At a moment when artificial intelligence agents are beginning to execute purchases, negotiate contracts, and move money with minimal human intervention, the payments industry faces a foundational question: who governs the machine? Wally Mlynarski, chief executive of Elavon — the payments processing subsidiary of U.S. Bank — has a clear answer. Trust, not transaction velocity, will determine which institutions shape the next chapter of commerce.
Speaking to the emerging phenomenon of agentic commerce, Mlynarski laid out a philosophy that runs deliberately counter to the prevailing silicon-valley instinct to move fast and iterate later. Safety and soundness, he argued, are not compliance footnotes to be appended after deployment — they are the architecture upon which any durable agentic payments infrastructure must be built. In an industry where a single breach of consumer confidence can unravel years of market positioning, that is not a conservative stance. It is a strategic one.
What Is Agentic Commerce, and Why Does It Matter Now?
Agentic commerce refers to the capacity of artificial intelligence systems — so-called AI agents — to act autonomously on behalf of users in commercial environments. These agents can browse, compare, select, and pay for goods and services without moment-to-moment human direction. For payments processors, this represents both an enormous opportunity and a structurally novel risk surface. Unlike a human consumer who can pause, question, or reverse a transaction, an AI agent operating within pre-set parameters may execute a series of irreversible financial actions at machine speed. The governance frameworks that underpin traditional card-present and card-not-present transactions were simply not designed with this paradigm in mind.
That gap between the existing regulatory and technical architecture and the demands of autonomous transaction flows is precisely where Mlynarski's emphasis on trust becomes operationally significant. Payments rails will need to be adapted — or entirely re-engineered — to provide identity assurance, fraud detection, and liability clarity in environments where the "customer" making a purchase is software, not a person standing at a terminal or typing into a checkout field.
Paze Signals Institutional Readiness
Against that backdrop, Mlynarski's disclosure that mobile wallet Paze is "ready to do pilots" in the agentic commerce space carries considerable weight. Paze, the bank-backed digital wallet developed as a consortium initiative among major U.S. financial institutions, has positioned itself as a trusted-identity layer for online transactions — a credential that takes on heightened importance when the entity presenting that credential may be an autonomous software process rather than a human account holder.
The readiness to pilot in this space suggests that at least some corners of the incumbent banking infrastructure are moving from theoretical discussion to operational experimentation. Pilots, by their nature, are controlled environments designed to stress-test assumptions before broad deployment. In the context of agentic commerce, that methodical approach aligns precisely with Mlynarski's stated philosophy: establish the trust architecture first, scale the velocity later.
This is a meaningful posture for an institution of Elavon's scale. As a subsidiary of U.S. Bank — one of the largest financial institutions in the United States — Elavon processes payments across a vast merchant base, giving it both the data and the institutional standing to influence how industry norms around agentic transactions take shape. If Elavon's pilots with Paze succeed in demonstrating that AI-agent-initiated transactions can be authenticated, monitored, and reconciled with the same rigor applied to human-initiated payments, the resulting framework could become a reference standard for the broader industry.
The Stakes of Getting This Wrong
The urgency underlying Mlynarski's trust-first argument becomes clearer when one considers the failure modes. An agentic commerce ecosystem built on speed without adequate identity and fraud controls is a high-value target for social engineering at the AI layer — adversarial prompts designed to redirect autonomous agents toward fraudulent transactions. It creates ambiguity around liability when unauthorized purchases occur. And it risks triggering regulatory intervention that could stifle legitimate innovation for years.
The financial services sector has lived through analogous inflection points before. The early rollout of real-time payments networks surfaced persistent questions about irrevocability and fraud recovery that regulators and institutions are still working through. The expansion of buy-now-pay-later products generated consumer protection concerns that ultimately invited congressional scrutiny. Agentic commerce, operating at AI speed and potentially at far greater scale, compresses the timeline in which those lessons must be learned and applied.
What This Means
Mlynarski's positioning of Elavon and Paze at the intersection of institutional trust and agentic innovation reflects a broader truth about technology adoption in regulated industries: the institutions that define the safety standards tend to define the market. By framing the agentic commerce debate around safety and soundness before scale, Elavon is not being cautious — it is being deliberate. For merchants, technology partners, and regulators watching this space, the signal from U.S. Bank's payments arm is worth calibrating against: in agentic commerce, the fastest player does not necessarily win. The most trusted one does.
Written by the editorial team — independent journalism powered by Codego Press.