The global retail payments infrastructure has spent nearly a decade perfecting the unified commerce playbook. However, a disruption has arrived to challenge the foundational architecture of the ecosystem: agentic commerce. As autonomous artificial intelligence agents transition from basic conversational search tools into transactional entities capable of independently evaluating products, constructing carts, and executing payments on behalf of humans, the core definition of the consumer journey is changing.
A global research study commissioned by Visa Acceptance Solutions and conducted by PYMNTS Intelligence exposed a profound gap in industry readiness. Surveying 75 major payment acquirers across the United States, the United Arab Emirates, and Brazil from January 16 to January 30, 2026, the data revealed that while technology providers push for rapid automation, the institutional gatekeepers of the global payments grid are intentionally flashing a yellow light. The move from human-initiated to agent-initiated transactions represents an optimization challenge requiring an overhaul of rules, identity frameworks, and risk systems.
The Existing Foundation: Unfinished Omnichannel Homework Stalls the Future
The transition toward agentic commerce is heavily constrained by an existing, unresolved market reality: the persistent friction inside traditional unified commerce infrastructure. While eight in ten payment acquirers report that they possess the internal technical capability to offer seamless digital-to-physical shopping solutions, scaled merchant deployment remains unevenly distributed.
The audited transaction data highlights a steep disconnect between payment rails and real-world merchant adoption:
- The Integration Divide: Globally, only 32% of merchant clients currently operate an identical payment experience across their digital and physical storefronts, leaving nearly 70% of the retail landscape fragmented.
- The Structural Bottlenecks: This slow deployment is driven entirely by modernization costs and integration timelines, rather than a lack of market demand. Half of all surveyed acquirers (51%) explicitly state that the sheer expense of implementation holds back their merchant clients, while 41% cite implementation time as the primary operational blocker.
- The Regional Realities: These cost barriers are particularly acute in the UAE, where 68% of acquirers cite integration costs and 64% cite prolonged implementation timelines. In the United States, the metrics remain similarly binding, with 52% of acquirers citing cost bottlenecks and 44% citing timing constraints. Furthermore, 39% of US operators pinpoint cross-channel payment infrastructure fragmentation as a direct internal barrier.
This data establishes that the current merchant ecosystem is still struggling to link basic human-to-human physical and digital flows. Forcing non-human, autonomous AI agents onto this legacy infrastructure risks multiplying existing fraud exposures, tokenization failures, and point-of-sale operational limitations.
The Present Dilemma: Rational Hesitation Over Distributed Trust
Because agentic flows challenge historical assumptions regarding consent and transaction authorization, the vast majority of global payment gatekeepers are actively resisting a blind rush into market scaling. The underlying data reveals a heavily fractured market sentiment, with only a small first-mover cohort of 23% proactively executing strategies to actively accelerate customer and merchant readiness.
The remaining 77% of the market has adopted a highly conditional, wait-and-see response to agentic commerce. This cautious segment is divided almost equally among three distinct strategies, with 27% explicitly waiting until consumer demand signals clarify, 24% holding back until they receive clear demand from their merchant clients, and the final 27% opting to adapt selectively only after following the leadership of dominant business platforms and large merchants.
This widespread hesitation represents a rational corporate response to severe structural and policy uncertainties. Half of all surveyed organizations (49%) state that scaling agent-initiated commerce is impossible without explicit regulatory and compliance certainty.
Furthermore, 45% of the market notes that traditional liability frameworks must be entirely redesigned to account for situations where an AI agent acts out of scope or executes an erroneous transaction—a demand that spikes to 56% among infrastructure-mature operators in the United States.
The industry is currently comfortable evaluating non-final, easily reversible advisory use cases like automated cart building (56% internal evaluation) and real-time inventory exposure (56%). However, when it comes to true autonomous execution, such as end-to-end agent-led fulfillment orchestration, internal evaluation drops significantly to 32%, reflecting deep anxiety over unmapped operational liabilities.
The Shared Governance Allocation: Who Bears the Risk?
A critical point of discussion across corporate boards is the chaotic diffusion of responsibility regarding who validates, authenticates, and secures an agentic transaction. Financial networks, issuing banks, acquirers, and AI software providers are locked in a complex negotiation over risk attribution.
According to the Visa study, the perceived allocation of operational responsibility breaks down into highly contested silos:
- The Frontline Acquirer Role: Four in ten acquirers (40%) acknowledge that frontline fraud prevention falls squarely within their domain, while 29% accept primary responsibility for credential security and tokenization in agentic flows.
