The rules governing how AI agents spend money don’t exist yet. A new coalition launched today is working to fix that before it becomes a real problem.

Rain, a stablecoin payments startup, announced the Agentic Payments Alliance (APA) on August 18, 2026, bringing together 26 organizations to build the infrastructure standards for a world where AI agents autonomously execute financial transactions on behalf of businesses and individuals.

Founding members include Visa, Mastercard, Fiserv, Circle, Solana, Remitly, Fireblocks, Uniswap Labs, Shift4, Chainalysis, and more than a dozen others across traditional payments, crypto, and fintech. The coalition is structured as a working collective, run by its members rather than controlled by Rain alone.

Why This Coalition Exists Now

AI agents are already buying things. Businesses are deploying autonomous agents to book travel, reorder supplies, pay invoices, and manage subscriptions. But the payment infrastructure handling those transactions was built for humans, not machines.

When a human clicks “buy,” identity is established through login credentials, device trust, behavioral patterns, and sometimes multi-factor authentication. When an AI agent executes a purchase, none of those signals apply cleanly. The agent may be running on a server. It may be acting on behalf of multiple users simultaneously. It may be operating across currencies, countries, and payment rails that have never seen this type of actor before.

The APA was formed because those infrastructure decisions are being made right now, often by individual companies working in isolation. Before the standards become locked in by incumbents or accident, the Alliance wants every major player in the same room.

Three Problems the Alliance Is Solving

Early work will focus on:

Agent authorization. How does a merchant or payment network verify that an AI agent genuinely has permission to make a purchase? The current answer is: it largely doesn’t. Most authorization frameworks assume the entity initiating a transaction is a human with a card or bank account tied to an identity. AI agents operating on delegated authority from a business need a different model.

Fraud detection. AI-driven purchasing patterns look nothing like human ones. An agent booking 40 flights in a row for a corporate travel program is legitimate. An agent cycling through payment methods trying to find one that clears is fraud. Existing fraud systems were trained on human behavior and often can’t tell the difference.

Loyalty and rewards portability. If an AI agent books a hotel on behalf of an employee, where do the points go? Does the agent even know a loyalty program exists? As agents handle more of the transaction layer, the rules around incentive structures need to catch up.

These might sound like edge cases. They aren’t. McKinsey projects between $3 trillion and $5 trillion in global agentic commerce by 2030.

What the Alliance Will Actually Do

The APA plans to publish shared research and frameworks, test emerging standards for agent identity and authorization, and engage regulators on the questions agentic commerce raises that existing financial law doesn’t cleanly address.

Members get early access to Rain’s Agentic Startup Program, an accelerator for early-stage companies building specifically for this space. The first cohort of five startups will present at a demo day open to Alliance members.

This isn’t a standards body that will produce a finished specification in six months. Think more along the lines of how PCI-DSS emerged from card payment fraud in the early 2000s: slow and contested, but ultimately the foundation the industry built on.

What This Means for Business

If your business uses AI agents for anything that touches purchasing, invoicing, or finance, the authorization question will arrive sooner than expected.

The most immediate risk is authorization confusion. When an AI agent inside your accounts payable workflow tries to cut a check for an unusual amount, does the payment processor pause it? Does your bank flag it as suspicious? Does it clear and nobody notices for 90 days? These aren’t hypothetical scenarios for companies actively running AI-driven financial operations.

Practically, this creates three action items for business leaders:

1. Audit what financial authority your agents actually have. Most businesses that have deployed AI agents haven’t explicitly defined what those agents can and can’t spend. That gap is a security surface.

2. Track the standards that emerge from the APA. The frameworks this coalition produces will likely become the compliance baseline for enterprise AI payments. Building systems that align early means less rework when requirements solidify.

3. Treat agent identity like any other financial control. If an agent can authorize transactions, it needs access controls, audit logs, and spending limits. Most current AI governance frameworks don’t go that far.

One note of caution: the Alliance’s founding members include the companies that build and sell the payment rails. That gives them obvious incentives in how standards get written. This is how most infrastructure standards get built, and it doesn’t make the output bad. But independent businesses should engage with what emerges critically rather than treating it as neutral technical specification.

The fact that Visa and Mastercard joined before the first meeting is a signal that this problem is being taken seriously at the highest level of the payments industry. Whether the APA produces meaningful standards in two years or five, the agentic payments infrastructure gap will need closing. This is the first organized attempt to do it.

---

Enterprise DNA helps business leaders understand and act on developments like this one. Omni Ops helps you build and govern AI agent workflows that are ready for the infrastructure changes ahead. Book a discovery call to talk through what this means for your operations.