HSBC and Ant Digital Trial AI-Agent Payments with Tokenized Deposits

The way payments work could be entering a new phase—one in which artificial intelligence agents do more than recommend products or automate tasks. In a recent technical trial, HSBC and Ant Digital demonstrated how AI agents could discover services, initiate transactions, and make micropayments using tokenized bank deposits.

The experiment brings together several important developments in finance and technology: agentic artificial intelligence, tokenized money, automated service discovery, real-time settlement, and built-in transaction risk controls. While the trial remains a technical demonstration rather than a consumer product launch, it offers a glimpse into how autonomous digital systems may eventually participate in the economy.

How AI-Agent Payments Work

Traditional online payments generally require a person to select a product or service, enter payment details, approve a transaction, and wait for processing. AI agents could change that process by allowing software to complete certain steps independently based on a user’s instructions, preferences, and spending limits.

For example, a user might ask an AI agent to find the best available data service, book a short-term digital subscription, or purchase access to a specific computing resource. The agent could then identify compatible providers, compare offers, confirm the terms, and make a small payment without requiring the user to manually complete every stage.

In the HSBC and Ant Digital trial, the agents were designed to discover available services and make micropayments. This is particularly relevant for machine-to-machine commerce, where software systems may need to pay each other frequently and in small amounts.

Why Tokenized Deposits Matter

The payments in the trial were supported by HSBC’s tokenized deposits. Tokenized deposits represent money held in a bank account in a digital token format, while remaining connected to the underlying banking system and subject to the bank’s controls.

This approach differs from using a traditional cryptocurrency or an unbacked digital token. Tokenized deposits are intended to combine the familiarity and regulatory structure of commercial bank money with the programmability and speed of blockchain-based settlement. They can potentially be used to automate payment conditions, support immediate transfers, and provide clearer records of transactions.

For businesses, this could make it easier to manage automated payments while maintaining a connection to existing treasury, compliance, and accounting processes. Instead of treating digital assets as an entirely separate financial system, tokenized deposits may serve as a bridge between established banking infrastructure and emerging programmable applications.

Real-Time Settlement and Automated Risk Checks

One of the most important elements of the trial was the combination of real-time settlement and transaction risk checks. AI agents may be able to act quickly, but speed alone is not enough for financial transactions. Automated systems must also determine whether a payment is authorized, legitimate, and within defined limits.

Risk controls can include checks on the identity of the parties, transaction size, payment frequency, service terms, and the agent’s permissions. These safeguards are especially important when agents operate with limited human supervision. A system that can make payments independently must also be designed to prevent unauthorized purchases, fraud, duplicate transactions, and accidental overspending.

Real-time settlement may also improve the experience for service providers. A provider could receive payment immediately after delivering a digital service, reducing settlement delays and potentially allowing it to offer resources on a pay-as-you-go basis.

Potential Use Cases for Agentic Commerce

AI-agent payments could support a wide range of future applications. Digital platforms might use agents to purchase cloud computing capacity only when it is needed. Connected devices could pay for data, network access, or maintenance services automatically. Businesses could deploy agents to handle recurring procurement tasks, while consumers could authorize agents to manage subscriptions or make small purchases according to preset rules.

Micropayments are particularly interesting because conventional payment systems are not always efficient for very small transactions. If processing costs and delays are too high, providers may be unable to charge for brief access, individual data requests, or small increments of usage. Tokenized deposits and automated settlement could make these business models more practical.

Challenges Before Wider Adoption

Despite the promise, several challenges must be addressed before AI-agent payments become common. The first is trust. Users need to understand what an agent is permitted to do and how they can reverse or dispute a transaction.

Interoperability is another major issue. Agents, banks, payment networks, and service providers will need shared standards so that systems can identify one another, communicate transaction terms, and confirm payment securely. Without common technical frameworks, agent-based commerce could remain fragmented.

Regulation and accountability will also be essential. Questions may arise over who is responsible when an agent makes an incorrect payment, misunderstands an instruction, or interacts with a fraudulent service. Financial institutions will need clear policies governing identity verification, data protection, consumer safeguards, and automated decision-making.

A Step Toward Programmable Banking

The HSBC and Ant Digital trial highlights a broader shift toward programmable banking, where financial transactions can be embedded directly into software workflows. Rather than treating payments as a separate step that happens after a decision, future systems may allow payments to occur automatically as part of an agent’s activity.

The technology is still developing, and a technical trial does not guarantee immediate commercial deployment. However, it demonstrates how banks and technology companies are exploring a future in which AI agents can interact with markets, services, and financial infrastructure on behalf of people and businesses.

If the industry can establish reliable safeguards, interoperable standards, and clear user controls, tokenized deposits may provide a practical foundation for this emerging form of digital commerce. AI agents would not simply help users decide what to buy—they could eventually discover services, negotiate access, and settle payments in real time.