British banking giant HSBC and Ant Digital Technologies have conducted a technical verification test of a system that allows AI agents to access digital services and execute micropayments using tokenized bank deposits. The transactions were settled in real time on the Jovay Testnet, a layer-2 blockchain testing environment. The pilot combined HSBC’s Tokenised Deposit Service with Ant Digital’s Anvita Flow network, which coordinates service discovery and payment execution for AI agents. In this demonstration, an AI agent selected a digital service and completed a payment defined as a micropayment, typically less than $2. HSBC provided the settlement capabilities and real-time risk checks, while Ant Digital managed the coordination of service access. Both companies emphasized that the test was limited to technical verification and did not constitute a commercial launch or a live customer offering.
This development occurs amid broader industry efforts by established banks to integrate AI-driven financial transactions. In March, Santander completed an AI-agent-initiated payment using Mastercard’s Agent Pay infrastructure within its live payment systems. Swiss digital asset bank Sygnum followed in May with a test of AI-agent-driven transactions on a blockchain mainnet, requiring customers to approve each transaction. Similarly, CaixaBank utilized Visa Intelligent Commerce for an AI-agent-initiated card transaction. However, skepticism remains regarding the adaptability of traditional infrastructure. Augustus Bank CEO Ferdinand Dabitz argued in May that legacy clearing banks rely on decades-old systems designed for human operations rather than automated, around-the-clock transactions. Meanwhile, research firm Citrini Research suggested in an October report that autonomous AI agents could increase demand for programmable financial infrastructure, noting that traditional systems may need adaptation as AI agents handle transactions across applications.
The collaboration between HSBC and Ant Digital highlights a critical phase in the maturation of agentic finance: moving from theoretical frameworks to controlled technical verification using existing banking rails. By leveraging tokenized deposits and a layer-2 testnet, the partners are addressing the fundamental challenge of integrating high-frequency, low-value AI interactions with the rigid compliance and settlement requirements of traditional banking. This approach suggests that incumbents are prioritizing risk management and regulatory alignment over rapid deployment, recognizing that AI agents require deterministic settlement finality and real-time fraud detection mechanisms that current fiat systems struggle to provide natively at scale. The use of a testnet rather than a production environment underscores the cautious stance major institutions are taking toward autonomous financial actors.
Institutional adoption of AI-driven payments faces significant structural hurdles, particularly concerning the compatibility of legacy infrastructure with 24/7 automated operations. While banks like Santander and CaixaBank have demonstrated proof-of-concept integrations with card networks, the persistent argument from challengers like Augustus Bank indicates that retrofitting decades-old clearing systems may be inefficient compared to building purpose-native architectures. The market impact will likely depend on whether traditional banks can successfully abstract the complexity of blockchain settlement behind familiar interfaces without compromising security. Regulators will closely monitor how liability is assigned when autonomous agents initiate transactions, especially in cross-border contexts involving tokenized deposits. The success of these pilots may determine whether the future of agentic commerce relies on hybrid models bridging fiat and crypto infrastructure or if new, fully programmable financial layers will displace traditional intermediaries.
