How generative AI is redefining payments

Digital payments have undergone a remarkable transformation over the last decade, yet many of the core challenges remain surprisingly persistent. False declines continue to cost merchants billions. Authentication flows remain rigid and unintuitive. Fraud shifts faster than rule-based systems can adapt. And payment stacks struggle to keep up with rising customer expectations for speed, transparency and security.

Today, generative AI (genAI) is emerging as a technology capable of reframing many of these long-standing challenges. But its significance isn’t simply tied to automation or cost efficiency. In payments, genAI has the potential to transform the intelligence, adaptability and resilience of the entire authentication and risk ecosystem.

The cost of payment failures: A silent crisis

Payment failures rarely make headlines, yet they are one of the most persistent and expensive problems in modern commerce. Recent research illustrates the scale of the issue. A study by Checkout.com found that businesses across the US, UK, France and Germany lost more than $50 billion last year due to legitimate transactions being incorrectly declined — false declines that consumers had no reason to expect.

What makes these losses so damaging is not only their magnitude, but their downstream impact on customer trust. In the same study, nearly 45% of consumers said they would not retry a payment after a false decline and 42% said they would not return to the merchant at all. A single failure at the point of purchase can end a customer relationship entirely.

Separate research from PYMNTS Intelligence estimates that US ecommerce merchants placed roughly $157 billion in sales at risk in 2023 due to false declines and more than half of that revenue may never be recovered despite outreach and remarketing efforts. These numbers challenge the assumption that fraud prevention is the primary source of conversion loss; in many cases, the greater threat is the friction introduced by overly cautious systems misclassifying legitimate buyers.

Authentication introduces friction of its own. Multiple surveys with merchants globally show that checkout interruptions — particularly in the authentication stage — remain a leading reason for cart abandonment. PYMNTS’ annual research on merchant checkout innovation notes that payment friction is consistently ranked as one of the most critical customer-experience problems, with merchants citing approval volatility, excessive authentication steps and processing failures as key contributors.

A separate Riskified survey from 2025 found that 85% of merchants still struggle to balance strong fraud prevention with a seamless customer journey, with many estimating that up to 5% of legitimate orders are falsely declined each year.

At the same time, fraud itself is becoming more sophisticated. Synthetic identities, multi-step social-engineering attacks and coordinated account-takeover patterns continue to rise. The 2025 Global eCommerce Payments and Fraud report from the Merchant Risk Council highlights that merchants increasingly cite fragmented data, outdated tooling and slow model update cycles as some of their most significant vulnerabilities in combating modern fraud.

Taken together, these statistics point to a systemic issue: Payments today break for reasons that span technology, design and data fragmentation — and the cost of those breaks is substantial.

Why authentication systems break — and why rules alone can’t keep up

Authentication sits at the heart of most friction in the payment lifecycle. Historically, authentication logic has been built on prescribed, rule-based flows. These flows are deterministic by design: “If 3-D Secure authentication method fails, retry,” or “If OTP fails, fallback to a different method.” While rooted in good intent, these rigid sequences struggle to accommodate context. They don’t interpret subtle behavioral signals, adjust to issuer preferences or adapt to patterns across millions of similar transactions.

Compounding the problem is the fragmentation of data. Device integrity information, behavioral telemetry, historical success rates, regional requirements and processor feedback often live in isolated systems. Without a unified view of the customer and the transaction, authentication becomes binary and brittle. The result: Customers who should be exempted from authentication experience unnecessary friction, while genuinely risky behavior sometimes slips through simply because the underlying rules are outdated.

Legacy architecture plays a role as well. Many payment stacks still rely on static workflows encoded directly in application logic — code paths that cannot pivot dynamically when issuers change strategies, when new authentication methods emerge or when fraud patterns evolve. The rigidity of these systems means that even minor changes require extensive engineering work, slowing the pace of innovation.

Where genAI changes the trajectory

Generative AI offers a path forward because it excels at pattern recognition, contextual reasoning and dynamic decision-making — all areas where traditional authentication systems fall short. Unlike static rules, genAI can synthesize inputs across the entire payment ecosystem: device metadata, browsing behavior, past transaction outcomes, real-time risk indicators, issuer responsiveness and even subtle anomalies in how a user interacts with the checkout page.

This broader context allows genAI to recommend or autonomously select the most appropriate authentication method for each transaction. A customer who has a strong device fingerprint, a low-risk history and a consistent behavioral pattern might be routed into a frictionless flow. Another customer who shows signs of account takeover risk may be guided into a stronger, step-up authentication like biometric verification. Instead of forcing every customer down the same path, authentication becomes adaptive and individualized.

In risk-decisioning, genAI improves precision by identifying patterns across vast behavioral signals that humans cannot feasibly encode into rules. This capability helps reduce false positives — the leading cause of false declines. Over time, a genAI-driven risk engine learns from every outcome, continually refining its understanding of intent, fraud and customer behavior.

GenAI also transforms the operational side of payments. Because it can analyze logs, error patterns and processor behaviors at scale, it is well-suited to diagnose failures, identify root causes and recommend routing adjustments. In many cases, it can proactively shift traffic away from failing systems or recommend fallback methods that are more likely to succeed — making payment infrastructure inherently more resilient and self-healing.

Why this transformation matters for CIOs

For CIOs, the conversation around genAI cannot be reduced to experimentation or isolated point solutions. Payments sit at the intersection of revenue, customer trust and regulatory scrutiny. Improving authentication outcomes, reducing false declines and increasing approval rates have direct financial impact. Even a modest improvement in approval rates can translate to tens or hundreds of millions of dollars in recaptured revenue for large merchants.

But the benefits go beyond revenue. Adaptive authentication improves customer experience by eliminating unnecessary friction. Better risk scoring strengthens fraud prevention without damaging conversion. More resilient infrastructure reduces operational burden and outage-related loss.

In an era where customer loyalty is fragile and competitive differentiation increasingly depends on seamless digital flows, the intelligence of the payment system becomes a strategic asset.

The road ahead: Building a genAI-ready payment ecosystem

The shift toward genAI-driven payments will not happen overnight. It will require unified data infrastructure, modular orchestration layers and robust AI governance frameworks that ensure transparency and regulatory compliance. But the direction of the industry is clear. Authentication flows will become dynamic rather than static. Risk engines will become adaptive rather than reactive. And payment stacks will evolve from brittle pipelines into intelligent, self-optimizing systems.

The organizations that invest early in this transformation will not only unlock higher approval rates — they will build payment foundations that can evolve with customer behavior, fraud landscapes and regulatory environments. They will be better equipped to deliver the seamless, secure and reliable experiences that modern commerce demands.

Payments are no longer just a backend function. They are a critical touchpoint in the customer journey, a direct driver of revenue and, now, one of the most promising areas for advanced AI to reshape the enterprise.

For CIOs, the opportunity is clear: GenAI is not simply an enhancement to the payment stack — it is the future architecture of authentication, fraud detection and building customer trust.

[The views expressed here are the author’s own and do not represent those of Meta.]

This article is published as part of the Foundry Expert Contributor Network.
Want to join?