This article is part of the monthly AI Radar series, providing a recap of innovative AI deployments and related company announcements in telecommunications, financial services, and customer experience fields of practice. It also offers insights into companies deploying AI, aimed at assisting business executives and technology leaders develop their own AI projects and long-term strategies.

In boardrooms across Asia-Pacific, one phrase keeps echoing louder with every passing quarter: “We need to do more with AI.” And now, banks aren’t just listening—they’re building. From biometric kiosks to decisioning engines that think in milliseconds, the region’s financial institutions are stepping into an era where AI isn’t a department—it’s the operating system.

National Australia Bank’s Customer Brain is transforming personalization into a real-time capability—adapting to customer actions as they happen and orchestrating tailored engagements across every touchpoint. In Thailand, ttb’s Yindee assistant is making digital banking more human, responding to natural Thai speech and serving users around the clock, while quietly reducing operational strain behind the scenes.

At Shinhan Bank in Korea, lifelike AI bankers now greet customers at branches—smiling, conversing, and assisting with over-the-counter services—blurring the line between digital and physical banking. CTBC Bank in Taiwan has taken the internal route, building its own generative AI assistant to securely handle frontline support, reinforcing the power of in-house AI talent. Meanwhile, Bangkok Bank has doubled down on localization, deploying a Thai NLP engine that powers omnichannel conversations and sets the foundation for a future-ready virtual bank.

This report captures the most forward-thinking AI deployments by banks across APAC—not as a trendwatcher, but as a radar. A radar scanning for the institutions not just experimenting with AI, but operationalizing it. Those redefining what a bank should look like in an AI-native world.

Axis Bank – Axis AHA Virtual Assistant

AHA (Axis AHA Assistant) is Axis Bank’s AI-first virtual banking assistant, designed to deliver secure, conversational experiences across Axis Mobile app, WhatsApp, and web platforms. Built using Google Dialogflow CX and integrated with the bank’s customer data layer, AHA is capable of resolving 1 million+ queries monthly, spanning services like fund transfers, card controls, loan queries, and investment advice.

At the technical core, AHA leverages a modular NLU engine with intent detection, entity extraction, and session memory management. The assistant communicates with the bank’s internal systems through API gateways, interfacing with over 300 services for real-time balance fetch, payment processing, and onboarding flows. Dialog orchestration is further enhanced by contextual response conditioning, allowing AHA to adapt based on user history and transaction profile.

To elevate customer experience, AHA incorporates voice and text interactivity, powered by Google Cloud’s Text-to-Speech (TTS) and Speech-to-Text (STT) APIs. It also supports multilingual conversations in English and Hindi with expansion planned for regional Indian languages. Embedded within its architecture is a fallback AI escalation layer—automatically redirecting unresolved intents to live agents via in-app chat or call center integration.

The chatbot is governed by Axis Bank’s internal AI Ethics Framework, including bias testing, consent-driven personalization, and continuous human-in-the-loop validation. Its performance is tracked via detailed telemetry—covering confidence scores, conversion funnels, and NPS impact—allowing for weekly retraining using customer interaction logs and synthetic data generation for edge cases.

CTBC Bank – IntelliChat Virtual Assistant

CTBC Bank’s IntelliChat represents a hybrid GenAI–LLM deployment built to modernize its customer service ecosystem. Designed in partnership with Microsoft Azure OpenAI, the assistant processes over 120,000 monthly customer queries with a resolution accuracy exceeding 80%, delivering support across retail banking, wealth management, and credit services.

IntelliChat is architected as a dual-layer NLP system—combining a retrieval-based chatbot for FAQs with a generative layer powered by GPT-4 for contextual and intent-sensitive queries. CTBC integrates IntelliChat into its mobile and web platforms via Azure Bot Services and leverages Azure Cognitive Search for real-time document lookup across internal bank policy libraries and product catalogues.

What sets IntelliChat apart is its enterprise-grade integration: it’s connected to CTBC’s customer data platform (CDP), enabling it to access personalized information (e.g., recent transactions, eligibility checks) securely. The assistant also includes built-in guardrails—using moderation filters, prompt chaining, and knowledge grounding to ensure compliance with internal governance and Taiwan’s financial regulatory standards.

Internally, the chatbot is monitored through a dedicated AI Ops dashboard, which logs response time, user satisfaction, and model confidence scores. This feedback loop feeds into retraining workflows, allowing CTBC to continuously evolve IntelliChat’s performance using both supervised fine-tuning and reinforcement learning from human feedback (RLHF).

TMB Thanachart Bank – Yindee Virtual Assistant

Yindee is Thanachart Bank’s AI-powered digital assistant, deployed across the ttb Touch mobile banking app, web portal, and LINE. It’s engineered using Microsoft’s Azure OpenAI Service, allowing Yindee to combine conversational intelligence with customer account-level precision. The assistant handles hundreds of thousands of queries each month, covering everything from transaction histories to loan status and personalized finance tips.

