Blog
Agentic commerce rewrites payment choice
Be the payment option agents can see—and customers can trust
5-minute read
May 20, 2026
In 2030, we estimate that more than 30% of online commerce could run through AI agents, representing close to $3.1 trillion in transactions.1 Those projections should immediately focus the minds of every payments leader on strategy. Why? Because the entire logic of how payments compete is about to change.
For decades, payments platforms mattered because they converted human intent into completed transactions. But they didn’t influence what a customer chose to buy. Now they do. As people enable AI agents to take control of the purchase journey, agents can compress every step from discovery to checkout into a single, optimized and automated decision layer, including payment selection.
Specifically, based on the shopper’s and/or merchant’s preferences, agents are expected to route payments toward the most cost-effective or rewarding option—autonomously optimizing payment method selection. The implications for payments players are significant. If half of the transactions currently routed through higher-cost payment methods such as cards are redirected to lower-cost alternatives, the impact could reduce revenues by $7.2 billion.2
As a result, the playing field across payment methods is simultaneously more level and more competitive, spanning account-to-account (A2A) payments, pay-by-bank solutions, traditional transfers and emerging payment forms such as stablecoins and central bank digital currencies (CBDCs). The question for payments leaders is no longer just “how do we convert?” It is: “How do we become the payment option that agents see, trust, value and select?”
The evolving landscape of agentic payments
With agentic payments, people define the parameters that govern how money can be spent, including limits, permissions and preferences. Agents interpret these parameters, evaluate options and execute transactions when conditions align.
Today, a wave of consumer payment agents is growing, from Google’s AI-powered shopping experiences to PayPal enabling AI assistants by integrating its wallet with secure agent token systems. In parallel, commercial use cases are advancing quickly in more controlled environments, where platforms Coupa, Oracle and SAP already support automated procurement and payment within predefined enterprise policies. These examples trace a clear maturity curve: from assist, to delegate, to autonomous agents executing payments within defined guardrails.
As agentic payments mature, several key shifts occur at the same time:
- Discovery moves upstream. Agents evaluate payment choices in parallel with product features, surfacing new payment options and preselecting an optimized method early in the workflow.
- Payment choice becomes algorithmic. Selection is optimized across rewards, fees, constraints and consumer preferences, with agents dynamically selecting or recommending new payment products based on context.
- Data flows first through the agent layer. The AI agents that sit between consumers, merchants and financial institutions may review, interpret or act on transaction insights before passing them to merchants or financial institutions.
That means payment providers and their methods like cards and wallets can no longer depend on visibility at checkout. Suddenly, success depends less on influencing human behavior directly at checkout and more on aligning with how agents evaluate and execute purchasing decisions. Figure 1 shows how the customer role shrinks as agents compress the buying process.
Figure 1. Agentic commerce reshapes the payments flow
Already, payments players are developing agent-delegated journeys that range from simpler individual transactions to more complex scenarios such as bundled purchases, parameter-based buying and agent-to-agent transactions. Soon these will incorporate additional financial services such as lending, alongside alternative payment methods such as A2A and stablecoins (see Figure 2).
Figure 2. AI Agents expanding into new use cases
The protocols challenge
Agentic payment protocols still lack standardized and mature rules, creating challenges for governance and auditability. Today, organizations are stitching together solutions instead of building true end-to-end auditable frameworks. They rely on API-based authorization, stored credentials, pre-approved rules and human checkpoints, where authority remains static rather than dynamically verifiable.
In practice, what’s needed is a ‘chain of intent’ that translates legal authorization into cryptographic constraints that people can audit and enforce in real time. Once established, that chain of intent allows people to delegate payment authority to agents, so that legal authorization is “always on,” with agents initiating and completing transactions without needing human approval at each step. People stay in the lead, monitoring results and handling exceptions, but agents take on the repetitive work.
The legacy challenge
To remain unmissable and selectable, payments providers also need to ensure that their capabilities integrate into the environments and flows where agents make decisions. The key constraint here is legacy infrastructure. Most payments stacks were not built for real-time, continuous, API-driven flows. Today, 7 out of 10 banks are not fully prepared to support agentic payments; legacy systems lack the flexibility required to integrate into agent ecosystems.3 And agent-enabled procurement and vendor workflows create new dynamics, namely:
- Transaction frequency increases. Payments shift from discrete events to continuous flows
- Execution becomes automated. Rules and data, not manual inputs, trigger decisions
- Speed accelerates. Decision cycles compress from minutes to milliseconds
Encouragingly, the market is responding: Half of payment institutions are already upgrading infrastructure to support the expected increase in agent-initiated transactions.3 To compete, organizations should invest in real-time processing, API-first architectures and scalable transaction capabilities to meet the demands of agentic payments.
Trust as table stakes
Delegating the payment decision is the least likely element in the purchase process to achieve full autonomy in the near term. Just 12% of consumers are willing to allow agents to handle payment choices today, according to our research.4
But those numbers will change, and soon if payment organizations are proactive about strengthening the trust consumers have in them through transparency, control and auditability.
Many people already accept a constrained form of delegation, through mechanisms like automatic bill pay, where payments are executed under explicit consent, predefined rules and the ability to intervene. As trust increases, people will enable agentic payments to extend this familiar model from fixed schedules to more dynamic decision-making.
