The Bank of Thailand (BOT), founded in 1942, is the central bank responsible for ensuring economic stability and sustainable growth in Thailand, a nation of over 70 million people.
Overcoming SQL complexity to unlock critical insights
In 2024, the BOT launched the Regulatory Data Transformation (RDT) program, aimed at creating a comprehensive data repository to provide a detailed understanding of Thailand’s economic landscape and support informed policy decisions.
Central to this initiative was a vast trove of loan-level data, encompassing over 1,000 fields and 25 million loan accounts per month. Cloudera was selected for its robust data management and analytics tools, which are essential for handling the large and complex RDT dataset.
Cloudera was implemented in a hybrid cloud environment, allowing BOT to manage and analyze the extensive loan-level data effectively. However, traditional methods of managing and analyzing this extensive dataset required specialized Structured Query Language (SQL) coding skills. SQL is a programming language used to manage and manipulate databases, but its complexity created bottlenecks, limiting access to critical insights. Analysts struggled to efficiently extract meaningful information due to the sheer volume and complexity of the data.
Recognizing these challenges, the BOT needed to upskill its workforce through initiatives like the Data Expert Program. However, traditional training methods and knowledge transfer from domain experts were time-consuming and inefficient. These issues highlighted a pressing need for a solution that could democratize access to data, enabling employees without extensive technical backgrounds to harness the power of this information.
Streamlining data queries with natural language processing
To unlock the full potential of the data stored on the Cloudera platform, the BOT developed the SQL Coding Copilot, an AI-powered tool that acts as a translator, allowing anyone regardless of technical skill proficiency, to analyze and extract insights from data.
The Copilot converts natural language queries into SQL code, streamlining the analysis process and making it more efficient and accessible. Employees use the Copilot by typing questions such as “What is the total loan amount for the last quarter?”. The Copilot then translates this query into the appropriate SQL code and retrieves the relevant data. The process eliminates the need for complex coding, lowering the barrier to entry and encouraging curiosity and exploration among employees who were previously daunted by data analysis.
With its ability to understand schemas in the Cloudera data lakehouse, two extra features — “metadata copilot” and “Brainstormer” — help users unfamiliar with the thousands of fields in RDT to quickly learn, locate targeted data, and get answers to their economic questions. “Brainstormer” leverages both the agentic paradigm for generative AI and a reasoning model that breaks down analytic questions into multiple tasks and steps using a multi-agent approach.
Building on these capabilities, SQL Coding Copilot laid the foundation for RDT Copilot. It enables users to ask analytic questions informed by economic and RDT data stored in the data lakehouse, further simplifying the process of deriving valuable insights.
Looking ahead, BOT is exploring opportunities to implement text-to-insight capabilities and autonomous analytic agents that seamlessly integrate with Cloudera and the Copilot. These advancements will enable the generation of analytic dashboards, business intelligence visualizations, and actionable insights with unprecedented ease, further enhancing the BOT’s data-driven decision-making processes.
Additionally, the integration of the RDT Copilot with the internal knowledge management chatbot and the SupAI platform, the generative AI Q&A with the minutes of meetings of the board of directors of all financial institutions has further enhanced regulatory compliance by providing quick and accurate answers to regulatory questions and assisting examiners in identifying potential areas of non-compliance.
By providing a comprehensive view of the credit landscape, the tool enables BOT examiners, financial analysts, and economists to proactively identify and address emerging risks, ensuring the stability and resilience of the financial sector. This, in turn, fosters investor confidence, promotes responsible lending practices, and contributes to monetary and supervision policies for sustainable economic growth.
Unlocking the full potential of a data-driven culture
The AI capability unlocks the full potential of the data stored in Cloudera, enabling the BOT to gain deeper insights into the financial landscape and make more informed decisions that positively impact Thailand’s economic development and stability. The RDT Copilot family provides deeper insights into loan-level data and enables the bank to better assess and mitigate risks, make data-driven decisions on monetary policy, and rapidly evaluate vulnerabilities and stress-test scenarios.
Since its launch, the RDT Copilot has empowered a wider range of employees to contribute to data-driven decision-making by enhancing the coding and data analytics skills of over 100 examiners in BOT’s Data Expert Program. This has facilitated a data-driven culture across the organization and significantly reduced the time and resources required for data analysis, promoting a more efficient and engaged workforce.
“Our vision is to create a truly data-driven culture, where every decision is informed by accurate and timely insights,” said Dr Art Chaovalitwongse, Head of Data Management & Analytics, BOT. “As we continue to innovate, we aim to make data even more accessible and actionable for all our employees. This will empower everyone at the Bank of Thailand to harness the power of data, leading to better decisions and a stronger, more resilient financial ecosystem for Thailand.”
