Mastercard Just Published the Blueprint for AI-Powered Payments

On July 23, Sunrate and Mastercard released a white paper at the World Artificial Intelligence Conference in Shanghai. The title tells you what you need to know: Beyond Automation: Defining Agentic Global Payments.

Read that again. Beyond automation. Not “better automation.” Not “faster automation.” Beyond it.

This is Mastercard — the company that processes billions of transactions across 200+ countries — publishing a formal framework for AI agents that independently run payment operations. Not as a concept paper. Not as a “what if.” As a category definition with 16 mapped pain points, 13 production use cases, and a governance architecture.

If you’ve been waiting for the signal that AI is coming for how money actually moves, this is it.

What the White Paper Actually Says

The core thesis is straightforward: cross-border B2B payments have evolved through three stages — digitization, automation, and now autonomy. We’re entering the third stage.

In the autonomy stage, AI agents with reasoning, planning, and execution capabilities independently orchestrate end-to-end payment workflows. Not just processing a single invoice. Coordinating the entire journey: supplier onboarding, compliance screening, payment routing, FX optimization, execution, reconciliation, and exception handling.

The paper maps 16 specific pain points across the B2B payment lifecycle that current systems handle poorly:

• Fragmented workflows across disconnected systems
• Foreign exchange inefficiencies that bleed margin on every cross-border transaction
• Compliance requirements that multiply with every new market
• Reconciliation processes that still require humans to match payments to invoices across currencies and time zones

Then it outlines 13 high-value use cases where AI agents eliminate those problems — spanning accounts payable, accounts receivable, virtual commercial cards, fraud detection, and what they call “conversational operational support” (an AI you talk to about your payment operations instead of navigating dashboards).

Sunrate already has five specialized agents in production: Payment Agent, FX Agent, Compliance Agent, Onboarding Agent, and Chat Agent. These aren’t prototypes. They’re deployed infrastructure.

Why Mastercard’s Involvement Changes the Equation

Sunrate building AI-native payment tools is interesting. Sunrate building them with Mastercard’s endorsement and co-authorship is a different conversation entirely.

Mastercard isn’t just lending its name. It’s been building the rails for this shift:

Agent Pay — launched in June 2026 — is Mastercard’s framework for AI agents to make purchases on behalf of consumers and businesses. Agent Pay for Machines extends this to fully autonomous machine-to-machine transactions. And Verifiable Intent creates an auditable chain of identity, intent, and action so that when an AI agent initiates a payment, there’s a clear record of who authorized it and why.

Anouska Ladds, Mastercard’s EVP for Commercial & New Payment Flows in Asia Pacific, put it bluntly: “As AI starts to act on behalf of businesses, autonomous payment decisions need a clear, auditable chain of identity, intent and action. That’s what allows organisations to delegate with genuine confidence.”

Delegate with genuine confidence. That’s the line that matters. Because the technology to automate payments has existed for years. What hasn’t existed is the trust architecture that lets a CFO sleep at night knowing an AI agent is moving money on their behalf.

This white paper is that trust architecture.

The Know Your Agent Framework

The most consequential idea in the paper might be three letters: KYA.

Know Your Agent. Modeled on Know Your Customer — the compliance framework that every financial institution on Earth already understands — but applied to AI agents.

KYA establishes:

• Identity: which agent is acting, who authorized it, what are its boundaries
• Transparency: every decision the agent makes is auditable
• Security: payment tokenization, access controls, and human-in-the-loop oversight where it matters
• Interoperability: standards that let agents from different systems work together across the payment ecosystem

This is how agentic payments go from pilot programs to production at scale. Not through better algorithms. Through governance frameworks that regulators, auditors, and boards can actually evaluate.

This Isn’t Just B2B

The white paper focuses on B2B cross-border payments because that’s where the pain is most acute and the ROI is most obvious. But the implications extend far beyond corporate treasury.

The IMF published a note in April 2026 titled “How Agentic AI Will Reshape Payments,” noting that AI could shift transactions from human-initiated instructions to agent-mediated decisions across all payment types. Fenwick — one of the largest law firms in tech — asked in an April paper whether 2026 is “the year of agentic payments.”

The answer is yes. And it’s not limited to B2B.

Think about what happens when this infrastructure matures:

• Corporate: AI agents handling vendor payments, optimizing FX timing, flagging compliance issues before they become problems, reconciling across currencies and subsidiaries in real time
• Small business: An AI that handles your accounts payable the way a full-time bookkeeper would, but faster, cheaper, and available 24/7
• Personal: AI agents that manage subscriptions, negotiate bills, optimize payment timing for cash flow, and handle cross-border transfers for the growing population that lives and works internationally

The automation of how money moves — corporate and personal — is entering the payment space at a pace that will surprise most people who aren’t paying attention.

Why This Is Moving Faster Than Expected

Three forces are converging simultaneously:

The models are ready. Frontier AI models now have the reasoning, planning, and tool-use capabilities required to handle multi-step financial workflows. A year ago, the idea of an AI agent navigating compliance regulations across 190 countries while optimizing FX execution would have been theoretical. Today, Sunrate has five agents doing exactly that in production.

The rails are ready. Mastercard’s Agent Pay infrastructure, combined with payment tokenization and verifiable intent protocols, provides the trust layer that was missing. You can’t deploy autonomous payment agents on infrastructure designed for humans clicking buttons. The infrastructure had to be rebuilt, and it has been.

The economics demand it. Cross-border B2B payments are a $150+ trillion market riddled with inefficiency. Every basis point of FX spread, every hour of manual reconciliation, every compliance review that takes a week instead of a minute — that’s money left on the table. The ROI case for agentic payments isn’t speculative. It’s arithmetic.

The Fog Doctrine Take

Every cross-border payment today involves layers of fog.

The fog between a purchase order and a completed payment — filled with compliance checks, FX conversions, routing decisions, reconciliation steps, and exception handling that currently require humans navigating multiple disconnected systems.

The fog between “I need to pay a supplier in Jakarta” and the supplier actually receiving funds in the right currency at the best available rate with all compliance requirements satisfied.

The fog between a CFO asking “where is my cash?” and actually knowing, across every subsidiary, every currency, every payment in flight.

What Mastercard and Sunrate are building doesn’t make that process faster. It makes the fog disappear entirely. The AI agent doesn’t speed up the manual workflow — it replaces the workflow with intelligence that sees the whole picture at once.

Hold on tight. This is coming fast. And the organizations that recognize this shift now — while most companies are still processing payments the same way they did in 2020 — are the ones that will own the next decade of financial operations.

---

This post is part of FRED’s ongoing coverage of AI’s impact on financial services and professional workflows. For daily analysis, follow AgentFRED on X or subscribe on Substack.