Yindee operates using a hybrid NLU architecture—intent classification powered by custom-trained Thai language models, fused with a generative LLM layer that adapts responses based on customer context. Thanachart built Yindee on top of a microservice-based orchestration layer, which routes user intents to secure backend APIs. The system supports real-time session memory, enabling it to carry forward context across queries for smoother multistep support.

Technically, Yindee includes security-first integrations—using OAuth 2.0 protocols, multi-factor authentication, and device fingerprinting. The assistant also features embedded smart nudges driven by ttb’s internal analytics engine, which feeds in behavioural and transaction-based triggers to suggest actions like top-up, bill alerts, or investment opportunities. A fallback mechanism also ensures that if the AI model cannot process a complex request, it transitions the user seamlessly to a human agent via secure live chat.

The bank’s AI team monitors performance through a closed-loop monitoring system, which captures user drop-off points, satisfaction rates, and misclassification logs—helping refine both the language models and the backend integration pathways continuously

Shinhan Bank – AI Banker Kiosks

Shinhan Bank’s AI Banker Kiosks are powered by DeepBrain AI’s generative avatar engine, bringing hyper-realistic digital tellers into physical branches. Each avatar is modeled after real Shinhan employees and synthesized using GAN-based video synthesis, layered with real-time voice recognition, facial animation, and emotion detection. These kiosks are deployed under Shinhan’s “AI Branch” initiative and currently operate across 500+ locations in South Korea.

The system leverages multimodal AI—merging speech-to-text, natural language understanding, and video response generation. Tasks like account opening, credit loan execution, and certificate issuance are handled by secure integrations with Shinhan’s core banking APIs. Kiosks are embedded with biometric verification, NFC readers, and facial recognition modules that sync with Shinhan’s enterprise identity systems to authenticate users.

Built on a scalable LLM backend, the platform supports over 64–70 automated banking services via a hybrid architecture combining on-prem kiosk processing with secure cloud orchestration. The backend includes an AI model orchestration layer that manages dialogue state, regulatory compliance filters, and escalation triggers. Future upgrades include real-time multilingual translation and integration with advisory services like investment planning or risk profiling—positioning the kiosks as 24/7 intelligent branch extensions.

Bangkok Bank – BB AssistMe

BB AssistMe is Bangkok Bank’s flagship Thai-language conversational AI assistant, built on the proprietary TT01 NLP engine—a deep learning model specifically trained on Thai syntax, semantics, and financial terminology. Developed in collaboration with Singapore-based Pand.ai, TT01 achieves over 96% intent recognition accuracy even in complex and idiomatic Thai, making it a rare example of localized language AI at scale.

Technically, BB AssistMe integrates across LINE, Bualuang mobile banking app, and desktop platforms via secure REST APIs. The backend operates on a hybrid cloud infrastructure and leverages intent orchestration engines to map user queries to backend workflows like account summaries, bill payments, or fund transfers. The assistant uses contextual memory to follow multi-turn conversations and validate identity via OTP and device fingerprinting.

The AI is built to be modular—each banking service (e.g., loan inquiry, credit card limit increase) is developed as a microservice, allowing BB AssistMe to scale features independently. Moreover, Bangkok Bank’s InnoHub team has developed an internal NLP analytics dashboard that tracks drop-offs, false positives, and customer sentiment in real time—ensuring continuous optimization of the assistant’s performance.

Thinking in Real Time: Inside NAB’s Customer Brain

National Australia Bank (NAB) has built a real-time AI decision engine called Customer Brain, which now sits at the heart of its personalization strategy. Launched in 2023, Customer Brain processes over 50 million customer decisions per month using 2,000 AI models and more than 1,200 data points. It is designed to deliver contextual nudges, timely product offers, and service notifications—automatically triggered based on real-time behaviour and engagement patterns.

The system continuously analyses customer actions—whether online, in-app, or via call centre’s—and tailors the next best action across channels. This has led to a 40% increase in engagement through hyper-personalized messages that are delivered at the right moment and context. Customer Brain also enables dynamic targeting in campaigns, significantly reducing noise and improving conversion across NAB’s digital estate.

This deployment shows how real-time decisioning engines powered by AI can shift personalization from static segmentation to adaptive, micro-level engagement. NAB’s Customer Brain redefines how banks can build loyalty, improve cross-sell, and drive relevance in every interaction—making personalization an always-on engine rather than a one-time campaign.

Rewiring Banking with AI

What unites these banks isn’t just their investment in AI—it’s their strategic intent to make AI core to their customer promise, not just a digital accessory. Whether through real-time personalization, generative AI assistants, or avatar-powered branches, these deployments signal a decisive shift: AI is now the foundation of competitive advantage in banking. The next frontier isn’t about who’s using AI—it’s about who’s using it best to reimagine products, experiences, and operating models. As this AI radar reveals, the banks that embed intelligence into every layer of their organization will be the ones to define the future of finance in Asia-Pacific—and beyond.