The Challenge
A Gurgaon-based e-commerce company selling fashion accessories was spending ₹4L/month on a 10-person support team handling repetitive order status, return policy, and product availability queries. Response time averaged 4 hours.
Our Solution
We deployed a RAG chatbot trained on their product catalog, order database, and return policies. The chatbot integrated with Shopify and their logistics API to provide real-time order tracking. Multi-language support was added for Hindi and English.
How We Built It
- 1
Analysed 3 months of support transcripts to identify the 12 highest-volume query categories and the data sources needed to answer each one.
- 2
Built a unified knowledge base combining their product catalog (Shopify), return policy, shipping guide, and size charts, with nightly syncs to keep inventory data fresh.
- 3
Developed a RAG pipeline with hybrid search — vector similarity for semantic queries and BM25 keyword search for exact matches like order numbers and SKUs.
- 4
Integrated with the Shopify API for live order status and the logistics provider's API for real-time tracking updates.
- 5
Added Hindi and English language support with automatic language detection based on the customer's query.
- 6
Deployed a fallback escalation flow: if the chatbot could not resolve a query after two attempts, it handed off to a human agent with the full transcript.
Integration & Architecture
The chatbot connected to Shopify via its REST API for order lookups and to the logistics provider's tracking API. The knowledge base was rebuilt nightly via a scheduled job that pulled the latest product catalog and policy documents.
The Results
- 80%Auto-Resolved Queries
- ₹4L/moCost Savings
- 2 minAvg Response Time
- 4.7★Support Rating