How payments infrastructure is becoming a strategic asset

The payments industry has long operated on false dichotomies—scale or agility, proven or innovative, fast or safe. Marqeta CEO Mike Milotich recently joined Alex Johnson on the Fintech Takes podcast to discuss why the "either/or" mentality is dissolving across a few key areas in payments, and how Marqeta is enabling payment possibilities for customers across many industries and use cases.

Modern payment experiences are designed so consumers never think about the payment itself, yet this simplicity hides enormous back-end complexity. "The more seamless and almost invisible payments become on the front end, the more complicated it becomes on the back end," Milotich explains. Marqeta provides the payment building blocks that our customers can put together in unique ways to deliver value to their users, whether through creating new revenue streams, deepening engagement, or improving access to capital. Our primary objective is to provide our customers with more control and agility in issuing payment credentials.

The agentic commerce opportunity

As we're seeing in the headlines, the payments industry is beginning to explore agentic commerce: AI agents conducting transactions on behalf of users. While the underlying infrastructure exists today, 2026 will likely be the year of experimentation and identifying meaningful use cases where the value justifies the investment.

Marqeta's Model Context Protocol (MCP) server provides an integration layer that allows AI agents to directly interact with payment infrastructure through standardized APIs. This creates a direct connection between AI applications and issuing functions, enabling agents to pull balances, initiate transactions, and receive real-time signals—all while operating within defined policies.

Marqeta's safety controls for agent card issuance

Capability	Description
Issue virtual cards	Instantly provision ready-to-use virtual cards for agent checkout
Set spend and velocity controls	Define merchant categories, amount limits, and transaction frequency
Real-time authorization decisioning	Approve or decline transactions at the moment of purchase
Dispute transactions	Initiate and manage disputes for overcharges or unauthorized purchases
Manage and review transactions	Track spending against budgets and reconcile purchases automatically

To understand how AI agents will spend on a user's behalf, corporate expense management is a great example since it already governs delegated spending at scale. "An AI agent could create and configure a card with very specific instructions from that user in terms of what it wants to go purchase," Milotich explains.

Single-use virtual cards as a security mechanism

Single-use virtual cards are credentials generated for a single transaction that cannot be reused. This design ensures that even if card credentials are exposed after a purchase, they cannot be used again—minimizing the blast radius of any potential compromise. Businesses can use metadata to tie each card back to the specific task, order, or agent that triggered it, creating a complete audit trail.

Spend controls for agent transactions

Spend controls constrain what an agent card can purchase. Key parameters include:

Merchant category controls: Restrict the card to specific business types (e.g., a travel booking agent can only charge airlines and hotels, not restaurants or retail stores)
Amount limits: Set maximum transaction or daily spending thresholds
Velocity controls: Limit the number of transactions within a time period
Geographic restrictions: Constrain where transactions can occur

Trust-level framework for credential scope

Depending on trust level, credentials take different structures:

High trust: Broad permissions with standard spend controls for established, verified agents
Medium trust: Tighter constraints with lower limits and narrower merchant categories
Low trust: Single-purpose virtual cards with narrow parameters scoped to execute only the specific purchase the user intended

This approach follows a "just-in-time" model where the system issues credentials for a specific task with limited scope and duration, minimizing risk if a credential is intercepted.

Real-time authorization and post-transaction monitoring

Marqeta's Real-Time Decisioning capability allows businesses to define when and how agents move funds, set permissions at the card level, and introduce fraud controls at transaction authorization. This positions the platform as a core defense against evolving threats in agentic commerce.

Post-transaction, businesses can detect unusual spending patterns, track spending against budgets, reconcile purchases automatically, and catch or dispute overcharges—providing complete oversight of agent activity.

Agent card use cases with safety controls

Automated business spend: Developers use agents to handle routine purchases with merchant category restrictions and amount limits
Expense management: Fintech providers embed agent-issued cards to control and reconcile employee spend in real time with velocity controls
Travel booking: Vertical SaaS platforms issue agent cards to SMB customers with cards restricted to airlines and hotels only
Procurement automation: Agents purchase supplies with single-use cards tied to specific purchase orders
Subscription management: Agents manage recurring payments with cards scoped to specific vendors and amounts

Evolution of BNPL

According to Marqeta's 2025 State of Payments report, BNPL usage has expanded beyond large, discretionary purchases to include everyday spending like groceries and routine purchases. Milotich noted that Marqeta is seeing a shift in how people are using BNPL, and it's "becoming a much more broadly used case. People are using it much more thoughtfully as a cash flow management tool." Consumers are turning to BNPL to help them strategize their spending, regardless of category.

In addition, BNPL is moving from

checkout options to card features themselves. With Marqeta and Visa Flexible Credential, Affirm and Klarna now issue debit cards that let users elect installments anywhere the network is accepted. Milotich noted this effectively unlocks tens of millions of merchants at once, enabling cardholders to choose how and when they pay.

Building for the flexible, AI-enabled payments era

The competitive focus in payments is shifting from rewards and rates to payment choice and flexibility. As Milotich emphasizes, "You don't have to make that trade-off or sacrifice like you've had to make in the past. We can maintain a highly flexible and nimble platform that is also delivered at scale." Those who win will see payments as strategic infrastructure that enables scale, flexibility, and new AI-powered, user-driven experiences.

To listen to the full podcast episode, click here.

Subscribe to our newsletter

Subscribe for the latest news, updates and trends.

Unsubscribe at any time. By entering your work email, you agree to receive marketing emails from Marqeta. California residents can learn more about their rights here.

Discover more insights