Octopus is integrating new capabilities with the existing architecture of a transport and retail payments platform used by millions across Hong Kong, while maintaining the reliability of its services.
In an interview with Frontier Enterprise, Tim Ying, Chief Executive Officer of Octopus, described how the company is gradually expanding its use of AI and the platform itself while maintaining control over how changes are introduced and who remains accountable for their outcomes.
How is AI being applied across the Octopus ecosystem today?
Octopus has operated across Hong Kong for nearly three decades, so we do not treat AI as a standalone technology project or adopt it for its own sake. We introduce it selectively to support what already works while maintaining the reliability and simplicity that millions of people in Hong Kong depend on every day.
We use AI to strengthen network resilience and improve the accuracy and efficiency of transaction processing. We also use it to make customer services easier to use and more relevant over time. Beyond routine automation, AI-enabled workflows give our teams more capacity to experiment, develop services, and expand new capabilities while maintaining the operational controls required of critical payment infrastructure.
As these capabilities develop, including more agentic forms of AI, governance remains central to how we use them. Greater autonomy requires clear operating limits, appropriate human oversight, and accountability for outcomes.
What types of customer or operational data have become more valuable because of AI?
The most useful insights increasingly come from linking interactions across channels, such as a digital wallet transaction and a contactless tap at a turnstile. Viewed as transitions people make throughout the day, these interactions provide a more complete picture of customer needs and operational patterns.
Octopus processes interactions across Hong Kong’s physical and digital economies, including transport, retail, mobile applications, online services, and cross-boundary activity. AI allows us to identify patterns across interactions that were previously considered separately. Instead of analysing only individual transactions, we can examine customer journeys and use those findings to make services more timely, relevant, and convenient. The same analysis can inform operational planning, improve service responsiveness, and help us anticipate changes in demand.
The aim is not to identify individual behaviour, but to interpret aggregated patterns. This analysis must operate within privacy, security, and governance requirements, with safeguards incorporated into the design and deployment of AI from the outset. Individual transactions show where people are, while journeys can indicate where their needs may be heading.
Which legacy systems create the greatest constraints on innovation?
Our main constraint is not a particular legacy system, but the risk involved in modernising a platform that already operates reliably at scale and is used throughout the daily life of a city.
Many organisations modernise systems that serve a defined group of users. Octopus is modernising infrastructure used by millions of people every day, which affects how we approach each stage of the process. Its existing architecture was designed for reliability. Rather than replacing it outright, we are integrating cloud-native platforms, open APIs, and AI capabilities with that infrastructure while maintaining the reliability that users expect.
Changes are required both in customer-facing services and in the core infrastructure, where errors can have wider consequences. This requires incremental renewal, rigorous testing, and carefully managed integration rather than wholesale change. Each upgrade must improve the platform without disrupting the services people rely on.
How do you balance automation with human oversight in customer-facing services?
AI can accelerate decisions, but accountability cannot be outsourced.
AI is useful for processing information quickly, operating at scale, and recognising patterns. Human oversight remains essential for judgement, accountability, and empathy, particularly in complex, sensitive, or unexpected circumstances that an automated system may not anticipate.
We use automation where it can improve consistency, responsiveness, and convenience. This involves setting limits on what automated systems can do, monitoring their performance, and referring matters to a person when individual judgement or support is required.
What technology investments became necessary as Octopus evolved beyond transit payments?
We invested in making our technology and operating model more adaptable rather than concentrating on a single system or platform.
As Octopus moved beyond transit payments, we invested in hybrid cloud infrastructure, API-driven platforms, and a more developer-focused way of working. These changes allow us to integrate with a broader range of partners, develop services more efficiently, and respond to changing customer expectations.
The aim was to develop the payments network into a platform that could support a broader range of services. These investments have enabled Octopus to expand into areas such as ride-hailing and travel beyond Hong Kong while maintaining the reliability and ease of use of its existing services. They also allow us to introduce capabilities gradually, with the control and operational visibility required of payment infrastructure.
These investments complement the Octopus card rather than replace it.
