Pelago, a travel experience platform established by the Singapore Airlines Group, has integrated generative AI (GenAI) to automate assistance for customers throughout the booking process, from itinerary planning to post-booking enquiries.
It has integrated an AI-powered travel assistant to solve customer queries and help create support tickets directly on the platform.
This eliminates the limitations of traditional support channels.
Pelago’s director of data, Akshay Bhonde told iTnews Asia that after evaluating vendors including Freshdesk and Verloop, the company selected the Yellow.ai platform to manage and analyse customer interactions and optimise service delivery.
Previously, the average turnaround time to resolve customer enquiries was two to three hours.
Now, this has been reduced to 30 seconds, said Bhonde.
According to Bhonde, by automating responses, Pelago has seen a 60 percent decrease in the number of product enquiry questions that agents need to address individually.
This has freed up time for agents to focus on handling urgent requests and providing personalised support where necessary, he added.
The company has integrated solutions on WhatsApp and the web platform.
It has plans to expand integration to other channels depending on requirements for specific markets.
While the majority of Pelago’s team is based in Singapore, it has employees working remotely from Indonesia, Australia, and Malaysia.
The platform allows users to book activities across the globe, with over 200,000 listed activities.
Underlying technology
According to Bhonde, the underlying technology of the LLMs involves training on large, general-purpose datasets.
He added that Pelago’s in-house team fine-tunes LLMs with proprietary travel-related data including information about activities in various countries, pricing details, and contextual data.
The team regularly refines and enhances LLMs through a feedback loop utilising customer interactions and analytics to ensure the travel assistant delivers accurate and relevant responses, said Bhonde.
This helps Pelago deliver personalised and efficient services to users and travel agents, enhance customer experience and streamline internal processes.
Yellow.ai’s co-founder and chief product officer, Rashid Khan, said the company’s GenAI platform, YellowG, allows Pelago to utilise multiple LLM models for tasks including query understanding, sentiment analysis, and response generation.
This makes conversations human-like by incorporating emojis and visual elements, said Khan.
He added the platform offers a user-friendly interface, enabling complex workflows without coding and allowing for coding extensibility when needed.
The platform offers post-conversation analytics, enabling Pelago to gain insights into user interactions, such as booking inquiries and user preferences, Khan said.
Mitigating hallucination
According to Bhonde, a primary concern was mitigating the risk of “hallucination”, where LLMs might provide inaccurate information, leading to negative user experiences and financial losses.
To address this, Pelago’s team engaged in extensive prompt engineering and testing, iterating multiple times to fine-tune the LLMs’ responses to ensure accuracy and reliability.
Another challenge includes adapting the solution to Pelago’s specific use cases and customer interactions, which requires an understanding of user behaviour, said Bhonde.
The Yellow platform’s feedback loops analyse interactions and feedback, he added.
It helps in personalising the travel assistant’s responses, incorporating additional information and optimising actions including booking cancellations or voucher requests.
Khan said organisations need to prioritise the top three or four use cases for which customers are enquiring, derived from call volumes, chat interactions, and emails.
Customised messaging
Pelago plans to expand channel integrations, improve omnichannel support, and use Yellow.ai’s capabilities for proactive engagement through customised messaging for growth and retention.
The company plans to tap GenAI to automate content creation and enhance SEO.
This will help identify duplicated product listings across suppliers, ensure coherence and reduce confusion, Bhonde said.
Pelago employs GenAI for review augmentation, categorisation, and summarisation, he added.
Additionally, the company uses GenAI to analyse and categorise user reviews, providing concise summaries of customer feedback.
This automated process highlights aspects of user experiences, including hospitality quality and specific preferences.
By streamlining review analysis, Pelago enhances the accessibility of user feedback and gains actionable insights into customer preferences, said Bhonde.
This helps in product development, marketing strategies, and customer service initiatives to enhance responsiveness, he added.