When Hong Kong’s Banks Use AI to Fight AI: The New Frontier of Customer and Employee Experience
Picture this: A fraudster creates a deepfake video of your bank’s CEO, instructs an employee to wire millions, and disappears before anyone realizes the deception. Sounds like science fiction? It happened in Hong Kong in January 2024, costing one firm $25 million. This isn’t a distant threat anymore—it’s today’s reality, and it’s forcing financial institutions to fundamentally rethink how they protect both customers and employees.
The Hong Kong Monetary Authority just announced something remarkable. On October 15, 2025, they unveiled the second cohort of their Generative AI Sandbox, bringing together 27 use cases from 20 banks and 14 technology partners. Among them, Ant Digital Technologies stands out not just as a participant but as a technology partner helping reshape how banks in Asia’s financial hub approach artificial intelligence. But here’s what makes this different: these aren’t simple automation projects. They’re pioneering “AI vs. AI” strategies—using artificial intelligence to check the work of other AI systems, creating layers of defense against increasingly sophisticated fraud.
This marks a turning point. Financial institutions are moving from asking “Should we use AI?” to “How do we use AI responsibly while defending against AI-driven threats?” The implications for customer experience and employee experience are profound.
The Deepfake Crisis Hitting Financial Services
Deepfake fraud surged 3,000% in 2023 alone. Banks now face criminals who can clone voices with just 15 seconds of audio, create convincing video impersonations, and bypass traditional security measures designed for a pre-AI world. The average cost of deepfake-related fraud reached nearly $500,000 per incident in 2024, with large enterprises experiencing losses exceeding $680,000.
Hong Kong police intercepted a fraud network in April 2025 that used AI to merge scammers’ faces onto stolen IDs, enabling them to open bank accounts. Losses exceeded $193 million. According to research firm Proofpoint, 99% of customer organizations monitored in 2024 were targeted for account takeovers, with 62% experiencing at least one successful attack.
The threat isn’t theoretical anymore. It’s systematic, scalable, and getting worse. Financial institutions must respond with equal sophistication—which is exactly what the HKMA’s GenAI Sandbox enables.
AI vs. AI: A Governance Revolution
Traditional quality control meant humans checking human work. Then came AI checking human work. Now we’ve entered a new phase: AI checking AI work. This “AI vs. AI” approach represents a fundamental shift in governance strategy.
Here’s how it works in practice. When an AI system generates a customer service response, approves a transaction, or flags potential fraud, another AI system immediately audits that decision. It checks for accuracy, consistency, bias, and adherence to compliance requirements. The second AI acts as a quality control layer, catching errors before they reach customers or regulators.
Arthur Yuen, Deputy Chief Executive of the HKMA, describes the second cohort as “a major leap forward in making AI adoption safer and more robust.” The Sandbox serves as a testing ground where banks can conduct adversarial simulations—using AI to stress-test and fortify systems against sophisticated digital fraud. Several participants are exploring how defensive AI can detect deepfake attempts during video verification, voice authentication, and identity checks.
This layered approach delivers benefits across the entire customer and employee journey.
Transforming Customer Experience with Responsible AI
Ant Digital Technologies brings three key platforms to Hong Kong’s financial institutions, each addressing different aspects of the customer experience challenge.
Agentar AI Agent Platform represents the company’s full-stack, finance-grade development environment. Banks using this platform with Ant’s financial large language models achieve performance improvements exceeding 10% in critical areas like campaign planning and customer engagement. The platform enables institutions to move beyond basic chatbots to intelligent agents that reshape core operational functions.
For Fubon Bank (Hong Kong), Ant Digital Technologies is leveraging Agentar alongside its mPaaS LUI module to explore an AI-powered assistant designed for personalized, secure, and interactive mobile banking. This isn’t about automating simple tasks—it’s about creating experiences that understand context, anticipate needs, and adapt to individual customer behaviors.
Mobile banking personalization has become essential, not optional. Research shows 53% of consumers expect their financial provider to leverage data to personalize their experience, while 76% are more likely to choose a bank offering personalization. But personalization must balance effectiveness with trust. Customers increasingly demand transparency about how AI makes decisions affecting their finances.
Modern AI-driven personalization analyzes transaction patterns, behaviors, and preferences to anticipate needs and offer targeted solutions at critical moments. Banks can provide real-time, context-aware guidance—suggesting budget adjustments after certain purchases, alerting customers to potential overdrafts before they occur, or recommending savings strategies based on spending patterns.
The key difference? These interactions feel helpful rather than invasive because they’re built on transparent governance frameworks.
Defending Digital Identity in a Deepfake World
Identity fraud represents one of the most pressing challenges in financial services. Traditional KYC processes relied on facial recognition and voice verification, but deepfake technology has challenged both methods. Banks now need multi-layered defense systems.
