HSBC and Ant Digital Run End-to-End AI-Agent Payment Test

HSBC and Ant Digital Technologies conducted a demonstration in which an artificial intelligence agent identified a suitable digital service and executed a payment using tokenized deposits provided by the British lender. The test relied on Ant Digital’s Anvita Flow network to handle service discovery and execution while HSBC managed settlement through its Tokenised Deposit Service and performed real-time risk assessments. All activity took place on Jovay Testnet, a layer-2 blockchain environment chosen for its ability to simulate near-production conditions.

The transaction involved a micropayment, defined by participants as a transfer below $2. Settlement occurred instantly on the blockchain testnet, eliminating the delays typical of traditional batch processing. HSBC contributed both the deposit tokenization and the immediate compliance checks, while Ant Digital’s system coordinated the automated flow that required no manual intervention once the agent initiated the request.

Micropayments Align With AI-Agent Commerce Needs

Micropayments suit scenarios where an AI agent must make frequent, low-value transactions across applications without human oversight. The HSBC-Ant Digital test targeted exactly that use case, focusing on speed, automation and programmable execution. By pairing tokenized deposits with blockchain settlement, the workflow demonstrated how banks can integrate existing payment infrastructure with emerging agent technology.

Real-time risk controls formed a core part of the demonstration, ensuring that suspicious patterns or policy breaches could be detected and addressed during settlement. This integration of compliance into the automated process addresses a key requirement for scaling AI-driven finance beyond controlled labs.

Broader Industry Experiments With AI Agents

The HSBC trial joins a series of bank-led initiatives exploring how artificial intelligence agents might initiate financial transactions on behalf of users. Santander completed Europe’s first live end-to-end payment executed by an AI agent in March using Mastercard’s Agent Pay infrastructure and the bank’s own payment systems. In May, Swiss digital asset provider Sygnum reported the first live AI-agent-driven digital asset transactions on a blockchain mainnet, with explicit customer approval and signing required for each action. Spanish bank CaixaBank also ran an AI-agent-initiated card transaction through Visa Intelligent Commerce and existing merchant rails.

These efforts reflect different technical approaches. Some leverage traditional payment networks with agent tooling, while others test tokenization and direct blockchain settlement. HSBC’s demonstration emphasizes programmable rails and tokenized deposits, offering one path among several for enabling always-on agent activity.

Legacy Systems Face Questions for 24/7 Agent Workflows

Despite the momentum, questions remain about whether established banks can quickly adapt decades-old clearing and settlement infrastructure originally built for human-operated processes. In a May interview, Augustus Bank CEO Ferdinand Dabitz noted that traditional systems were not designed for continuous, automated transaction execution driven by software agents. He pointed to his bank’s ongoing work on a US model built around stablecoins and AI-native operations as a potential alternative path.

HSBC’s test does not resolve these concerns. The exercise stayed within a testnet environment and included basic risk checks, but it leaves open the question of scalability, auditing and operational readiness for higher volumes or more complex consent flows. The pilot signals that banks are actively experimenting with agent-compatible architectures, yet the gap between laboratory success and production deployment remains wide.

Programmable Infrastructure Could Accelerate Blockchain Adoption

Investment research firm Citrini Research recently argued that autonomous AI agents operating continuously across applications will drive demand for financial systems that can execute logic programmatically. Traditional setups designed around human users may require evolution to support machine-initiated money movement. Citrini wrote that blockchain networks can deliver the always-on rails needed for such activity.

In the Oct. 8 report Breaking The Wall, the firm stated: “AI agents move programmatically, 24/7, across applications, and it’s only logical that money and financial assets eventually will, too.” HSBC and Ant Digital’s test operationalizes elements of that thesis by combining tokenized deposits, real-time settlement and agent-driven service access in a single workflow. The next steps for the industry will likely involve tracking whether these demonstrations expand to include clearer user consent models, broader service categories and measurable performance under sustained loads.