Skip to main content
Normalized for Mintlify from knowledge-base/aiconnected-apps-and-modules/modules/kb-generator/ai-knowledge-training-system.mdx.

AI Knowledge Training System

The Insight

The Skin Beauty AI works exceptionally well not because of the model, but because of the structured knowledge base. This wasn’t scraped content—it was carefully authored educational content designed to:
  1. Map customer concerns to services
  2. Help customers self-identify their needs
  3. Explain the “why” behind each service
  4. Guide decisions, not just list features
The goal: Replicate this quality for any business from just a URL.

The Transformation Pipeline


Admin UI Flow

Screen 1: Getting Started

Screen 2: Analysis Progress

Screen 3: Review Services

Screen 4: Review Knowledge (Expandable)

Screen 5: Conversation Style

Screen 6: Test & Launch


Data Schema


AI Prompts for Each Transformation Step

Prompt: Entity Extraction

Prompt: Knowledge Enhancement (per service)

Prompt: System Prompt Generation


Technical Implementation

Crawling Options

  1. Firecrawl - Best for JS-heavy sites, returns clean markdown
  2. Apify - More control, handles complex sites
  3. Custom Puppeteer - Full control, more maintenance

Processing Pipeline


Cost Estimation

Per training run (typical business with 10-20 services): This is a one-time cost per business setup, with optional re-training when they update their website.

Success Metrics

A well-trained AI should:
  1. Recommend relevant services for any concern mentioned
  2. Never hallucinate services or prices that don’t exist
  3. Maintain brand voice consistently
  4. Guide toward booking naturally
  5. Handle edge cases gracefully (unknown concerns, competitor questions)
  6. Self-identify triggers resonate with actual customers

Future Enhancements

  1. Automatic re-training - Monitor website for changes, re-crawl weekly
  2. Analytics feedback loop - Learn from which recommendations convert
  3. Industry templates - Pre-built concern maps for common business types
  4. Multi-location support - Different services/prices per location
  5. Seasonal adjustments - Automatically suggest seasonal services