Ant Digital Technologies’ ZOLOZ platform addresses this challenge head-on. The company collaborates with partners to explore how ZOLOZ AI solutions detect and prevent identity fraud using GPU-accelerated infrastructure for both model training and deployment. The goal: building a scalable defense layer for financial institutions navigating increasingly complex digital identity challenges.
ZOLOZ uses biometric verification, real-time identity checks, and AI-driven fraud detection to enhance KYC compliance. Unlike simple systems comparing a user’s photograph to their ID, ZOLOZ includes sophisticated measures detecting fake and spoofed IDs. It reads information from ID cards via optical character recognition, spots signs of fake IDs, and requires users to move and blink rather than submit still photos—making fraud significantly more difficult.
The platform’s real strength lies in its historical data cache. By comparing verification attempts with IDs, user photos, and transaction history it has seen before, ZOLOZ detects fakes more quickly and accurately. Changes to hairstyle, makeup, or environment aren’t enough to fool the system.
In 2024, ZOLOZ launched ZOLOZ Deeper, specifically designed to combat AI face-swapping attacks on facial recognition systems. The solution uses mobile sensor verification, multi-dimensional dynamic risk control, and multimodal deep learning models to address AI security threats across system, application, and server layers. Its algorithm is trained on 300,000 test samples and constantly undergoes offensive and defensive evaluations.
For CX professionals, this technology isn’t just about security—it’s about preserving trust. When customers know their bank uses advanced AI to protect their identity, they feel more confident conducting digital transactions. The experience becomes smoother because legitimate customers face fewer friction points while fraudsters encounter insurmountable barriers.
Elevating Employee Experience Through AI Augmentation
While customer-facing AI gets most attention, employee experience transformation may deliver even greater impact. Banks implementing AI virtual assistants handle 40% of employee IT inquiries with 67% faster resolution times. Institutions deploying comprehensive AI workplace platforms report 25% improvements in employee satisfaction.
Financial institutions increasingly recognize that AI should augment employee capabilities rather than replace human judgment. Bank employees using AI features on mobile devices can summarize or organize notes from important client meetings, quickly retrieve product information, conduct research, or answer complex customer questions in real time.
The Agentar platform enables this augmentation at scale. Ant Digital Technologies’ risk control agent can autonomously generate risk models from business data and expert knowledge, producing results surpassing experienced professionals and improving modeling effectiveness by 10%. The marketing agent enables intelligent upgrades to retail banking operations, enhancing marketing performance and operational efficiency by as much as 20%.
This represents a fundamental shift in how financial institutions approach employee productivity. Rather than viewing AI as a cost-cutting tool for eliminating positions, forward-thinking banks use it to elevate their workforce—freeing employees from routine tasks so they can focus on complex problem-solving, relationship building, and strategic thinking.
Employee experience and customer experience are intrinsically linked. When bank employees have powerful AI tools supporting their work, they deliver better customer service. They respond faster, provide more accurate information, and handle complex situations with greater confidence. The technology removes frustration from their daily work, reducing burnout and improving retention.
The Broader Hong Kong Innovation Ecosystem
Ant Digital Technologies’ participation in the HKMA GenAI Sandbox reflects a broader commitment to Hong Kong’s fintech ecosystem. The company established its international headquarters in Hong Kong in April 2025, joining the Office for Attracting Strategic Enterprises as a Key Enterprise Partner.
With a global network of over 300 partners and more than 10,000 enterprise customers served, Ant Digital Technologies has scaled its AI solutions to support digital transformation throughout the Guangdong-Hong Kong-Macao Greater Bay Area. The company actively participates in the HKMA’s Project Ensemble, which explores wholesale central bank digital currency and tokenization use cases.
Project Ensemble represents another dimension of Hong Kong’s innovation ambitions—creating infrastructure for seamless interbank settlement of tokenized deposits and assets. Ant Digital Technologies joins other Architecture Community members including HSBC, Standard Chartered, Bank of China (Hong Kong), and Hang Seng Bank in developing standards supporting interoperability among wholesale CBDC, tokenized money, and tokenized assets.
This collaborative approach—bringing together regulators, banks, technology companies, and academic institutions—creates an environment where innovation can flourish under appropriate oversight. The HKMA’s Sandbox model allows participants to experiment with cutting-edge technologies while receiving supervisory and technical guidance throughout trial processes.
Hong Kong is positioning itself as a global hub for responsible AI innovation in financial services. AI-driven fintech investment surged 68% in the first quarter of 2025, defying global downturns, with AI projects claiming over 50% of the city’s HK$3.1 billion in fintech investments. The HKMA has prioritized generative AI in its “FinTech Strategic Plan 2025-2030,” backing innovation with HK$500+ million in support.
Navigating Implementation Challenges
Despite promising developments, financial institutions face significant challenges implementing AI responsibly. Security and data privacy concerns rank as the primary obstacles, with 39% of banks identifying them as significant issues. Additionally, 33% of respondents highlight a lack of AI skills or expertise within their workforce, and 30% cite difficulties measuring return on investment.
