In brief
Jim Munn, CIO of United Asia Finance, argues that consumer lenders in Hong Kong must fight AI-powered fraud with AI of their own. He describes a three-part strategy: a document-fraud checker, a customer-facing generative AI chatbot, and a real-time data lake. Each part supports the others, and he judges them by business outcomes rather than technology for its own sake.
The central tension is speed versus safety. Munn wants to use AI to move faster, to build software and answer customers, while rules are still being written and old systems resist change. His answer is steady governance: a cross-functional review group, continuous risk assessment, guardrails on what the chatbot may say, and hard budgets on AI usage costs.
Questions
What is the AI document defender at United Asia Finance?
It is a tool that uses AI to check the documents loan applicants submit, helping staff spot fake ones more consistently than human eyes alone. The host says it cuts loan verification from hours to seconds, and Jim Munn says it lowers the risk of fraudsters becoming customers.
How does the company's generative AI chatbot avoid giving wrong or off-topic answers?
Munn says it is restricted to United Asia Finance products and set up to refuse other topics. It also reads the company's product documents, so it knows which promotions are current and does not offer expired ones.
Why does a real-time data lake matter for a lender?
Munn says older data warehouses use data that is a day old, which is too slow for instant lending decisions. A real-time data lake combines large amounts of fresh data, so the company can price loans better, check that customers can repay, and send personalised offers as events happen.
How is the company handling AI rules that are still being written?
Munn says it built its own AI governance framework about a year ago, covering ethics, personal data, compliance and cybersecurity. A mixed IT and business group reviews each AI use, and risks are reassessed continuously and escalated when they exceed the company's risk appetite.
How does Munn justify the cost of AI projects to the board?
He says return is not only money: brand awareness, operational efficiency and cybersecurity count too. He also says AI token costs must be budgeted, with a backup plan if the budget runs out and the chatbot stops answering.
Facts
Stated by the speakers. Each quote matches the transcript, which shows what was said, not that it is true.
Deepfake fraud incidents in Hong Kong's financial sector rose sharply, culminating in a US$25 million incident in early 2024, according to the host.
“deep-fake fraud incidents in Hong Kong's financial sector has surged dramatically in recent years, culminating in the shocking US $25 million incident in early 2024”
0:03:15 · Estelle Quagg · Deepfake fraud · 2024
According to the host, Jim Munn delivered three major innovations in two years with an IT team of 100 at United Asia Finance.
“In just two years, with a dedicated IT team of 100, Jim has delivered three transformative innovations”
0:00:41 · Estelle Quagg · IT delivery
The host describes United Asia Finance's three innovations as an AI document defender that cuts loan verification from hours to seconds, Hong Kong's first generative AI chatbot in financial services, and a real-time data lake.
“an AI document defender that cuts loan verification from hours to seconds, Hong Kong's first generative AI chatbot in financial services, and a real-time data lake”
0:00:41 · Estelle Quagg · Three innovations
Jim Munn started his banking career more than 20 years ago at UBS.
“I started my banking career 20 plus years ago in UBS and has been on a journey with UBS and front office international.”
0:01:44 · Jim Munn · Career
Jim Munn has been Chief Information Officer of United Asia Finance for four years.
“and also move on to be our lending business right now and has been in UA Finance for four years as CIO.”
0:01:44 · Jim Munn · Career
Hong Kong's government released its policy statement on responsible AI in financial markets in October 2024, according to the host.
“Hong Kong's government only released its policy statement on responsible AI in financial markets in October 2024”
0:06:44 · Estelle Quagg · AI regulation · 2024
United Asia Finance built an AI governance framework about a year before the interview, covering technology and cybersecurity as well as ethics, personal data and compliance.
“we have built AI governance framework here one year ago and this framework will cover not just technology, cybersecurity perspective”
0:07:18 · Jim Munn · AI governance
United Asia Finance uses a working group of IT staff, business users and subject-matter experts to review its AI use cases periodically.
“And we have a working group, not just IT, but also business users, SME, to review all these AI use cases from time to time.”
0:07:56 · Jim Munn · AI governance
Jim Munn says building an application skeleton once took days, whereas his team now prompts AI agents to create the code.
