

The Question-Answer Format Works Best

"The Q&A format is the most effective for knowledge bases because it replicates exactly how customers formulate their queries."

Example:

Q: What is the delivery time for domestic shipments?

A: Domestic shipments take 3 to 5 business days for urban areas and 5 to 8 days for rural areas. Free shipping on purchases over 1,400,999 MXN.

It includes specific and concrete data.

Avoid vague information. The chatbot needs accurate data.

Bad example: "Our prices are competitive and we offer shipping nationwide."

Good example: "The Aurora CRM Plan costs 1,800 MXN/month and includes 3 users, 1 AI agent, and 800 AI responses per month."

Organize by Topics, Not by Department

Structure your business by thinking about how your customers search for information, not how your company is organized.

Topic-based approach (most effective):

• All about shipping
• All about payments
• All about products

Knowledge Base Maintenance and Updating

Establish a Review Schedule

• Weekly: Prices, promotions, availability
• Monthly: Policies, procedures, contact
• Quarterly: Full review
• Immediate: Critical changes

Version Control

Keep a record of:

• Document name
• Date of last update
• Responsible for maintaining it
• Next review date

Gap Analysis from Unanswered Questions

Check:

• Questions that triggered transfer to agent
• Low-rated conversations
• Recurring topics not covered
• Questions about new products

"Every unanswered question is an opportunity to strengthen your knowledge base."

Team Feedback

Your team is an invaluable resource:

• Agents report new frequently asked questions
• Sales share common objections
• Support documentation for technical solutions

Quality Tests for your Knowledge Base

Testing with Boundary Cases

• Ambiguous questions
• Multiple questions in one message
• Questions with spelling errors
• Negative questions
• Comparative questions
• Questions with incomplete context

Precision Verification

Verify that the answer:

• Be factually correct
• This updated
• Be complete
• Be consistent
• Cite the correct source

Real-World Scenario Testing

It simulates complete conversations as a real customer would, verifying consistency throughout the conversation.

How Aurora Inbox Makes Creating Your Knowledge Base Easy

Aurora Inbox connects to WhatsApp in two ways: through QR code with regular WhatsApp (ideal for getting started quickly without paperwork) or through the WhatsApp Business Cloud API (for advanced features such as mass messaging and catalogs). In both cases, the knowledge base works in the same way.

Training with PDF Documents

Upload PDFs directly. The system automatically processes the content and indexes it for semantic searches.

Training with URLs

Provide URLs from your website. Aurora Inbox automatically crawls the content of each page.

Product Catalog Training

Upload your structured catalog with names, descriptions, prices, categories, and images.

Combination of Sources

You can combine all these sources simultaneously for a seamless conversational experience.

Checklist for Creating Your Knowledge Base

• Contact information and full location
• Updated opening hours
• Catalog with detailed descriptions
• Current price list
• Return, shipping, warranty, and cancellation policies
• 20-30 frequently asked questions with answers
• Step-by-step purchase process
• Accepted forms of payment
• Shipping Information
• Differentiating factors and value proposition
• Current promotions and discounts
• Appointment scheduling procedure
• Support information

Frequent questions

How many documents do I need to get started?

It is recommended to start with 3-5 essential documents: general information and schedules, catalog with prices, policies (shipping, returns, warranty), FAQ with 20-30 questions, and procedures (how to buy, schedule).

In what format should I upload my documents?

"PDFs with selectable text offer the best results." Avoid PDFs that are scanned images. URLs also work great and can be re-tracked when you update.

How often should I update the database?

Prices and promotions are updated weekly if you have rotating offers. Policies and procedures are updated monthly. A full audit is conducted quarterly.

How do I know if my knowledge base is working?

Review the percentage of conversations resolved without transfer (target: 70-80%). Analyze satisfaction ratings. Review conversations where you failed. Conduct periodic tests.

What happens if the chatbot responds with incorrect information?

The problem is usually in the knowledge base. Identify the incorrect document, update it with the correct information, and test to verify the answers.