We recently hosted an internal hackathon called the GPT Hackathon, where talented Agodans from various teams, such as Tech, Product, and Customer Experience Group, put their skills to the test by building innovative applications utilizing the cutting-edge language model, ChatGPT. It was a great opportunity for our colleagues to learn and stay ahead of the game with the latest technology while working with a team to pitch exciting, game-changing ideas.
Over 40 participating teams explored use cases of ChatGPT in the travel tech space, ranging from conversational interfaces for booking hotels and generating recommendations for supplier dashboards to personalized travel itineraries and improved customer service. At Agoda, we believe in exploring the newest technologies while having fun and collaborating, and our GPT Hackathon was a perfect blend of both.
In this blog, we will take a closer look at some of the top projects, as well as key insights and takeaways from this exciting event.
Hotel Gini (Property Page Chat)
“This team showed the best use of language generation.” Dries Geebelen, Senior Director, Head of Data Science
Their pitch: The property page contains an overwhelming amount of information, which can result in cognitive overload, difficulty finding answers, and the possibility of visitors leaving the page. Integrating a chatbot with GPT will provide timely information on the property page and reduce potential drop-offs.
How: With the help of the GPT-3 davinci model, this team designed and engineered a bot to provide quick and human-like responses to users, resulting in a higher conversion rate from property page to booking. Their solution scans all property content (including reviews) and, with the right context, provides users with the answers they’re looking for.
SupportHero (Support Assistant)
Winning: Data Diviner
“This award was given for the best use of historical data. This team was able to use already existing data to create their bot.” Idan Zalzberg, Chief Technology Officer
Their pitch: Support teams frequently spend their time resolving repetitive tickets, which can lead to longer wait times for end users due to the workload on the support team. Using GPT can help alleviate the burden of answering repetitive questions for support teams at Agoda.
How: As Slack is heavily used at Agoda, this team leveraged the existing support channels by creating a second brain from the highly valuable information available in the existing conversations. They extracted all conversations from the Slack support channel and then created a Slackbot that, when called, finds similar conversations and passes them as context to GPT. The bot then asks GPT to answer the new questions using the given context.
The team also published an article to explain the project. You can read it here.
Agodata (Data Analyzer)
Winning: Jedi Mind Reader Award
“They showed the best use of NLP (Natural Language Processing). Their idea allowed users to translate plain English into queries.” Varuntida Varutbangkul, VP of Product
Their pitch: Over 650 agodans run 500K “simple” queries each month that require analytical understanding to analyze the results. GPT can help create database queries using simple natural language, analyze the data result, and save 200 hours of manpower through that.
How: With ChatGPTs’ advanced comprehension of datasets and column descriptions, this team used ChatGPT to swiftly detect crucial insights, trends, and anomalies and even generate spontaneous queries, allowing users to interact with the results from the analysis or ask further questions in a friendly chat format. This saves time and increases accuracy.
Team Random (Email Summary)
Winning: Venture Capitalist Dream
“Their idea was most likely to get funded. The idea for this team is worth exploring more.” Ittai Chorev, Chief Product Officer
Their pitch: Agents spend 18% of their time generating email and messaging responses for users. Using GPT can improve agent efficiency by reducing the time agents spend responding to customer queries related to bookings.
How: Agents are continuously getting queries from customers related to booking, either in the form of emails or messaging, and they spend considerable time responding to these customers. This team used OpenAI’s GPT models to help generate complete email responses based on information entered by agents or directly from the customer’s query. This can reduce the time it takes for agents to respond to these queries.
AIST (Supply AI Assistant)
Winning: Most Likely to Give a TED Talk Award
“This team was able to grab the audience with the shortest summary of their solution right at the start of the presentation. Concise and precise.” Tim Hughes, VP of Business Development
Their pitch: Over 50% of customer tickets can be resolved through basic research and data. Leveraging GPTs capabilities can expedite the process of addressing customer support tickets.
How: Implementing automated responses for frequently asked questions using GPT can significantly cut down the time required by agents to address such queries. This, in turn, will expedite the overall process, leading to quicker response times and 24/7 availability.
Excited to learn and use the latest technologies? Join us in our tech team!
Engineering Manager (Full Stack) (Bangkok based, relocation provided)