Rearchitecting the Payments Value Chain for the AI Era

AI is everywhere today. From boardroom discussions to product roadmaps, banks and payment providers are racing to understand how it can unlock growth, improve efficiency, strengthen risk management, and transform customer experiences. Yet one challenge persists: How do organizations move beyond AI pilots to deliver meaningful business outcomes at scale?

To explore this question, Opus Technologies CMO Nischala Murthy sat down with Arvind Ronta, veteran payments executive and Opus Advisory Board Member. Arvind offered expert insights on the evolving role of AI in payments. Having worked at the intersection of systems governance, innovation, and customer experience at Visa and other leading institutions, Arvind brings a practical perspective on what separates AI ambition from AI impact.

Drawing on decades of experience launching card, credit, and embedded finance solutions on a global scale, he cuts through the hype to highlight where AI is delivering tangible value today and where the greatest opportunities lie ahead. He also shares his perspective on a broader industry reality, i.e., while many financial institutions have initiated AI programs, relatively few have successfully scaled them across the enterprise.

The opportunity, however, is significant. According to the Stanford AI Index 2026, financial services stands out among industries for its advanced adoption of AI in core operational areas such as risk and compliance. The next wave of value creation will come from extending AI across customer experience, operations, product innovation, and decision-making.

In this conversation, Arvind shares practical lessons, emerging trends, and strategic considerations for banks, fintechs, and payment providers looking to translate AI investments into sustainable business impact.

Question. AI is one of the most talked-about technologies in payments today. Where are you seeing the most practical and measurable use cases emerge?

Arvind Ronta: The most practical AI use cases in payments today fall into three buckets: fraud and risk decisioning, operational automation, and authorization optimization. On fraud and risk, AI is already embedded at scale in networks and banks, scoring transactions in near-real time using behavioral, device, and transactional signals, such as fingerprints, mouse movements, and shipping and billing patterns, that humans simply cannot synthesize as quickly or as accurately. Banks can use models to automatically approve, decline, or flag these transactions. This is consistent with what we see in the broader market, where AI-powered risk and security tooling is one of the earliest and most mature enterprise applications of AI.

The second bucket is operational automation. Think of a financial institution that used to rely on, say, 50 fraud analysts manually triaging rule-based alerts; with AI, that same organization can now run with maybe ten analysts who focus on exceptions while the models auto-approve 98–99% of legitimate transactions. This is where you start to see measurable improvements in cost-to-serve and cycle times.

The third bucket is what I would call authorization optimization. Here, AI is used to build real-time risk-scoring engines that analyze contextual data across the transaction lifecycle and decide in milliseconds whether to approve, decline, or step up a transaction. The net effect is better authorization rates, fewer false declines, and a smoother experience for customers, which card networks and large processors are already pursuing aggressively.

Question. In an AI-driven payments ecosystem, how can providers create differentiated customer experiences while preserving the trust, security, and compliance that underpin the industry?

Arvind Ronta: Customer service is probably the area where the promise and the pitfalls of AI are most visible. You see state-of-the-art models that can do incredible things, and then the same model fails on a simple, intuitive question. For example, Kevin Mannion, Founder/CEO at Workstuff AI, recently reported on LinkedIn that “Claude just told me that there is no ‘d’ in Monday or Friday” when asked how many days of the week have ‘d’ in them. That is a good reminder that AI has its flaws and in customer-facing use cases, many things can go wrong if you do not design the system carefully.

The balancing act is between speed and personalization on one side and trust, security, and compliance on the other. The companies that are doing this well treat AI as a connective tissue between those two worlds rather than choosing one over the other. For instance, Visa and Mastercard are working on “agent identity” frameworks, sometimes summarized as “know your agent” (KYA), that treat AI agents as first-class identities in the ecosystem, with explicit access controls and permissions.

