United Overseas Bank (UOB) and Singapore’s Agency for Science, Technology and Research (ASTAR) announced on Nov. 18 that they are exploring a multilingual artificial intelligence system designed to assist relationship managers during customer conversations. The project will use ASTAR’s MERaLiON model and examine how AI can support product disclosures, language interpretation and post-call compliance reviews across UOB’s regional operations. PR Newswire APAC
The development places AI in a sensitive part of banking: direct conversations between customers and the people advising them about financial products.
Rather than presenting the system as an automated banker, UOB’s announcement describes a tool intended to support employees. Relationship managers will remain accountable for the conversations and advice they provide.
That distinction matters. In financial services, an AI system that helps interpret a customer’s question or reminds an employee about a disclosure must still be accurate, explainable and subject to human oversight.
What UOB and A*STAR are developing
The proposed system is called the Compliance Intelligence Engine and Evidence-Grounded Post-Call Compliance Review System.
It is built around MERaLiON, a multilingual model developed by A*STAR. The model is intended to process speech and language patterns found across Southeast Asia, including conversations that move between languages or use local expressions.
UOB said the system could interpret customer questions across different languages, accents and communication styles. During a conversation, it could provide relationship managers with prompts about relevant product information and risk disclosures.
The second component comes after the call. The system could review conversation evidence to assess whether important information was communicated clearly and whether fair-dealing standards were met.
This is different from simply producing a transcript. A transcript records what was said; a compliance review attempts to help assess whether required information was properly communicated.
UOB and A*STAR have not presented the project as a completed operational system. Their announcement describes an exploration of how the technology can be enhanced and applied safely at scale. PR Newswire APAC
MERaLiON’s language capabilities extend across ASEAN
MERaLiON supports multilingual processing and code-switching across 11 ASEAN languages, according to the announcement.
The listed examples include Malay, Bahasa Indonesia, Thai and Vietnamese, alongside English, Mandarin, Cantonese and Singlish.
Code-switching occurs when speakers move between languages or dialects during the same conversation. This can happen naturally in multilingual communities, but it can make automated transcription and interpretation more difficult.
For a regional bank, this capability could be useful when customers explain financial needs using a mixture of formal language, local expressions and familiar terms.
UOB said it serves more than eight million customers across ASEAN. The five markets identified for the initiative are Singapore, Malaysia, Thailand, Indonesia and Vietnam. PR Newswire APAC
One technical detail needs careful interpretation. The announcement also refers to MERaLiON achieving the lowest average word-error rate among open-source models across languages. Its footnote identifies seven languages for that particular comparison. That benchmark should not be confused with the broader 11-language processing capability.
Why the post-call review could matter
Banking conversations can involve more than explaining a product’s basic features.
Customers may need to understand fees, risks, eligibility requirements and other important conditions before making a financial decision. A missed or unclear disclosure can create problems for both the customer and the institution.
UOB’s proposed system is intended to provide an additional review layer by examining whether important information was communicated.
For the bank, this could help identify areas where relationship managers need additional coaching. It could also reduce some manual translation and post-call review work, although the announcement does not provide measured productivity gains.
The project therefore has two potential functions: supporting employees while conversations are taking place and helping the bank examine the quality of those conversations afterwards.
The wider banking industry is also exploring how AI affects customer discovery and service. WWC previously examined how AI-generated answers are changing the visibility of Philippine banks. UOB’s project addresses a different part of the relationship: the interaction between a customer and a bank employee.
Human accountability remains central
UOB’s announcement explicitly states that relationship managers remain accountable for the conversations and advice they provide.
That is an important boundary. An AI prompt should not become a substitute for an employee’s professional judgement, particularly when a conversation concerns financial risks or a customer’s individual circumstances.
The system’s usefulness will depend on the quality of its language processing, the relevance of its prompts and the reliability of its post-call assessments.
A tool that misunderstands a dialect, misses context or incorrectly identifies a disclosure could create additional work rather than reduce it. These are risks to be evaluated during development and testing, not problems that the announcement says have already occurred.
UOB also has an existing governance framework. Its published data-privacy information describes a Group Data and Artificial Intelligence Governance Framework, an AI Ethics and Model Governance Council and a Data and AI Committee. The bank says its AI governance work incorporates the Monetary Authority of Singapore’s Fairness, Ethics, Accountability and Transparency principles. Data Privacy
This governance context is relevant because customer conversations may contain personal and financial information. Any practical deployment will need to address how such information is handled, accessed, retained and protected.
A wider push to bring AI into banking operations
The MERaLiON initiative is part of UOB’s broader effort to expand the use of advanced technology across its business.
In April 2025, UOB announced a three-year memorandum of understanding with Accenture covering the adoption of generative AI and agentic AI. That collaboration included ambitions involving business processes, productivity, risk management and more personalised customer experiences. UOB Group
The bank has also announced a separate collaboration with Amazon Web Services to explore a dedicated cloud environment in Singapore and AI-driven banking innovation across ASEAN. These projects show that UOB’s technology strategy extends beyond one customer-call application.
The wider shift is also visible in UOB’s participation in live tokenised Singapore-dollar deposit transactions, a separate initiative involving digital financial infrastructure.
The MERaLiON project, however, focuses on customer-facing work rather than payment settlement.
What happens next?
UOB and A*STAR said they would continue assessing how MERaLiON could be applied safely and responsibly across the bank’s regional operations.
The memorandum of understanding signed on Nov. 18 provides a framework for further exploration. The announcement does not give a confirmed commercial launch date or say that the system is already operating across all five markets.
The next meaningful milestones will be evidence of testing, validation, operational safeguards and any decision to expand the system into live customer interactions.
For customers, the important question will be whether the technology improves clarity and consistency without weakening privacy or human accountability. For the bank, the challenge will be demonstrating that the system works reliably across languages and financial contexts.
The broader AI conversation in banking also includes data protection. WWC has examined Singapore’s warnings about giving AI agents access to private accounts. UOB’s project is not the same as a personal AI agent, but both developments underline why access controls and clear responsibility matter when AI interacts with sensitive information.
Frequently asked questions
1. What is UOB’s MERaLiON AI project?
It is a proposed multilingual customer-call intelligence system intended to assist UOB relationship managers with customer conversations and post-call compliance reviews.
2. What is MERaLiON?
MERaLiON stands for Multimodal Empathetic Reasoning and Learning in One Network. It is an AI model developed by Singapore’s A*STAR.
3. Which countries are included in the UOB initiative?
The five markets identified in the announcement are Singapore, Malaysia, Thailand, Indonesia and Vietnam.
4. Will AI replace UOB relationship managers?
The announcement says relationship managers will remain accountable for their conversations and advice. The technology is intended to complement their professional judgement.
5. When will UOB launch the AI system?
No confirmed launch date was announced. UOB and A*STAR said they would continue exploring how the system could be developed and applied safely at scale.
