Scaling AI-Native Fintech: Jiangming Yang (Ant International)
Chief Innovation Officer @ Ant International
Ant International’s Chief Innovation Officer, Jiangming Yang, is on a mission to redefine what AI-native financial services look like. From launching the Alipay+ GenAI Cockpit to deploying agentic AI across payments and compliance, he’s pushing the boundaries of how fintech can embed intelligence at scale - securely, reliably, and with real-world impact.
Q: With AI evolving so rapidly, what’s helped Ant stay ahead — not just in model innovation, but in turning that into real, high-impact financial use cases? What do you think sets Ant’s approach apart in such a fast-moving space?
Jiangming: Firstly, Ant International is a leading financial technology provider, which means that we have industry expertise and insights into what our partners want when it comes to financial AI.
A good example of how we’ve turned that expertise and insight into innovation is our Alipay+ GenAI Cockpit Platform, which allows fintech companies, banks, and digital wallet partners to build modular, customizable AI agents that they can easily integrate into their services. Our AI toolkit also means that we’ve built in financial rules such as dispute resolution and banking protocols. This AI-as-a-Service model lets our partners flexibly deploy AI and move towards AI-native commerce in a scalable manner while remaining compliant with financial regulations.
Q: You’ve described the future of finance as ‘agentic’, with AI agents running everything from onboarding to fraud checks. With the Alipay+ GenAI Cockpit now live, what are some of the biggest wins or breakthroughs you’ve seen in making AI genuinely useful and trusted in complex financial workflows? What are the main challenges?
Jiangming: When it comes to the finance industry, financial AI could in fact be considered the “holy grail” of AI because of the following challenges: the security stakes are high, especially given the rise of deepfakes and the risk of hallucination in AI models. On top of that, AI use-cases require rapid decision-making and action in specialized scenarios that tap on an up-to-date knowledge base and deep expertise.
For us, we’ve seen success with our Antom Copilot AI agent, which helps merchants implement, orchestrate and optimize their payment methods. Specifically, integration became 90% faster, with a 90% satisfaction rate in issue resolution. In fact, when our AI agent is used to handle service requests, the quality of answers matches or exceeds human experts in 70-80% of the cases. While we consider this a “win”, we are of course always striving to improve on our AI models to deliver even better experiences for our partners.
Q: The Cockpit brings together 20+ LLMs, Ant’s own FX prediction models, and rich fintech-specific knowledge. How do you think about the right balance between using general-purpose models and building tailored, domain-specific ones - and where have you seen the biggest payoff?
Jiangming: Whether to use a general-purpose model or a tailored, customized one ultimately depends on what the needs of a partner might be. That’s the great thing about Cockpit – we have offerings that can meet the needs of most firms, be it a startup or SME, to a more established financial institution.
For SMEs, they can get started quickly with pre-built agents without the need to build their own model from scratch, while larger firms may want a more customisable solution that better fit their workflow.
Ultimately, the key metrics of successfully deploying AI in fintech can be measured by two criteria – whether the AI exceeds human expertise to bring qualitative improvements in efficiency, security and reliability, as well as whether that AI is scalable and therefore more accessible to a wider base of clients.
Q: Ant International operates at the intersection of global scale and deep local nuance. How do you decide when and where to localize products, and when to export proven playbooks from China? What frameworks or signals guide you in balancing standardization with customization across such a diverse set of markets?
Jiangming: For our AI agents, there are no “proven playbooks” per se as every market is different, with varying financial regulations and requirements. As such, depending on the market, the agents powered by Cockpit are equipped with the relevant domain expertise and can be further optimized based on our partner’s specific and localized needs.
Take KYC (Know Your Customer) as an example – it’s a process that organizations use to verify the identity of their customers to prevent fraud or money laundering. KYC requirements vary by market. For instance, some markets rely on document checks while others require additional biometric data like facial recognition as part of the KYC process. At Ant International, our anti-deepfake electronic Know Your Customer (e-KYC) tool has an interception success rate of over 99%.