Hong Kong-based financial infrastructure company PhotonPay and Mastercard have completed a live operational demonstration of an agentic payment, in which an AI agent autonomously initiated and executed a real-world transaction on behalf of a user. The demonstration marks a step in the development of AI-driven payment infrastructure in Hong Kong.
In the demonstration, a PhotonPay tokenised card was provisioned for use by an AI agent, which independently identified, selected, and booked a ride through hoppa, a global mobility provider, completing the transaction within a test flow. The end-to-end process covered payment initiation, authentication, and settlement without direct human intervention at the point of execution.
Infrastructure and security framework
PhotonPay's programmable financial infrastructure provided the underlying foundation for the transaction, using real-time settlement capabilities and API orchestration to enable millisecond-level responsiveness to financial instructions. Mastercard contributed authentication and security mechanisms through Mastercard Agent Pay, enforcing user authorisation, identity verification, and data protection throughout the transaction process.
The global head of financial institutions at PhotonPay said the demonstration connected intelligent decision-making with reliable financial execution. Through close work with partners such as Mastercard, the company is enabling AI agents to operate with the same trust and assurance as conventional payment flows, while delivering the speed and flexibility that programmable money enables.
Development roadmap
PhotonPay has indicated it is deepening its investment in agentic payment infrastructure, with a focus on building a programmable financial layer integrated with global payment networks. The company's stated architecture combines real-time settlement, on-demand liquidity, and API-native orchestration, designed to allow AI agents to route transactions dynamically, optimise foreign exchange execution, and complete payments with reduced friction.
The company is also working with Mastercard to develop standards and trust frameworks for global agentic commerce, with the aim of bridging programmable, digital-native capabilities with the regulatory and security requirements of established financial infrastructure.
Industry context
The demonstration sits within a broader industry effort to define the infrastructure layer for agentic commerce, in which autonomous software agents initiate and complete commercial transactions without direct human involvement at each step. Several parallel initiatives are emerging across the payments sector, including protocol-level work on machine-to-machine payment standards and platform-level integrations that embed payment execution directly into AI agent workflows.
A key challenge for agentic payments is establishing trust and compliance frameworks that satisfy regulatory requirements while enabling the speed and autonomy that define the agentic model. The involvement of an established network such as Mastercard in providing authentication infrastructure addresses part of that challenge, though the broader regulatory and liability frameworks governing AI-initiated financial transactions remain at an early stage of development across most jurisdictions, including Hong Kong.