“In old days, there were days to build a skeleton of an application. And today we just use a lot of prompts as a genetic solution to just ask the agent to create their code”
0:04:48 · Jim Munn · AI coding
According to the host, a Hong Kong Monetary Authority FinTech 2030 finding is that integrating AI with legacy infrastructure is among the most resource-intensive barriers to innovation.
“One of HKMA's own FinTech 2030 findings is that integrating new AI solutions with legacy infrastructure remains one of the most resource-intensive barriers to innovation”
0:09:22 · Estelle Quagg · Legacy systems
Jim Munn says some legacy components at United Asia Finance are not scalable or safe enough to interact with AI.
“Not every legacy system is compatible with AI. In my case, there are some legacy systems components that is not really scalable and safe to interact with the AI.”
0:09:50 · Jim Munn · Legacy systems
Jim Munn says United Asia Finance modernises legacy components to be cloud-native and API-friendly so AI can interact with them.
“But we have to modernize different types of legacy components to be cloud-native, API-friendly, become a friend of components to interact with the AI.”
0:10:26 · Jim Munn · Modernisation
The host says close to 90 percent of financial institutions globally cite data management problems as their biggest AI adoption hurdle.
“close to 90 % of financial institutions globally cite data management, fragmented sources, quality issues, and regulatory constraints as their biggest AI adoption hurdle”
0:11:55 · Estelle Quagg · Data management
Jim Munn says a data warehouse using end-of-day data is not helpful for instant decisions in fast-paced lending, which a real-time data lake addresses.
“we're talking about data warehouse that takes end of day, end minus one data. It will not be very helpful to make instant decision in the fast-paced lending industry.”
0:12:28 · Jim Munn · Data lake
Jim Munn says United Asia Finance uses a data lake to combine large data sets for underwriting rather than relying on one or a few data points.
“Having a data lake to mash all this data together, not just making a bit underwriting decision, not just one or a few data points”
0:16:40 · Jim Munn · Underwriting
The host cites surveys showing only about one in three Hong Kong users consider a digital bank their primary account.
“recent surveys now show only around one in three Hong Kong users consider a digital bank their primary account”
0:13:47 · Estelle Quagg · Digital banking
Jim Munn says digital banks have higher agility in launching new products and testing new interfaces and gamification.
“have a higher agility to roll out this new product and have testing tasting new user interface experience gamifications”
0:15:09 · Jim Munn · Digital banks
The host says Jim Munn's CIO 100 Awards 2025 submission describes three interconnected innovations he led.
“I gathered from your CIO 100 Awards 2025 submission that you led three interconnected innovations.”
0:15:32 · Estelle Quagg · CIO 100 Awards · 2025
Jim Munn says the AI document defender identifies fraudulent documents more consistently than human checking and lowers the risk of fraudsters becoming customers.
“there will be more consistent way to identify fraud documents and help the operator, which is efficiency of our staff, minimizing the risk of having a fraudster in becoming our customers”
0:16:40 · Jim Munn · Document defender
Jim Munn says United Asia Finance's chatbot uses retrieval-augmented generation (RAG) over its product information to answer customer inquiries about the latest offers.
“we've generally produced all those RAG to understand our existing product offering and can answer the customer this kind of general inquiry, the latest offer inquiry immediately.”
0:17:26 · Jim Munn · Chatbot
Jim Munn says the chatbot is meant to avoid presenting expired time-limited promotions to customers.
“For example, our offer is a time-limited offer. And if a customer is asking an offer, we won't send them an already expired campaign.”
0:18:13 · Jim Munn · Chatbot
Jim Munn says a key challenge was setting guardrails so the chatbot answers only about United Asia Finance products.
“One of the key points is I want the board to focus on answering only our products. We have to pay certain guardrails”
0:19:51 · Jim Munn · Chatbot guardrails
The host says the innovations were reported to have reduced fraud attempts and call center workloads and improved customer satisfaction.
“Fraud attempts markedly decreased, call center workloads reduced and customer satisfaction improved.”
0:20:30 · Estelle Quagg · Outcomes
Jim Munn says AI token usage should have a budget and a reaction plan for when it is exceeded.
“If you should have a budget on these AI tokens and exceeding, you have to have a reaction plan”
0:21:40 · Jim Munn · AI costs