Planning Perfect Holidays Faster with Generative AI

Our impact

• Enhanced operations by enabling Hotelplan’s sales staff to provide faster and better recommendations to customers for their travel experiences with the Generative AI tool
• Improved the efficiency and speed of Hotelplan’s sales staff to provide customized information to customers
• Accelerated the time to provide expert recommendations to travel questions from hours to minutes

 

The challenge

Imagine being able to answer detailed customer questions and needs with the connected travel expertise of 500+ travel experts in seconds. 

Hotelplan’s booking team is made up of travel experts, with deep knowledge about specific countries and locations. This means that when a new customer wants to plan a holiday outside of their expertise area, they consult other team members to be available to help recommend the perfect itinerary, which can take some time. 

Hotelplan wanted to build a tool that leverages all of their experts’ internal knowledge, combined with external, structured, and unstructured data using semantic search, and provides instant responses.

Powered by Generative AI, the tool would allow internal sales staff to advise customers on their perfect holiday instantly.

“For the past 3 years, we have looked at ways to best share our staff’s expert know-how across all our locations. With an AI chatbot we can finally achieve this.” – Lukas Karrer, Director of Direct Business, Hotelplan

 

Our solution

Using Google Cloud’s latest AI technology, Datatonic collaborated with Hotelplan Suisse to develop its Generative AI-powered knowledge sharing chatbot, which: 

 

• Ingests data from 10+ internal and external sources with an automated pipeline for ingestion of new versions of the data sources
• Uses guardrails to prevent undesirable outputs, such as hallucinations or harmful content
• Enables users to save their current chat history and load previous chat histories in a new session

 

For the team delivering this innovative project, this involved:

• Optimizing model outputs through prompt tuning and multiple rounds of UAT to improve performance
• Refactoring to separate the Gradio frontend from backend logic such as LLM calls
• Carrying out automated unit tests to verify basic functionality, e.g., APIs, data availability, and processing logic
• Creating production-ready CI/CD pipelines for automated deployment
• Enabling logging of inputs and outputs from the LLMs in Hotelplan’s Google Cloud project, available in BigQuery

With this solution being rolled out, Hotelplan will accelerate its sales process and serve personalized recommendations to customers faster through its experts.

To leverage GenAI further, Datatonic is also working with the Marketing and Content teams to help speed up writing location descriptions and blogs for its website. Previously this would require time from a copywriter or sales staff to turn a report of destinations into a blog, severely limiting output capacity. With GenAI, this process will be much faster; teams will be able to quickly turn the travel report into a blog, with a choice of three different tones of voice depending on the specific brand.

“Delving into GenAI with Datatonic has been a fascinating experience. In this fast-paced and constantly evolving environment, we collaborated closely to test the boundaries of what’s possible and unlock its potential.” – Lukas Karrer, Director of Direct Business, Hotelplan