New to Nonstop: Join the Journey

AI is beginning to move from advising customers to acting for them, and that shift could reshape how payments are initiated, authorised and trusted.

Agentic commerce is emerging as a new trend in digital commerce, enabling AI agents to do more than recommend products or answer questions. Increasingly, AI can act on a customer’s behalf throughout the purchasing journey, introducing both new opportunities and new challenges for banks, merchants and payment providers.

Following the recent HPE Payments Agentic Commerce demonstration at HPE Discover, Martin Lomax (HPE FSI Payments Specialist) discusses what agentic commerce means, how it works, and what organisations should be doing now to prepare.

What is agentic commerce and why does it matter now?

Agentic commerce is the next evolution of digital commerce, where AI agents do more than recommend products or answer questions. The agent can ‘mediate’, that is, actively participate in the purchasing process on behalf of the customer.

That may include:

• Discovering products
• Comparing options
• Checking availability
• Applying customer preferences
• Initiating checkout
• Completing payment within defined controls

This shift is significant because current e-commerce relies on customers actively searching, selecting and paying themselves. In contrast, agentic commerce allows customers to express what they need and delegate more of the purchase journey to an AI agent. That creates opportunities for convenience and personalisation, but it also introduces new technology requirements and important business questions around liability, consumer protection and trust.

How is AI already transforming the payments industry?

AI is already influencing multiple areas of the Payments ecosystem. The most mature use case is fraud detection, where machine learning models assess transactions in real time and provide risk scores that contribute to authorisation decisions.

Beyond the real-time transaction flow, organisations are increasingly applying AI to:

• Anti-money laundering (AML) screening
• Dispute and chargeback triage
• Customer service automation
• Operational efficiency
• Data analysis and reporting

Like most industries, the Payments industry is also adopting broader enterprise AI capabilities to enhance productivity, streamline operations and improve decision-making across the business.

What does a real-world agentic commerce transaction look like?

A real-world agentic commerce transaction typically begins with a conversation between the customer and an AI agent.

For example, a customer might say: “Find me an anniversary gift under €200 that can be delivered by a trusted merchant before the weekend.”

The AI agent interprets the request, searches available products and merchants, compares options, and presents a shortlist for consideration.

Once the customer selects a preferred option, the process moves into purchase authorisation. At some point, the customer must explicitly authorise the AI agent to act on their behalf. This is likely to involve a form of Strong Customer Authentication (SCA) and the creation of delegated permissions that define what the agent is allowed to do.

Those permissions could include:

• Maximum transaction value
• Approved merchants
• Product categories
• Spending frequency
• Time limits

The AI agent can then complete the purchase within those agreed rules, creating a balance between customer convenience and appropriate control.

Can you walk us through the customer journey demonstrated in the HPE Payments Agentic Commerce demo?

The important point for the demo is that the payment platform is no longer just processing a transaction submitted by a merchant. It is recognising that the purchase was initiated by an authorised AI agent and validating it against a customer-approved mandate.

The customer journey we implemented closely mirrors a real-world agentic commerce experience, although the demonstration deliberately focuses on the agentic commerce elements rather than implementing a full Strong Customer Authentication process.

The Shopping Experience

The customer interacts with an AI shopping agent connected to several simulated e-commerce stores selling consumer electronics and computing products.

The agent searches across participating merchants, identifies suitable products and presents options before proceeding with a purchase.

AI Mandate Validation

When the customer instructs the agent to complete the purchase, they provide a user identifier.

That identifier is linked to an AI Mandate stored within the Lusis Tango payments platform. The mandate contains predefined controls that govern what the AI agent is authorised to do.

These controls include:

• Transaction value limits
• Merchant whitelists
• Delegated purchasing permissions

Transaction Authorisation

Tango issues a purchase token to the AI agent, which is then passed to the merchant.

Rather than submitting traditional payment credentials such as card and PIN data, the merchant includes the token as part of its authorisation request.

Tango recognises the transaction as an AI-mediated purchase and validates it against the rules contained in the associated AI Mandate before making an authorisation decision.

Transaction Completion

Once approved, the purchase proceeds in much the same way as a conventional e-commerce transaction.

The key difference is that the merchant returns the outcome to the AI agent, which then communicates the result back to the customer.

What technologies and ecosystem partners are involved?