A critical watch-out for payments leaders: Agent-mediated payment transactions introduce new risks that can erode consumer trust, including agent impersonation, prompt manipulation, synthetic identities and automated fraud at machine speed. Fraud may shift from impersonating cardholders to impersonating agents and workflows. A large majority of financial payments leaders (78%) believe that fraud will increase significantly with the rise of agentic payments. Even more (87%) believe that trust will be the key barrier to agentic payment adoption. Yet 60% still rely on standard workflows to investigate agent-driven fraud.3
Organizations that enable identity verification, permissions and constraints, auditability and real-time anomaly detection are more likely to attract agents. Providers that lack these capabilities risk being bypassed.
Liability is a related issue. When an autonomous agent executes a disputed transaction, responsibility may be shared across agent providers, merchants, financial institutions and payment networks. Standards and protocols are emerging to enable interoperability and formalize interactions between agents and merchants/payment providers. But currently, perspectives vary across the industry. Some stakeholders view financial institutions as the primary control point, while others assign responsibility to merchants or platforms that deploy agents.
Resolving these disparate views will mark a critical milestone. Agentic payments demand ecosystem-wide rules for identity, consent, authorization and dispute resolution, not isolated controls within a single organization.
Future of Money
78%
of financial payments leaders believe that fraud will increase significantly due to agentic payments
87%
believe that trust will be the most significant barrier to agentic payments adoption
60%
do not have a dedicated response plan with forensic tools to investigate agent-driven fraud
Strategic plays as the agentic landscape develops
Looking ahead, three distinct strategic roles are emerging for payments players. How each role aligns to an organization’s strengths and ambitions will shape where it can compete and win.
01
Own the agent or become the payment choice
Some players will build their own customer-facing agent. This works when a company has deep category authority, controls meaningful first-party data, can orchestrate end-to-end commerce reliably and offers differentiated capabilities such as embedded financial advisory or treasury services for commercial customers. Alibaba is a strong example, leveraging its tightly integrated ecosystem across Taobao, Alipay and its Qwen AI assistant to enable end-to-end, agent-driven commerce where users can search, decide and complete transactions seamlessly within a single platform.
For most payment companies, success depends on embedding payment capabilities into agent workflows and becoming generative engine optimization (GEO)-ready. To become the payment choice, providers should compete on the factors agents optimize for: speed, cost, rewards, reliability and integration. Adyen and Stripe illustrate this strategy, positioning themselves as payment infrastructure providers for agentic commerce through AI integrations, agent-ready APIs and orchestration capabilities focused on speed, reliability and seamless execution.
02
Enable a trusted infrastructure for agentic payments
Payments infrastructure players need to support different protocols for all payment types, ensuring security and interoperability between existing and future protocols and networks. These players will enable choice across money types and networks, and build new capabilities, including initiation, multi-rail access, global acceptance, developer-first APIs, tokenization and credentials management.
Visa, Mastercard and American Express are strong examples, leveraging their global networks, APIs and tokenization capabilities to position themselves as trusted infrastructure layers for agent-driven commerce, including enabling secure AI-initiated transactions at scale. Some banks will participate in this area, but without API-first connectivity and real-time rails, they will struggle to compete.
03
Own the control layer
This layer governs how money is spent through rules, permissions and delegated authorization, supported by consent management, identity verification (user and agent), fraud controls, compliance and liability handling. Control also depends on seamless data sharing across ecosystem participants to manage fraud and maximize customer value.
Banks are well positioned here, anchoring trust through identity, account ownership, regulatory compliance and risk management. Card networks can act through tokenization, authentication and fraud monitoring that standardize how credentials are used across merchants and ecosystems. As wallets evolve, they, too, can play a role in the control layer: their proximity to users allows them to enforce granular permissions, authentication and spend controls when agents access payment credentials.
Where to focus next
Across these plays, the core capability stack remains consistent. Those that build the foundations will be best positioned to integrate, scale and compete in agent-driven ecosystems.
This list is non-exhaustive.
Think of these moves as strategic enablers. Organizations that make payments programmable and trustworthy will not only capture value now, but also better position themselves to identify and pursue new value pools as agents reshape the economics of payments.
They will be the first to spot and capture value across the different layers of the agentic commerce flow, for example (as mentioned earlier) by embedding financing where agents detect need. In the control layer, they may charge for identity, authorization, consent and liability management. Or, they may offer routing and optimization services as paid APIs to merchants and agents.
They will also be better able to identify emerging value pools in infrastructure (multi-rail access) and in data and performance signals that agents use when choosing providers. In ecosystem distribution, through partnerships that secure preferred placement in agent platforms. And in value-added services such as reconciliation and dispute management.
Agentic payments isn’t a “someday” scenario. It’s unfolding now and moving payments from the last step in the purchase process into the thick of it. It’s up to payments leaders to solidify their foundations and take bold steps toward capturing an expanded role and a profitable future.
Additional contributions from Hannes Fourie, Accenture Research, Payments Lead
Sources
1 Accenture estimate based on Global Data
2 Accenture estimate based on Global Data, using 30% of ecommerce transaction value multiplied by average interchange fees for card transactions globally
3 Accenture, Future of Money Survey, October 2025
4 Accenture, Consumer Pulse Research, January 2026