Agentic commerce: when money starts thinking for itself

Sulabh Agarwal, Managing Director and Global Head of Payments at Accenture, highlights the imminence of agentic payments and what must be done to ensure a smooth transition.

For years, the payments industry has focused on speed. Real time rails, instant settlement, and low latency have reshaped customer expectations. But the most important change ahead is not about moving money faster. It is about money moving more intelligently.

AI is shifting from the edges of financial services into the centre. Autonomous financial agents are beginning to make decisions about spending, saving, and investment on behalf of both individuals and businesses. This move from automation to genuine autonomy signals a fundamental change in how value flows through the economy. Increasingly, transactions will be initiated, checked, and completed by systems that interpret user preferences, real time data, and market signals without waiting for a human to act.

This change is already under way. In corporate settings, AI agents are starting to manage vendor payments, spot fraud, handle recurring billing, trigger chargebacks, and support liquidity planning. For consumers, similar tools will soon help households run their finances in the background, optimising bills and subscriptions as conditions change.

What makes this shift so significant is the relocation of intent. Payments infrastructure has always assumed that a person decides to make a payment. In an agentic world, that decision point becomes more diffuse. When a system triggers a payment based on a data pattern or a policy, it is no longer straightforward to say who the initiator is. That uncertainty raises broader questions about responsibility and trust.

Infrastructure designed for humans must now support machines

Most financial systems are built around predictable human behaviour. They assume clear moments of authorisation, defined lines of liability and transaction volumes that broadly follow human routines. Agentic commerce challenges these foundations.

First, volume patterns will change. Machine-initiated payments can occur in rapid bursts driven by data events rather than business hours. Several institutions upgrading their systems report that handling these spikes is becoming a key factor in resilience planning.

Second, governance needs to evolve. AI agents will follow rules, but unintended outcomes are inevitable. Institutions must be able to explain how these systems make decisions and provide reliable ways to audit and override them. Without transparent logic, confidence will deteriorate quickly.

Third, the industry must rethink consent. Traditional authorisation is a one-off action. Autonomous systems require ongoing permissions and safeguards that let users delegate authority safely, adapt it easily, and withdraw it entirely.

A regulatory framework built around human triggers will need to adapt

Much of today’s regulation rests on the idea that a human triggers every payment. Consumer protection, anti-money laundering rules and fraud oversight are all based on that assumption. As autonomous transactions grow, regulators will need to decide how existing obligations apply when no individual directly approves a transfer.

Several principles are already emerging, and systems will need to be explainable. Decision trails must be auditable. Customers must stay in control, even if they choose to let an intelligent agent act for them. At the same time, new fraud risks will appear, as criminals test the weaknesses of automated logic and synthetic identities. Yet AI also provides improved defence, from proactive anomaly detection to faster blocking of risky transactions.

A clearer technical foundation is emerging

The shift to agentic commerce depends on data that machines can interpret consistently. Richer transaction standards such as ISO 20022 and advances in tokenisation are creating the building blocks for safe, conditional, and programmable payments. These standards allow financial agents to interact with a level of precision that older formats simply could not support.

But technology leaders recognise that standards alone will not deliver autonomy. Many institutions still rely on legacy middleware, outdated APIs and compliance processes that cannot support fast, intelligent decision making. Some will manage through targeted upgrades. Others will need more substantial modernisation to build the intelligence layer required for autonomous activity.

The strategic question is whether institutions are ready

Having real time capability is no longer a differentiator. What will matter now is the intelligence that sits above it. Institutions that invest in AI-ready payment architectures, with adaptive controls and clear decision logic, will help set the norms of autonomous finance and capture early advantage.

The commercial potential is considerable. Agentic commerce could reduce friction in trade and treasury operations, continuously optimise liquidity, and support new types of programmable value exchange. Corporate users already see value in use cases such as automated vendor payment, supply chain finance, or usage-based billing. These are not marginal improvements. They can turn payments from a back-office process into a genuine driver of business performance.

Trust remains the decisive factor

Despite the enthusiasm around autonomy, the success of agentic commerce depends on trust. Both consumers and corporates want clarity on how decisions are made, how mistakes are prevented and who ultimately stands behind the outcome. Many expect regulated financial institutions to play a leading role in securing autonomous systems. This gives banks a natural advantage, but only if they reinforce that trust with transparency and strong safeguards.

The winners in this new landscape will be those who pair innovation with clarity and reliability. Customers will not simply want to know that a payment occurred. They will want to know why it occurred, on what basis, and with what protections.

A near-term reality

Agentic payments are approaching quickly. This is no longer a distant vision but something that will influence strategies and investment decisions within the next planning cycle. Machines will transact on our behalf whether the industry prepares for it or not. The real test is whether we can build the governance, trust, and intelligence needed to ensure this future is safe, resilient and widely accepted.

About the author

Sulabh Agarwal leads the Global Payments Practice at Accenture. He helps financial institutions and other organisations across the world navigate the shifting payments landscape, combat lost payments revenue, and control and mitigate operational risk throughout the payments value chain. Sulabh has led engagements in strategy, consulting, and transformations focused on payments.

 

About the company

Accenture is a leading solutions and global professional services company that helps the world’s leading enterprises reinvent by building their digital core and unleashing the power of AI to create value at speed across the enterprise, bringing together its proprietary assets and platforms, and deep ecosystem relationships.