Generative AI/LLM Based Customer Service and Advanced Product Recommendation
Engine For Ecommerce
The Challenge
A startup in the US aimed to provide boat owners with expertise on boat and engine maintenance through use of a GenAI based chatbot. The firm also aimed to generate revenue through highly personalized product recommendations for boat maintenance items in the context of the conversation with the synthetic expert. The Tatras team was asked to deliver on such a generative chatbot that can provide human expert level advice to boat enthusiasts.
Hypothesis
- Given complex engine and boat manuals, a RAG framework can provide expert level quality answers to questions around boat maintenance and engine trouble shooting.
- Structured data, such as product databases and CRM systems can be leveraged along with open and closed sets of unstructured information to optimize high quality personalized responses.
Execution
- A multi-agent architecture was developed.
- Agents were given access to specific knowledge bases around different OEMs such as engine manufactures manuals.
- Agents has access to structured database such as the product catalogue or CRM data.
- A query was routed to multiple agents that provide follow-up questions for refinement of recommendations or answers to the query.
Outcomes
- 80% positive feedback from consumer surveyed interactions.
- The chatbot is in production.
- It handles registration of users on the platform, provides advice on engine or boat maintenance issues, videos from open source libraries and recommendations of appropriate products.
- When required the chatbot routes the query to a human expert and learns from the interaction.
Project Highlights
25%
added revenue per chatbot
consumer interaction
consumer interaction