That kind of framework lets you say: this agent can perform A, B, and C, but if it wants to do D, it must check with a human or obtain additional authorization. On top of that, you can still personalize the customer experience based on how a customer shops and which controls they choose to delegate to the agent, all the while ensuring the agent operates within clearly defined, compliant boundaries. We are still in the early stages of this agentic model of customer experience, but the direction of travel is clear: AI will sit in the middle, mediating between hyper-personalized service and the non-negotiables of trust and regulation.

Question. What are the top AI priorities you are hearing from banks, issuers, acquirers, payment processors, and fintechs today?

Arvind Ronta: If you talk to leaders across banks, issuers, acquirers, processors, and fintechs, a few priorities keep coming up on AI. First, there is a very strong focus on fraud and risk – not just improving detection rates, but doing real-time risk monitoring at the volume and velocity of modern payments, including instant payment rails. Given rising cyber threats and the regulatory pressure around financial crime, AI in fraud is a necessity.

Second, there is growing interest in what I would call “agentic commerce” – agents that can participate in commerce on behalf of consumers and businesses. Some industry forecasts suggest that a meaningful share of e-commerce, as much as 15% to 25% in the US alone, could be influenced or initiated by AI agents by 2030, which translates into trillions of dollars of volume flowing through agentic channels. AI is dramatically changing how people buy – most consumers are using agents to discover, explore, and evaluate purchasing decisions. Visa and Mastercard are already building the identity and credentialing infrastructure to make that possible. I saw this coming while at Visa; we were laying early groundwork for agent-based commerce long before ‘agentic AI’ became an industry buzzword.

Third, there is a priority around modernizing data and platform infrastructure, so AI is not blocked by siloed systems. The Stanford AI Index 2026 notes that while 88% of enterprises report using AI, fewer than 10% have scaled it, largely because their data and workflow architectures are not designed for AI-first operations. Payments players are starting to internalize that lesson: they realize they need to treat data as infrastructure and rethink workflows if they want AI to go beyond pilots.

Question. Fraud remains one of the biggest challenges in payments. How do you see AI changing the way organizations prevent, detect, and respond to fraud?

Arvind Ronta: What we are seeing is a shift from reactive, rules-based fraud systems to more adaptive, preemptive AI systems that continuously learn. This is not entirely new. At Visa, for example, we were working with behavioral and biological indicators for fraud detection as far back as 2014 – but with Gen AI, the speed, scale, and usability of AI have changed dramatically with recent advances. Modern AI can synthesize behavior data, device ownership data, and transaction patterns into a single risk score that outperforms humans on hit rate and response time.

The other big change is how AI is reshaping dispute and case management. Instead of having large teams manually handle every step of dispute resolution, you can design an AI-driven framework in which a customer files a dispute via a chatbot, the AI performs first-level triage, and then routes the case to the next step. This could be arbitration or merchant evidence collection. This compresses resolution timelines and reduces friction for both customers and operations teams, which is exactly where many institutions are now experimenting. So the fact that Opus is building a case management tool aligns perfectly with this methodology for transforming disputes and service.

Of course, all of this assumes that the underlying platforms have been modernized. If you are trying to plug adaptive, continuously learning AI into a legacy fraud stack, you will quickly hit architectural and data constraints. That is where specialized partners, like Opus, that understand both the payments domain and the AI integration challenges can help organizations make this leap from isolated models to full-stack, AI-enabled fraud capabilities.

Arvind Ronta is a veteran payments executive and Advisory Board member at Opus Technologies. With over two decades of experience at Visa, Discover Financial Services, and Deloitte, he has led global product strategy in embedded finance, tokenization, and next‑gen credit and installment solutions that power leading fintech and Big Tech payment experiences. He has helped build the FI-led models behind Apple Pay with installments, Apple’s BNPL service, and Amazon (moving billions in payment flows). At Opus, Arvind works closely with the leadership team to align AI, real-time payments, and risk decisioning with the company’s core strengths in engineering execution. His role helps banks, networks, and fintechs turn emerging technologies into scalable, compliant products that move the needle on revenue and customer experience.