Agentic commerce: Driving smarter growth with every interaction

Agentic AI is one of the hottest topics in the payments space. To cut through the hype, Alla Gancz joined the Voice of MPE podcast to explore how the technology redefines retail consumer journeys, transforms checkout and payments experience.

The latest announcement from OpenAI and Stripe, introducing instant checkout and Agentic Commerce Protocol, marks a true turning point for retail and payments. For the first time, a conversational AI platform ceases to be merely an intermediary for information and becomes a transactional environment, where search, recommendation and checkout merge into a single, seamless experience.

What is Agentic AI? What sets it apart from today’s AI in payments?

Unlike traditional AI tools, which rely on humans to direct every step — providing inputs, interpreting outputs, and deciding what to do next — AI agents operate with autonomy and intent. Agentic AI refers to these goal-driven systems that don’t just react to prompts but can reason, plan and take action independently on behalf of humans. In contrast to today’s AI tools in payments — such as fraud detection models flagging anomalies, chatbots handling FAQs or recommendation engines suggesting offers — agentic AI can execute entire end-to-end processes. It doesn’t just assist with tasks; it delivers outcomes.

So agentic commerce will see agents undertake shopping tasks for customers, including finding products, tailoring recommendations and completing transactions automatically. 

Advances in AI allow these digital assistants to meet customers’ demands for seamless, customized shopping. Merchants’ priorities shift from getting their products to show up for search engines to ensuring products are front and center for AI systems. At the core of these processes is the integration of sophisticated payment solutions to facilitate smooth transactions.

That autonomous buying capability removes friction for the customer, but places additional burdens on merchants. They need to establish a common foundation to securely authenticate, validate and convey an agent’s authority to transact. While today’s payment systems generally assume a human is clicking the ‘buy’ button, autonomous agents will be able to initiate payments. 

AI agents transacting for users need a secure foundation to authenticate, validate and convey their transaction authority. Current payment systems assume direct human interaction but autonomous agents initiating payments challenge this, raising critical questions:

• Authorization. Proving that a user gave an agent the specific authority to make a particular purchase.

• Authenticity. Enabling a merchant to be sure that an agent's request accurately reflects the user's true intent. 

• Accountability. Determining accountability if a fraudulent or incorrect transaction occurs.

And we’re already seeing some attempts to address this. Take Google Cloud’s new Agent Payments Protocol (AP2), which provides a payment-agnostic framework for users, merchants and payments providers to initiate and transact agent-led payments. Google worked with more than 60 partners to create this protocol.

How will agentic AI impact consumer journeys?

For consumers, the promise is twofold:

• More choice — access to richer, more personalized and contextual recommendations.

• Less friction — purchases can be completed in seconds, without navigating multiple pages.

Customer journeys will become hands-off, hyper-optimized and invisible. Instead of manually checking out, renewing subscriptions or managing loyalty points, consumers will delegate these to AI agents.

For example:

• At checkout, an AI agent could compare multiple merchants in real time, find the best deal, apply loyalty points and pay seamlessly in the background.

• For subscriptions, agents could renew, cancel, or switch providers automatically based on price changes or usage — think of Netflix automatically pausing your subscription when you’re traveling.

• Loyalty becomes proactive: Delta Airlines is already experimenting with AI to optimize redemption. In the future, agents could automatically upgrade you to business class using your points, or ensure you never lose benefits because of missed deadlines.

This means far less direct interaction between merchants and customers. Instead, merchants will need to make their offers machine-readable and agent-friendly, because they’re now competing for placement in AI agents’ decision logic, not just on websites or apps.

What practical use cases are already emerging today?

It’s still early days, but we’re already seeing some powerful applications of agentic AI, for instance: 

• AI shopping assistants comparing prices and making purchases.

• Automated bill management handling due dates and payments.

• Dynamic fraud prevention that spots unusual patterns in real time.

• Personalized shopping agents curating deals and optimizing loyalty rewards.

Global payment leaders are moving quickly to make their networks agent-ready — providing developers and businesses with secure, scalable infrastructure for autonomous transactions and commerce:

• Visa’s Intelligent Commerce platform enables AI agents to securely discover, select and purchase products on behalf of consumers.

• Mastercard launched its own platform Agent Pay creating a standardized way for agents to initiate, authorize and settle payments.

• PayPal’s Agent Toolkit enables developers to build sophisticated agentic workflows that handle payments.

In the near term, as technology continues to mature, expect to see merchants experiment with the services they can through agentic channels.  

What opportunities do merchants have — and how can they get ready?

We’re already seeing agentic AI use cases move from concept to reality:

• Shopping assistants like Klarna’s AI (powered by OpenAI) can compare prices and even complete purchases.

• Automated bill management is happening in apps like Cleo or Mint, which already schedule payments and optimize balances — agentic AI will make this hands-free.