- The Network Custodianship: Payment networks are viewed as the absolute anchors for credential integrity, with 44% of the industry assigning them primary ownership over automated token security and ecosystem standard-setting.
- The Merchant Burden: Retailers are completely isolated when it comes to post-transaction friction, with an overwhelming 60% of acquirers stating that merchants must retain absolute liability for refunds, returns, and automated chargebacks.
- The AI Provider Vacuum: AI agent providers are assigned zero responsibility for financial fraud or transaction disputes, with their operational expectations strictly confined to providing decision logic explainability (52%) and assisting in basic subscription lifecycle tracking (31%).
This uneven distribution of risk proves that while AI platforms generate the inspiration and the purchase intent, they are legally insulated from the transactional fallout. If an autonomous agent triggers a fraudulent or corrupted bulk purchase order, the financial liability is mathematically funneled back onto the merchant and the acquiring bank.
The Future Solution: The Phased Roadmap to Automated Settlement
To transform this fragmented landscape into a commercially viable, trusted transaction engine, the financial ecosystem cannot deploy automation simultaneously. The report outlines a strict, sequential roadmap that moves from foundational policy to autonomous scaling over a multi-year horizon.
During Phase One, spanning the first twelve months, the market is forced to address policy and identity codification before a single line of automated settlement code is deployed at scale. This stage requires the industry to address regulatory uncertainty and establish binding identity verification standards for AI agents—a capability cited as an immediate priority by 43% of the industry and 64% of UAE acquirers. Authentication rules built around the assumption of a physical human presence must be rewritten to incorporate verified machine signatures.
In Phase Two, running from months twelve to twenty-four, technical teams must transition their capital toward building the operational trust infrastructure and non-human risk engineering. This requires upgrading legacy fraud-scoring models to evaluate non-human behavior patterns (a top 40% industry mandate) and deploying centralized governance dashboards to manage spending thresholds, merchant category restrictions, and time-bound consumer consent permissions.
By Phase Three, covering months twenty-four to thirty-six, the technical bottleneck shifts toward cross-actor orchestration and real-time token infrastructure. Capital must be directed toward launching interoperable message standards and reference APIs that span issuers, networks, and AI providers, ensuring real-time tokenization is dynamically supported across channels to preserve credential security during automated multi-merchant hops.
Finally, in Phase Four, extending beyond thirty-six months, true autonomous execution becomes durable as the ecosystem achieves end-to-end reversibility. Acquirers will unlock complex, high-value use cases like automated subscription lifecycle optimization and multi-merchant cross-business checkout coordination solely when dispute-resolution models are backed by automated evidence bundles, including consent logs and explicit intent metadata.
Strategic Actions for Corporate Financial Boards
The data compiled by Visa highlights that treating agentic commerce as a simple front-end software plugin creates severe financial exposure. To insulate transaction pipelines from unmapped liability, enterprise retail and fintech boards must execute three immediate pivots:
- Halt Autonomous Scripting Until Liability Frameworks Are Codified: Corporate governance teams must actively discourage the deployment of automated, agent-led purchase executions that lack human-in-the-loop authorization. Keep technical features strictly limited to reversible, advisory actions like cart building and real-time inventory exposure until payment networks and local regulators issue standardized operating rules regarding non-human transaction liability.
- Reallocate Capital from Marketing Gimmicks to Mid-Mile Risk Upgrades: Redirect engineering capex away from customer-facing conversational chatbots. Direct technical infrastructure teams to build deep-tech back-of-house capabilities: upgrading legacy fraud-risk scoring to analyze non-human machine behavior patterns, and constructing unified data backbones to mitigate cross-channel identity fragmentation.
- Mandate Machine Identity Binding Standards for All Partners: When negotiating integrations with external AI providers and business platforms, refuse models that rely on generic consumer credentials. Enforce a strict requirement for machine identity-verification and credential-binding protocols, ensuring every agentic transaction carries an unalterable, auditable cryptographic token that explicitly maps delegation scope and data-governance boundaries.
As retail continues to evolve across markets, the ideas shaping its future are increasingly being defined through global industry dialogue. Retail World Forum & Awards brings together senior retail leaders, technology innovators, and ecosystem stakeholders across high-growth markets to explore the strategies and innovations driving modern commerce—alongside a global awards platform. To partner, speak, or attend, log on to retailworldforum.com
Sources: Visa Acceptance Solutions “How Acquirers Prepare for Agentic Commerce” Report; PYMNTS Intelligence Multi-Country Cross-Channel Payments Survey Logs (March 2026); Global E-Commerce Architecture Risk Assessments.
