Summary
Sonata delivered a generative AI-powered agent assist solution that transformed the client's customer inquiry response process. By implementing natural language processing (NLP) with context filtering and retrieval-augmented generation (RAG) technology, the solution reduced average handling time by 50-60% while automatically handling 20-25% of customer inquiries, eliminating the need for resource scaling during peak seasons.
Client Overview
A leading travel and tourism enterprise specializing in leisure travel services, offering a comprehensive portfolio that includes flights, hotel bookings, and vacation packages. With annual revenues in the multi-billion-dollar range, the company operates across numerous international markets and manages a diverse suite of digital platforms that serve millions of travelers globally.
Headquarter
Hanover, Germany
Revenue
€23.167 B
Pressure Points
The client’s customer service operations were overwhelmed by long handling times and seasonal query surges. Agents spent significant time searching through disparate document repositories for brand- and destination-specific information, especially during peak seasons.
10–12 minutes average handling time per customer inquiry
25–30% of inquiries require brand- or market-specific responses
Inability to scale operations during 2X seasonal surge in query volumes
Solutions
Sonata implemented a comprehensive generative AI-powered agent assist solution that streamlined the entire customer inquiry response process. The system utilized advanced NLP with sophisticated context filtering to understand specific brand, source market, country, and booking segment details for each question. Through a RAG approach, the solution rapidly located, prioritized, and generated relevant answers from a tailored knowledge base, ensuring accurate, context-sensitive responses while reducing agent dependency.
EventBridge scheduler triggers automated article processing every set hour, triggering the "kb-articlesprocessor" function
Lambda "kb-articles-processor" sends GET request to Mindtouch to get a unsorted list of articles
Lambda gets articles from the response and creates individual kb-article-work-items that are put on the queue
Lambda "kb-articles-processor" upserts record in mindtouch-extraction-service-log with "pageid" as key
Lambda "kb-article-work-item-processor" gets messages from queue (each work item is processed individually)
Technology Used
Harmoni.AI features
- Sanitation of user PI information
- User consent & awareness
- Content moderation
- Data security – user authentication
- Dashboard telemetry
Results that Speak Volumes
Enhanced customer satisfaction through faster, accurate responses
Greater operational agility during peak periods
Improved consistency in brand- and region-specific query handling
Stronger compliance with user data protection and moderation standards
50-60%* Reduction in Average Handling Time (AHT)
20–25% of inquiries managed directly by the agent assist system
Eliminated need for seasonal resource ramp-up
