1. Project Meta
Develop an AI agent for automatic identification of product analogs and formation of cross-sells with integration into BAS KUP through the "Interaction System" module (Discussion).
2. Business Goals
- Reduce sales losses due to product unavailability
- Increase average check through cross-sells
- Standardize recommendations for managers
- Reduce dependence on manager experience
3. Solution Architecture
The AI service is implemented as a separate microservice with a REST API. BAS KUP serves as the data source and API client.
Components:
1. BAS KUP (source of products, orders, stock)
2. AI service (logic of analogs and cross-sells)
3. REST API (data exchange)
4. BAS integration module (configuration extension)
4. Core Functionality
4.1 Identification of analogs:
- Exact (exact analog)
- Close (close analog)
- Alternative (alternative solution)
4.2 Formation of cross-sells:
- Must-have
- Recommended
- Upsell
- Bundle
5. AI Trigger Points
- When opening an order
- When adding a product to an order
- When a product is out of stock
- When pressing the "AI Hint" button
6. API Format
Endpoint: /analogs
Endpoint: /crosssell
Method: POST
Exchange format: JSON
Example response:
analogs_exact[], analogs_close[], alternatives[]
crosssell_must[], crosssell_recommended[], upsell[]
manager_script, reason_short
7. Integration with BAS KUP
Implement through configuration extension without changing the standard database. BAS makes an HTTP request to the AI service and publishes the response in the "Discussion" tab of the corresponding document.
8. AI Limitations
- Do not invent characteristics
- Do not change prices or stock
- Do not provide delivery guarantees
- Use only data from BAS
9. Security
- Authorization via API token
- Logging of requests
- Access restrictions to internal data
10. Implementation Stages
1. Audit of product structure
2. Development of AI service
3. Development of extension for BAS
4. Testing
5. Launch
