Use Case Transportation
Generative AI travel assistant
GPTravel utilizes advanced AI technologies to deliver tailored travel and flight recommendations based on individual user preferences.
This intelligent travel assistant helps users plan their trips from start to finish, offering customized suggestions and valuable insights to ensure a seamless travel experience. GPTravel can search for flights to recommended destinations, ensuring users find the best travel options available.
Challenges
The project faces challenges arising from inconsistent email usage and device sharing, as well as difficulties in maintaining high recommendation precision due to variations in user behavior and limited data.
Data integration
Merging Google Analytics with CRM data is complex due to varied email usage and device sharing, complicating user identification.
Model Accuracy
Maintaining high recommendation accuracy is challenging, influenced by variations in user behavior and lack of data.
Solution
A chatbot based on a multi-agent system of LLMs has been implemented, where each agent performs a specific task. One agent handles the conversation, another creates dynamic queries, another provides destination… This system has been integrated into an API called from the airline’s website.
To control costs, access is currently limited to the airline’s premium customers. Response times have been minimized by using lighter models and reducing the verbosity of responses. Communication between agents has been optimized to minimize errors by parsing the output of each agent at the code level, ensuring smooth and accurate interactions.
Tech stack
Results
The chatbot maintains fluid communication with customers, addressing various airline-related topics. It plans trips, searches for flights based on preferences, and recommends destinations using the airline’s database. This is expected to increase customer satisfaction and boost revenue by redirecting customers to attractive flight options.
