Visa announced in April 2026 six AI tools to automate credit card dispute resolution across merchants, issuers and acquirers. This matters to UK financial services firms because disputes are now a compliance flashpoint, not just an operational nuisance. The FCA's Consumer Duty PS22/9 makes clear that firms must act in customers' best interests, and dispute handling—how fairly claims are assessed, how quickly they are resolved, how transparently outcomes are explained—is now scrutinized. Visa's tools promise to catch disputes before escalation and generate AI-powered responses. The scale is real: 106 million global disputes in 2025 alone. But Visa's approach reveals a critical flaw in how most payment networks and large tech firms think about regulated AI.

This story is part of a much larger pattern. Major infrastructure players—Visa, Mastercard, fintechs, cloud providers—are deploying generative AI at scale to solve operational headaches. They move fast, they iterate in production, they measure success by volume and speed. But they are not regulated like the firms they serve. A payment processor does not face an FCA investigation if an AI dispute response fails to document decision-making clearly, or if algorithmic bias disadvantages certain customer cohorts. The firms that use Visa's tools do. The gap between how unregulated tech companies deploy AI and how regulated service providers must govern it is now the biggest risk in financial services. It is the same pattern we saw with outsourcing, with third-party data breach risk, and with model risk frameworks in banking. Infrastructure providers move first; compliance follows; regulated firms get caught in between.

Here is Trovix's honest view: generic dispute automation without built-in governance is regulatory liability dressed as efficiency. Visa's tools may be excellent at pattern matching and response drafting. But they do not automatically create audit trails, decision rationale, bias testing or FCA-grade governance—the things that actually matter if a customer complains or if your firm is examined. Products like Harvey and Legora focus on legal reasoning and regulatory knowledge; Luminance focuses on document intelligence. None of these are dispute-specialist tools either, and none solve the core problem: regulated firms need AI that is *designed from the ground up* for compliance, not bolted on afterwards. Trovix Audit exists specifically because firms told us they deploy powerful AI tools and then spend months trying to document what those tools actually did, for whom, and why. That is a failure of implementation design, not just governance theatre.

If you are a mid-market financial services firm, insurer, law firm or accountancy practice, do not assume that adopting Visa's tools (or any vendor's AI) satisfies your compliance obligations. Ask three hard questions now: First, does the AI tool generate decision-level audit logs that explain *why* a specific outcome was reached for a specific customer? Second, does the vendor conduct ongoing bias testing and provide you with those results? Third, if a customer disputes the AI's decision, can you explain the decision to the FCA in non-technical language? If the answer to any of these is 'no' or 'that's on you to figure out', then you are outsourcing operational risk while keeping regulatory accountability. That is not a trade worth making. Trovix Watch tracks payment and financial services regulatory changes in real time; use it to stay ahead of where the FCA and PRA will go next on AI governance.

Source: CNBC