Markets
    Fintech

    Hong Kong Finance Firms' GenAI Use Hits 90% as 13% Deploy AI Agents, HKIMR Survey Finds

    The HKMA-linked research institute found banks, insurers and fund managers moving into agentic AI for fraud checks and compliance, while explainability and accountability worries hold them back.

    Share on X LinkedIn WhatsApp
    Entrance to Two International Finance Centre in Hong Kong's Central district
    Two International Finance Centre in Central, home to the Hong Kong Monetary Authority. Its research arm surveyed AI use at financial firms. Photo: LN9267 via Wikimedia Commons, CC BY-SA 4.0

    What to know

    • Current or planned GenAI adoption among Hong Kong financial institutions reached 90%, up from 75% in 2025, according to an HKIMR report released Nov. 17.
    • Thirteen percent of surveyed institutions have already implemented agentic AI and 56% plan to, while 54% cite explainability and operational risks as the top barriers.
    • In compliance, 86% have adopted or plan to adopt GenAI and 73% plan to raise spending on it; the survey covered 39 financial firms, including 16 banks holding 67% of deposits.

    HONG KONG

    Nine in 10 financial institutions in Hong Kong now use or plan to use generative artificial intelligence, up from 75% a year earlier, and 13% have already put AI agents to work, according to a survey published on Tuesday by the Hong Kong Institute for Monetary and Financial Research (HKIMR).

    The institute is the research arm of the Hong Kong Academy of Finance, which the Hong Kong Monetary Authority, the Securities and Futures Commission, the Insurance Authority and the Mandatory Provident Fund Schemes Authority set up together. Its report, titled "From Assistance to Autonomy," tracks how quickly the city's banks, insurers and fund managers are moving from tools that draft text and summarize documents to software that carries out multi-step tasks with little human input. Another 56% of the institutions surveyed said they plan to adopt such agentic AI, the report said.

    The Standard, which reported the findings the same day, cited the same headline figures. Enoch Fung, chief executive of the academy and executive director of the HKIMR, said the report showed "broader and deeper A.I. adoption across the financial services industry," according to the institute's statement.

    What the HKIMR survey found

    The jump came faster than the industry itself expected. In the institute's 2025 survey, firms had projected that adoption would reach 87% within three to five years. The 2026 figure of 90% already passes that mark, and only 10% of respondents said they had not put GenAI to institutional use, citing limits on resources and data.

    Most firms are not building alone. Two-thirds (67%) said they combine in-house work with outside vendors, while 20% develop their GenAI applications entirely in-house, the report said. None of the technology and regtech providers interviewed had built its own foundation model; they work on top of proprietary or open-source models instead.

    Agentic AI is going first into internal, process-heavy work: middle- and back-office operations, know-your-customer and anti-money-laundering checks, and risk management. Some 86% of market participants saw the ability to make decisions autonomously and raise efficiency as a key advantage. The worries are just as concrete. Among financial institutions, 54% named explainability and operational risks as the top challenges, and 49% pointed to accountability.

    Spending plans lean cautious. Over half of the firms (56%) intend to prioritize GenAI while exploring agents, 26% plan to balance investment between the two and 11% would put agentic AI first. On adoption, the remaining 31% of institutions said they had no immediate plans to deploy agents but were watching the market. Every service provider surveyed named multi-agent systems, in which specialized agents split up a complex job, as the top trend ahead.

    Digital banks and compliance teams move first

    The report singled out digital banks and insurers as quicker movers than their larger, traditional rivals. One digital-native respondent said it already runs agentic tools for real-time fraud alerts, loan document checks and debt collection, according to the HKIMR.

    Compliance is where adoption runs deepest. Some 86% of the financial institutions surveyed have adopted or plan to adopt GenAI for regulatory compliance, led by financial crime prevention and tracking regulatory updates. More than 70% said they want to test agents for customer onboarding, due diligence, transaction monitoring and fraud detection over the next three to five years, and 73% plan to increase their spending on GenAI for compliance.

    Banks report the most mature use. In the banking sector, 60% of respondents said they already use GenAI in compliance work, and the report linked that partly to better data availability and to longer engagement with regulators through programs such as the HKMA's GenA.I. Sandbox. The main obstacles firms listed were hallucination, bias and explainability, along with data privacy, cybersecurity, data availability and integration issues.

    The survey adds to other recent work by Hong Kong regulators on the topic, including the Securities and Futures Commission's survey of agentic AI use among brokers and the HKMA's Project SAGE.

    A small sample with heavy weight

    The sample is small but concentrated. The survey, run with KPMG Advisory (Hong Kong) from January to March 2026, drew 47 responses: 39 from the financial industry, including 16 banks, 11 insurers and 11 asset and wealth managers, plus eight technology and regtech providers. The 16 identifiable banks held 67% of Hong Kong's deposits at the end of 2024, and the 11 identifiable insurers wrote 45% of gross premiums, according to the report's appendix. The institute also held 19 interviews between January and July.

    The HKIMR set out five areas for policy work. It called for clear regulatory guidance, especially on explainability and accountability for frontier models, and for continued dialogue with the industry through the GenA.I. Sandbox++ and Project SAGE. It also urged privacy-by-design data practices, risk-based oversight, scenario testing and human checks suited to each use case, stronger operational resilience and a broader approach to training staff.

    On operational resilience, the report said scalable computing, supported by the government's planned Sandy Ridge Data Facility Cluster, would help firms detect and recover from AI-induced disruptions. The government plan cited in the report has the cluster coming into operation by 2029.

    Sources

    1. Hong Kong Monetary Authority (Hong Kong Academy of Finance and HKIMR release): Report on "From Assistance to Autonomy: A.I. Adoption, Governance, and Compliance in Hong Kong's Financial Services Industry" (Nov. 17, 2026)
    2. Hong Kong Institute for Monetary and Financial Research: From Assistance to Autonomy: AI Adoption, Governance, and Compliance in Hong Kong's Financial Services Industry (HKIMR Applied Research Report No.3/2026) (Nov. 17, 2026)
    3. The Standard: GenAI adoption in HK finance hits 90pc: report (Nov. 17, 2026)

    Capital Times Staff

    Capital Times newsroom, Hong Kong, Singapore and Jakarta

    Capital Times covers markets, business and economies across Asia.

    newsroom@capitaltimes.asia

    Our standards: Capital Times reports under the principles set out in our Editorial Standards and Ethics Policy. Spotted an error? Report it to our editors.

    Markets Data

    Hong Kong ClosedShanghai ClosedTokyo ClosedSeoul ClosedSingapore ClosedJakarta ClosedSydney ClosedMumbai Closed