The demonstration combines technologies developed by HPE Nonstop, Lusis and HPE Labs.

Three APIs were implemented as part of the solution:

• The Agentic Commerce Protocol (ACP) API, defined within Google’s Universal Commerce Protocol initiative
• An AI Mandate Management API
• A Transaction Authorisation API

The AI Mandate and Transaction Authorisation APIs were defined by Lusis and implemented within the Tango payments platform. They provide the mechanisms required for delegated purchasing permissions and AI-mediated payment authorisation.

The supporting integrations used by the shopping agent and participating merchants were developed by colleagues within HPE Labs.

Platform Deployment

The demonstration environment consists of:

• Lusis Tango Payments Hub running on HPE Nonstop
• Lusis AI Fraud running on HPE Nonstop
• Simulated e-commerce stores running on virtualised Linux infrastructure
• The AI shopping agent webapp running on virtualised Linux infrastructure and integrating with a large language model

Real-World Ecosystem Examples

Outside the demonstration, agentic commerce initiatives are already emerging across the payments ecosystem. One example is a French festival ticketing solution involving Crédit Agricole, Worldline, Mastercard and Weezevent.

The solution demonstrates how issuing banks, payment service providers, merchants and payment networks can work together to support AI-mediated purchases within a controlled environment.

You first presented the demo at HPE Discover. What customer feedback have you received so far?

The feedback has been remarkably consistent. Most organisations recognise that agentic commerce is likely to become an important part of the future payments landscape. The debate is not whether it will happen, but how quickly adoption will occur and what form it will take.

Customer reactions generally fall into two groups. Some organisations are moving aggressively to establish an early market position and explore new customer experiences enabled by AI-mediated purchasing. Others remain more cautious, primarily because of concerns around:

• Consumer protection
• Liability allocation
• Regulatory compliance
• Customer trust
• Potential negative customer experiences

Everyone acknowledges the change and the opportunity, but there are differences in their assessments of how quickly their organisations should move.

What should customers do now to prepare?

Many financial institutions, merchants and payment providers are already evaluating the implications of agentic commerce.

For those just beginning, it is important to recognise that competitors are unlikely to stand still. Organisations need to understand how agentic commerce could affect their customers, their operating model and their market position.

To prepare effectively, there are three key areas to focus on.

1. Technology readiness

Organisations should assess whether their systems, APIs, data platforms and operational processes are capable of supporting delegated purchasing models such as agentic commerce.

2. Liability, consumer protection and trust

Clear frameworks are required to determine responsibility when an AI agent acts on behalf of a customer. These need to cover dispute handling, reimbursement, consent and authorisation.

3. Customer journey evolution

With the other two areas defined, businesses should develop a roadmap for how their customer experiences may evolve over time, starting with relatively simple AI-driven transactions and progressing to more complex journeys.

For example, a simple search, compare and buy request could be implemented relatively quickly. More complex journeys may be better suited to a later stage in the roadmap, such as:

“Restock the household essentials for the next month, but optimise for total cost, delivery timing, loyalty benefits, sustainability preferences and my existing subscriptions. Don’t spend more than €350, avoid products with poor reviews, use any discount codes or benefits I’m entitled to, and only place orders that can arrive before next Friday.”

Where do you think agentic commerce will be in three years?

Agentic commerce is unlikely to evolve at the same pace across all markets. Adoption rates will vary based on factors such as:

• Regulation
• Culture and local consumer attitudes
• Market structure
• Payments infrastructure
• Availability of supporting ecosystems

In the near term, we’re likely to see a rise in ‘merchant-bound’ agentic commerce services. These are solutions where a merchant operates its own AI purchasing agent and the customer remains within that merchant’s ecosystem. The French festival ticketing system is a good example of this model.

Over time, broader ‘marketplace’ agents are likely to emerge, enabling customers to purchase from multiple merchants through a common AI platform. What seems less likely is the emergence of fully ubiquitous agentic commerce within the next three years.

The technology is advancing rapidly, but widespread adoption will depend on the industry’s ability to establish trusted frameworks for authentication, liability, customer protection and governance.

The direction of travel is clear; the pace is likely to differ significantly across countries, industries and customer segments. The most successful implementations are likely to emerge where banks, merchants, payment networks, technology providers and regulators start learning together: testing customer journeys, clarifying liability models and building the trusted payment infrastructure needed for AI-mediated commerce.