December 19, 2025
Agentic commerce means better, more personalized shopping experiences for consumers. It also means growth opportunities for financial institutions through more interchange fees, finance charges, and stickier customer experiences.
But this growth will continue to be constrained by undefined compliance and risk frameworks.
In our last piece, we covered the basics of agentic commerce and its main risks for issuers. Here, we’re breaking down potential solutions.
The Race to a New Standard: Emerging Protocols for Agentic Commerce
Visa recently recorded a 25% increase in malicious bot-initiated transactions globally, including a 40% jump in the U.S. It’s clear that adapting to fraud committed by agents is imperative, but how organizations adapt is still up for debate.
There are several emerging trends across solutions. For one, many in the industry are replicating familiar frameworks like KYC and KYB for agents. KYA (Know Your Agent) establishes a traceable digital identity for the agent and continuously monitors its real-time behavior against its declared scope and permissions to detect and flag malicious or non-compliant activity.
That said, even the biggest players are taking drastically different approaches to address risks in agentic commerce:
- Visa (TAP) verifies the trusted “who” behind the agent
- Google (AP2) proves intent cryptographically
- Mastercard (Agent Pay) restricts actions through tokens
- Stripe & OpenAI (ACP) provides infrastructure to reduce friction and ambiguity
Turning Risk into Reward: Agentic Commerce Risk Management
Most of the current solutions target the merchant and card network pieces of the puzzle, but it is worth examining what happens when the transaction hits the financial institution through the payment portal.
At the authorization stage, the decision is approve, decline, or refer based on confidence using the signals available in real time. Most payment processors use simple rules (e.g., velocity, card status, etc) to determine which transactions will then be evaluated for fraud.
But once an agent is compromised or manipulated, it can mimic the behavior of a trusted user operating at machine speed. Transactions are clean, consistent, and low-friction, which removes many of the behavioral signals issuers rely on today.
Issuers will then be forced to make harder, riskier decisions quickly with less signal. That’s why issuers need agent-aware signals at authorization time.
So the question becomes: how can we ensure that the issuer has sufficient confidence at authorization time? Here are two controls to consider:
- New Agentic Rule Books
Because agent-origin transactions remove or distort human behavioral signals, models trained on historical human behavior underperform at the point of authorization. In early stages of agentic commerce, deterministic authorization rules are more effective than fraud models because data is sparse and behavior shifts quickly. That’s why banks need new rules layers for agent-origin transactions. - New Fraud Controls for CNP (Card Not Present) Transactions
Issuers need new fraud controls designed to detect high-speed hijack attempts, reduce false declines, and optimize referrals by applying explainable risk overrides. The approach must center on adaptive, explainable controls that can move faster than model retraining cycles.
Preparing for What’s Ahead: Faster, Smarter Controls
Agentic commerce does not mean institutions need to invent an entirely new fraud taxonomy. It just means they need agent-aware controls and rules that improve authorization confidence in real time.
Stratyfy provides the missing intelligence layer for agentic commerce, giving processors and issuers the rules, scoring, and override controls needed to scale revenue safely. For more information, email sales@stratyfy.com.