• Dynamic fraud prevention is in pilots: Mastercard is testing AI models that adapt in real time to novel fraud patterns, rather than relying on fixed rules.

• Travel booking agents: Google is experimenting with AI agents that combine flights, hotels and payments into one optimized journey.

And we see there are three primary channels where retailers can activate agentic commerce:

• Business to agent (B2A) commerce to participate in agentic channels by making your product catalogs and checkout flows accessible to third-party AI platforms such as Perplexity or ChatGPT.

• Agent to consumer (A2C). Embed agentic experiences on your own site or app through branded shopping assistants (agent to consumer, or A2C) or natural-language interfaces — such as Saks Fifth Avenue’s AI agent — to both improve customers’ experience and conversion.

• Agent to agent (A2A). Power autonomous purchasing agents by enabling secure, software-triggered transactions via subscriptions or replenishment flows — such as a personal grocery AI agent automatically replenishing across multiple retailers when stock is low.

Looking ahead, retailers don’t need to rebuild their tech stack but extend it. Many already have the pieces in place: structured catalogs, APIs, personalization engines. Agentic commerce is about adapting those systems to interact with a new class of buyer — autonomous, software-based and always-on.

Nonetheless, agentic commerce introduces new complexities that need to be accounted for as retailers extend their tech stack. For instance, agents don’t operate like humans: they don’t browse, they query; they don’t click through pages, they act on structured prompts. That means the systems behind the scenes — catalog, checkout and payments — need to evolve in tandem.

Future outlook and key considerations for banks and payment service providers (PSPs)  

Platforms and ecosystems. The rise of agentic commerce could create a new layer of platforms that mediate between consumers, agents and merchants. While today’s tech giants have an early advantage, new entrants could emerge around AI models or trust infrastructure. What makes this shift different is that influence will move from controlling consumer interfaces to controlling how agents access data, make decisions and execute transactions.

Data and trust infrastructure. Behind the rise of agentic commerce is a growing race to build the trust layer that will let AI agents identify themselves, prove user consent and transact securely. Standards bodies are starting to define the foundations for digital identity and authorization. We’re already seeing major payment networks, cloud providers and fintech platforms try to extend these standards into commercial frameworks.

New revenue models. Payment players will continue to control the transaction and settlement layer but agentic commerce opens an additional opportunity. Providers can move up the value chain by enabling the services that agents depend on — verified identity, authorization, data validation and agent-to-agent transaction support. These capabilities can evolve into new revenue models built around trust, compliance and interoperability (on top of existing pure transaction volumes).

New KPIs. Agentic commerce will change how performance is measured. Instead of focusing on clicks or conversions, merchants will need to optimize for how well their products can be discovered, interpreted and transacted by AI agents. New KPIs will be very different versus what we know today and will emphasize data quality, structured catalog accessibility, API reliability and trust or verification scores.

User control and safety tools. As agents gain autonomy, users will need clear ways to set limits, permissions and preferences. These tools are likely to surface within banking apps, digital wallets and platform settings, allowing consumers to review, approve or override agent actions.

What risks and challenges come with Agentic AI in payments?

Key risks include:

• Fraud and social engineering. Just as merchants deploy agentic AI, fraudsters will too. We could see AI-powered bots launching social engineering attacks at scale. Payment networks need to build equally advanced defensive agents. 

• Liability and chargebacks. If my AI agent buys the wrong subscription or falls for a scam, who’s responsible — me, the merchant or the AI provider? Current chargeback frameworks aren’t designed for autonomous actors.

• Transparency and explainability. Agents can feel like “black boxes.” Regulators, merchants and consumers all need audit trails to see why an agent made a choice.

European regulators are already moving:

• The EU AI Act requires explainability and accountability.

• PSD3 (successor to PSD2) is expanding scope for strong authentication.

• The Digital Services Act (DSA) sets rules on transparency and consumer protection.

This could give Europe a trust advantage, but there’s a fine balance. Over-regulation risks slowing innovation compared to the US or Asia, where adoption may move faster. The winners will be those ecosystems that combine speed with trust.

What’s next?

This is why it’s a step-change, not a rebrand. AI is shifting from classification and prediction to autonomous action and adoption is happening faster than earlier waves. Fraud models and chatbots took years to mature and gain trust. But with agentic AI, the APIs, cloud infrastructure and regulatory frameworks already exist — so we’re likely to see deployment in months, not years, especially in payments where competition is fierce.

For payments firms, the call to action is clear:

• Build the rails and standards for agent-driven commerce

• Embrace or co-create protocols to enable agentic payments 

• Partner across ecosystems to shape adoption

The future of payments isn’t just faster checkout. It’s a new era where intelligent agents reshape the entire commerce value chain.