RAG AI Onboarding & Knowledge Bot for Café Staff
Diana Chykhabakh
RAG AI Onboarding & Knowledge Bot for Café Staff (n8n + OpenAI)
About the project
Client
A modern café chain with multiple locations focused on improving onboarding quality, operational consistency, and staff efficiency.
Challenge
Managers were overwhelmed by repetitive questions from staff, including:
How to process refunds
POS and payment terminal issues
Shift procedures and closing checklists
Inventory and supply questions
This led to onboarding delays, frequent mistakes, and increased management workload.
Solution
I built a fully automated RAG-based AI onboarding and knowledge assistant that delivers instant, accurate answers based strictly on internal documentation.
Core architecture:
OpenAI Vector Store for document embeddings and retrieval
n8n for workflow orchestration and Telegram integration
Telegram Bot API as the staff interface
The assistant retrieves verified knowledge from internal manuals and converts it into short, clear, step-by-step instructions.
How it works
Internal manuals are transformed into a structured vector knowledge base
Staff ask questions directly in Telegram
The AI retrieves relevant content using RAG (no hallucinations)
Answers are formatted for fast, on-shift use
The system operates in a closed, controlled workflow
Key highlights
True RAG architecture — answers only from verified internal content
Instant responses during live shifts
Zero hallucinations
Multi-location support
Covers refunds, POS, shift procedures, and daily operations
Results
Onboarding time reduced by 47%
Management interruptions reduced by 70%
Operational consistency improved across all locations
Staff gained 24/7 access to accurate procedures
Knowledge scaled without adding managerial overhead
Tech Stack
n8n
OpenAI (Vector Store + LLM)
Telegram Bot API
PDF-based internal documentation
Like this project
Posted Jan 8, 2026
Built a RAG-based AI onboarding bot that answers staff questions instantly, centralizes knowledge, and reduces training time for café teams.
