EX.IO Lab used a bank sandbox at Hong Kong FinTech Week on Nov. 3 to demonstrate an AI agent initiating a bank transfer and completing an investment order for a real-world asset product on the EX.IO platform, offering a practical example of how agent-led payments could connect banking rails with tokenised investment products.
The proof of concept combined product analysis, account funding and order execution in one workflow. Rather than requiring a user to move between a banking application and an investment platform, the AI agent handled the sequence inside the sandbox environment, ending with a subscription to an RWA product available through EX.IO.
The demonstration was a controlled technical test, not a commercial rollout, public product release or regulatory approval for autonomous investment activity. Yet it moves the discussion around AI agents beyond chat-based financial assistance and toward systems that can execute a defined transaction after receiving the required permissions.
A bank-to-RWA transaction in a single workflow
EX.IO described the test as an agentic payment proof of concept. The agent assessed an available product, arranged funding from a bank account and placed the order through EX.IO, linking functions that are usually separated across financial applications.
The final transaction involved an RWA product, a term generally used for blockchain-based representations of financial or physical assets. EX.IO said its RWA offerings are structured for different risk tolerances, expected returns and liquidity preferences, though the demonstration did not provide details on the specific asset selected, the amount subscribed or the banking institution involved in the sandbox.
The structure of the trial places bank transfers at the centre of agentic finance. Tokenised products have often been tested in settings where users already hold digital assets or tokenised cash. Connecting an agent to a bank account introduces a more familiar funding route, while also raising the stakes for authentication, spending controls and clear limits on what an agent can do without another user prompt.
A successful production version would need to establish more than a smooth customer journey. Banks, platforms and regulators would need common rules on how an agent proves it has authority to act, how payment credentials are protected, how instructions can be revoked and which party is responsible if a transaction is disputed or executed incorrectly.
Payment networks are building authorisation frameworks
Major payment companies have been developing infrastructure for this type of delegated transaction. Visa and Mastercard outlined agent-led payment initiatives in 2025 focused on user authorisation, payment credentials and ways for merchants to verify that a software agent is acting for a legitimate customer.
Google also introduced its Agent Payments Protocol, or AP2, in 2025. The framework sets out how agents, users and payment providers could exchange verifiable payment instructions, including for instant bank transfers. Such protocols are designed to preserve a record of the user intent even when the user is not manually confirming each step in a checkout process.
Those efforts address a problem that becomes more difficult when an AI system is allowed to act rather than merely recommend. A payment system needs to distinguish between a user asking an agent to research a product and a user authorising it to commit funds. In investment-related transactions, the difference can involve market risk, liquidity restrictions and suitability considerations as well as payment security.
EX.IO demonstration applies that issue to an RWA subscription, where an AI agent may need to interpret product information before it sends money and submits an order. The trial therefore sits at the intersection of two developing areas: tokenised financial products and payment systems designed for software agents.
Hong Kong is moving tokenisation trials toward transactions
Hong Kong monetary authorities have been creating controlled settings for experiments involving tokenised money and assets. The Hong Kong Monetary Authority launched EnsembleTX in 2025 to advance tokenised finance trials into real-value transactions within a managed environment.
Its initial areas of focus included using tokenised deposits to transact in tokenised money market funds, along with real-time liquidity and treasury management. That agenda shares a practical concern with the EX.IO test: tokenisation becomes more useful when the payment leg and the investment or settlement leg can operate together.
At Hong Kong FinTech Week 2026, the HKMA also announced Project SAGE and published a discussion paper on banks use of agentic AI. The initiative was presented as support for industry work on agentic finance within governance frameworks, reflecting regulators focus on how banks should supervise AI systems that could influence or initiate financial activity.
The EX.IO proof of concept does not show that autonomous agents can independently manage a customer portfolio or make unrestricted investment decisions. Its narrower achievement is demonstrating a linked workflow in which an agent can move from product selection to bank funding and order execution in a sandbox.
Licensed platform tests a new transaction model
EX.IO operates as a virtual asset trading platform licensed by Hong Kong Securities and Futures Commission. Its licence was dated December 2024, following a period in which the platform operated under Hong Kong deemed-to-be-licensed regime.
The platform was also identified as the only VATP selected for a programme run by Hong Kong Office for Attracting Strategic Enterprises. That positioning gives the demonstration relevance within Hong Kong attempt to develop regulated digital-asset and tokenisation infrastructure, rather than presenting it as an experiment detached from the city licensing system.
McKinsey has estimated that agentic commerce could orchestrate between $3 trillion and $5 trillion in global goods sales by 2030, provided AI adoption and merchant readiness continue. The estimate excludes services and business-to-business transactions and does not measure payment-processing revenue or the prospective size of the RWA market. Even so, it illustrates why banks, payment networks and technology companies are working on the authorisation layer before consumer-facing agent payments become commonplace.
For tokenised investment platforms, the appeal lies in reducing the operational friction between deciding to buy an asset and funding the purchase. The harder task will be ensuring that convenience does not weaken customer control. Hong Kong sandbox-led approach gives firms room to test those boundaries before agent-led bank payments are exposed to broader retail use.
Explore how tokenised assets reshape finance in 2026—read this RWA megatrend overview next.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
