A large European travel website has hundreds of customer booking requests daily. The majority are urgent, with customers about to start their journey or already in a foreign country. The increased cost pressure in the industry is forcing companies to invest into automation.
Conversations about bookings tend to get complex very quickly. There are a lot of specific rules and not many standards, meaning simple FAQ-based bots typically add little value for customers. Deep integrations into backend systems are required, as well as a way to train specific machine learning models for the travel industry vocabulary.
Using Rasa Platform, the company built a conversational travel agent which could handle complex interactions with customers. The agent is end-to-end automated with machine learning-based dialogue management and deep backend integrations. The company was also able to improve the accuracy of their language model by integrating real conversations.
The company went from idea to deployment in less than three months. Focusing on a few use cases with high volume of customer service on Facebook Messenger, after a short training phase the AI assistant was able to handle more than 25% of requests. Now, the company plans to expand into further use cases and conversational platforms like WhatsApp and live web chat.