HKBGE Enhances Upcoming Crypto Exchange Platform with Generative AI Chatbot on AWS

HKE Holdings, a listed fintech group focused on the Asia-Pacific region, established Hong Kong BGE Limited (HKBGE) to offer digital asset trading solutions for both retail and professional investors. In 2021, HKBGE applied for a virtual asset trading platform license under Hong Kong’s Securities and Futures Commission (SFC) to offer crypto exchange services.

Upon the anticipated launch of its trading platform, HKBGE foresees receiving a large volume of queries from thousands of interested customers once the license is granted by the SFC. The most common questions will likely relate to routine tasks such as account opening standards, checking balances, and deposit/withdrawal processes. Initial inquiries are expected to peak in the first few weeks post-launch. As the business expands, HKBGE anticipates a proportional increase in queries over time.

To help deliver a better customer experience, HKBGE looked to develop a generative artificial intelligence (AI) chatbot to provide personalized responses to frequently asked questions (FAQ) as customers visit the platform. The solution is designed to minimize resolution times and free up customer support staff to focus on more complex tasks and critical customer interactions. 

Since the license application in 2021, HKBGE has collaborated extensively with Amazon Web Services (AWS) on the development of its trading platform. Later in 2023, HKBGE’s IT team worked with AWS Solutions Architects to develop a retrieval-augmented generation (RAG) chatbot, which combines the advantages of retrieval-based models with generative AI models to provide more contextually relevant responses.

As part of the development process, the CS & IT team worked with AWS on the RAG chatbot’s search capability. This included looking at different databases, such as the open-source VectorDB. The two parties also conducted collaborative testing on Llama 2, a large language model (LLM) developed by Meta. The tests were performed on Amazon Elastic Compute Cloud (Amazon EC2) G5 instances with NVIDIA A10G Tensor Core GPUs.

Another part of the development process involved making sure the LLM could recognize inappropriate questions from bad actors. Cyrus Chan, network specialist at HKBGE, says, “We need to safeguard the chatbot from misuse by individuals with disruptive intentions.” This protection strategy was also part of a broader program to ensure the platform’s resilience.

Working with AWS solutions architects, HKBGE built its trading platform on AWS Well-Architected Framework best practices, including using multiple AWS Availability Zones and AWS Identity and Access Management (IAM) Identity Center features for business continuity.

In addition, HKBGE is able to offer secure and reliable digital asset trading services to professionals and institutions worldwide using AWS. “The AWS Availability Zones have provided high availability and assurance to our overall system architecture, which allows us to further develop the HKBGE trading platform without worries,” says Samuel Ng, tech operations director at HKBGE.

Today, financial services companies worldwide are adopting AWS solutions to accelerate innovation by delivering security and compliance at scale. In the case of HKBGE, it adopted Amazon GuardDuty, Amazon Inspector, Amazon Detective, and AWS Security Hub to maximize cloud security. Wilson Ho, chief operating officer at HKBGE, says, “The AWS team gave us a lot of support around compliance. For instance, AWS compliance specialists discussed with us how to meet the recommendations of the Securities and Futures Commission in Hong Kong.”

The platform also employs services like AWS Lambda, Amazon Simple Notification Service (Amazon SNS), AWS CloudWatch, and AWS CloudTrail for auditing. “Our integrated security approach on AWS provides real-time monitoring, ensuring adherence to regulatory standards and safeguarding the customer’s privacy,” adds Cyrus.

HKBGE is currently running its generative AI chatbot on AWS in a testing environment and plans to implement it with the launch of the trading platform, providing an excellent onboarding experience and platform for customers with personalized responses to FAQs. “Thanks to the flexibility of AWS, our generative AI chatbot can provide tailored answers to thousands of inquiries with speed and efficiency,” comments Wilson.

Once the chatbot is fully in production, HKBGE expects to reduce support costs for FAQs as much as 70 percent, freeing up resources to invest in staff training. Plus, the chatbot’s ability to operate 24/7 will provide round-the-clock personalized assistance to the company’s valued customers.

Beyond customer support, HKBGE plans to deploy the chatbot to address employees’ inquiries on internal processes and requirements. “We like the concept of the chatbot evolving into an internal knowledge base, providing snippets of information to enhance staff productivity,” says Wilson.

The HKBGE team is also exploring how generative AI can help customers in other potential aspects. “Generative AI on AWS has numerous potential use cases to enhance support for both our personnel and customers,” Wilson concludes.