Transparency and explainability emerge as critical success factors. Customers and regulators increasingly demand understanding of how AI systems make decisions affecting their finances. Black-box models that provide outputs without insight into their reasoning create compliance risks and erode trust.
Explainable AI addresses this challenge by providing insights into AI decision-making processes. Financial institutions using XAI can ensure fairness, accountability, and regulatory adherence. By making AI reasoning visible, banks can diagnose model behavior, fix errors, retrain on edge cases, and continuously improve outcomes.
Data quality and bias present ongoing challenges. Inconsistencies or errors in input data lead to flawed predictions and suboptimal outcomes. Outdated data results in misguided decisions. Financial institutions must implement rigorous data validation processes and cleaning mechanisms to ensure reliability.
Algorithmic bias poses both ethical and compliance risks. AI systems trained on historical data can perpetuate existing biases, affecting marginalized groups unfairly. Banks must actively address potential discrimination through rigorous testing, monitoring, and implementing techniques like resampling, reweighting, or algorithmic fairness constraints.
Talent gaps slow adoption. Organizations need professionals who understand both finance and AI—a rare combination. Banks should invest in targeted training programs, partnerships with universities, and strategic use of platforms that simplify AI deployment through low-code visual orchestration.
Best Practices for CX and EX Leaders
Based on emerging patterns from Hong Kong and global financial institutions, several actionable strategies can guide CX and EX professionals through AI adoption.
Start with high-impact use cases. Identify business processes that can be automated or augmented with AI and will deliver measurable impact. Focus on high-volume, low-complexity inquiries first to achieve quick ROI. For Ant Digital Technologies’ banking partners, this means beginning with identity verification, fraud detection, and personalized banking assistance before tackling more complex challenges.
Build governance frameworks from day one. Establish comprehensive AI governance structures encompassing ethical guidelines, compliance checks, and continuous monitoring. Create cross-functional committees with stakeholders from risk, compliance, IT, legal, and business units. The “AI vs. AI” approach demonstrated in Hong Kong’s Sandbox provides a model for building accountability into systems.
Prioritize transparency and education. Communicate openly about AI ethics policies and data privacy practices. Implement explainability features that detail how AI analyzes customer information to make recommendations. Invest in educational initiatives—workshops, webinars, and online resources—that empower customers and employees with knowledge about AI’s role, benefits, and safeguards.
Design for seamless human-AI orchestration. Create clear escalation rules and ensure context preservation during handoffs between AI and human agents. The goal isn’t replacing human judgment but augmenting it. Employees should view AI as a tool that makes their work more effective and satisfying.
Adopt a tiered, risk-based approach. Categorize AI models by risk level and apply stricter oversight to high-impact areas like credit scoring, fraud detection, and identity verification. Use lighter approaches for lower-risk applications. This maintains control while allowing flexibility for innovation.
Leverage sandbox environments. Test AI solutions in controlled environments before full deployment, ensuring they meet compliance standards and deliver intended outcomes. The HKMA’s GenAI Sandbox demonstrates how regulatory authorities can facilitate responsible innovation through structured experimentation.
Monitor, measure, and iterate continuously. Establish monitoring processes assessing AI accuracy across common and obscure scenarios. Track metrics including customer satisfaction, employee productivity, fraud detection rates, and operational efficiency. Use insights to refine models and expand successful use cases.
The Path Forward: Building Trust at Scale
The convergence of AI opportunities and AI threats creates an inflection point for financial services. Institutions that respond with sophisticated, responsible AI strategies will differentiate themselves through superior customer and employee experiences. Those that delay or implement AI carelessly risk falling behind competitors and exposing themselves to sophisticated fraud.
Hong Kong’s GenAI Sandbox represents more than a regulatory initiative—it’s a blueprint for collaborative innovation under appropriate oversight. By bringing together diverse stakeholders to tackle shared challenges, the HKMA fosters an environment where breakthrough solutions can emerge.
Ant Digital Technologies’ participation demonstrates how technology companies can partner with financial institutions to accelerate responsible AI adoption. The company’s platforms—Agentar, ZOLOZ, and mPaaS—address critical pain points across customer identity, personalized experiences, and operational efficiency. More importantly, they do so within governance frameworks emphasizing transparency, security, and continuous improvement.
For CX and EX professionals, this moment demands both ambition and humility. Ambition to reimagine what’s possible when AI augments human capabilities. Humility to recognize that AI introduces new risks requiring careful management. The financial institutions that thrive will be those that view AI not as a magic solution but as a powerful tool requiring thoughtful implementation, robust governance, and continuous refinement.
The question isn’t whether AI will transform banking—it already has. The question is whether financial institutions will transform responsibly, building systems that protect customers and employees while delivering genuinely better experiences. Hong Kong’s “AI vs. AI” approach offers a promising path forward, demonstrating that the best defense against AI-driven threats may be AI-driven governance. The future of financial services belongs to institutions that embrace this complexity and build trust at scale.
