Problem Statement:
The client’s existing system was limited by a predefined question and answer set for the chatbot. This led to poor user experience due to the chatbot’s inability to handle unexpected queries or complex issues. Additionally, there was no streamlined process for raising tickets, as this was done either through a separate functionality or manually by the agent, leading to inefficiencies and longer resolution times.
Solution Overview:
- Custom LLM Development: A large language model was trained specifically on the client’s question-and-answer sets. This custom LLM was designed to handle a wide range of queries by understanding the context and intent more accurately than traditional systems.
- Microservice for Automated Ticket Creation: A microservice was developed to automatically create support tickets based on interactions between the chatbot and customers. This service ensured that unresolved queries were systematically logged and tracked, reducing manual work and potential errors.
- Model Retraining with Feedback: The chatbot’s performance was continuously improved by retraining the LLM based on customer feedback and cases where the chatbot required agent intervention. This feedback loop enabled the system to evolve and become more accurate over time.
Tech Stack Leveraged:
HTML, CSS, Jquery, LLM, Custom Ticketing application (Laravel, MYSQL)
Benefits Delivered:
- Increased First-Contact Resolution (FCR): The AI chatbot resolved 60% more queries on first contact compared to the previous system, reducing the need for agent intervention.
- Reduced Agent Workload: By handling a broader range of queries, the chatbot reduced the volume of tickets needing human intervention by 45%.
- Improved Customer Satisfaction: Customer satisfaction scores improved by 30% due to faster and more accurate responses.
- Efficiency in Ticket Management: Automated ticket creation reduced the time taken to log issues by 70%, allowing agents to focus on more complex tasks.
- Reduced Response Times: Average response times decreased by 50%, leading to quicker resolutions and enhanced customer experience.
