Agentic commerce is moving from concept to infrastructure. The important shift is not simply that AI assistants may shop on behalf of consumers or businesses. It is that payments, identity, fraud controls, and merchant acceptance rules now need to distinguish between three actors in every transaction: the customer, the merchant, and the agent acting under delegated authority.

That makes 2026 a trust-infrastructure year for payments.

Visa, Mastercard, Stripe, Google, PayPal, Coinbase, and others are pushing agent-enabled commerce frameworks, while industry debate is quickly moving toward “know your agent” models and transaction standards. Visa and Mastercard are both competing and collaborating around agentic shopping standards, reflecting how strategically important this layer has become for networks and platforms. Sources such as Payments Dive and the IMF have described this as a fast-emerging standards race, not just a product feature race.

The commercial opportunity is clear. If AI agents can safely search, compare, negotiate, authenticate, and pay, merchants could see higher conversion, lower checkout friction, and more personalized purchasing journeys. For platforms, agentic commerce creates a new control point in the buying journey. For issuers and networks, it creates a new role: validating not only the payment credential, but also whether the agent was authorized to use it.

But the risk is equally clear. Agentic payments collapse several historically separate questions into one real-time decision: Is this the right user? Is this a legitimate merchant? Is this agent authorized? Is the purchase within policy? Is the behavior normal? Is the customer being manipulated?

That is why agentic commerce cannot scale on payment tokens alone. It needs identity, intent, and permissioning.

Digital identity wallets are becoming part of the answer. Google recently announced updates to Google Wallet and Google Pay, including expanded digital ID support, age verification, and direct checkout capabilities. The EU Digital Identity Wallet program and other verified credential initiatives point in the same direction: reusable, privacy-preserving identity signals that can reduce friction while giving merchants and financial institutions stronger assurance.

Fraud strategy also has to evolve. AI-enabled scams, synthetic identities, deepfakes, account takeover, and fraud-as-a-service are making static authentication weaker. Mastercard has emphasized real-time AI-based fraud detection to improve authorization decisions and reduce false positives. Thomson Reuters has also highlighted the 2026 fraud challenge as multi-channel, behavioral, and increasingly industrialized.

The second-order effect is important: faster payments and invisible checkout make fraud harder to unwind. In an agentic environment, the best risk systems will not simply approve or decline. They will adapt the journey: silent approval for low-risk behavior, stepped-up verification for uncertainty, spending limits for delegated agents, and human review when intent is unclear.

For payment leaders, the strategic question is no longer “Will AI agents buy things?” It is “Who will control the trust layer when they do?”

What Payment Executives Should Do Now

1. Define agent acceptance rules before volume arrives. Decide what an authorized agent can buy, at what limits, with what disclosures, and under what liability model.
2. Treat identity as dynamic, not a one-time onboarding event. Verified credentials, device signals, behavioral history, and consent records will matter more than static KYC files.
3. Upgrade fraud controls from transaction scoring to intent scoring. The goal is to understand whether the customer, agent, merchant, and purchase context align.
4. Pressure vendors on interoperability. Avoid being locked into a single agent, wallet, network, or orchestration layer before standards settle.
5. Prepare customer experience teams for new disputes. “My agent bought the wrong thing” may become a real service and liability category.

The winners in agentic commerce will not be the firms with the flashiest AI demo. They will be the ones that make delegated purchasing feel safe, explainable, reversible when necessary, and commercially useful at scale.

Sources and references used: Payments Dive, IMF, Google Wallet coverage, Mastercard fraud analysis, Thomson Reuters fraud trends.

Peter Jonathan Wilcheck
AI/ML Engine, Machine Learning, Deep Learning, Data and MLOPs