Visa's launch of six AI tools to automate charge dispute handling sounds like progress. In reality, it is an admission. The company processed 106 million disputes globally in 2025—a 35% surge since 2019—and most of them still moved through back-office systems run by humans reading documents and typing decisions. For UK payment processors, insurers, and financial services firms, this is both a warning and an opportunity. The FCA's Consumer Duty (PS22/9) now requires firms to demonstrate that their processes actually protect customer outcomes, not just comply with rules. Manual dispute handling fails that test. It is slow, inconsistent, and creates audit risk when the FCA asks how you handle edge cases.

What Visa's announcement really reveals is that the industry has been slow to automate where it matters. Dispute workflows are repeatable, rule-based, and fact-heavy—exactly the kind of work AI should do years ago. But financial services firms have been distracted by vendor promises about 'intelligent' systems that sound smart but don't actually reduce cost or time. They have bought tools like Harvey (trained on legal data but not payment processing rules) and Microsoft Copilot (good for chat, useless for systematic workflow) and discovered they still need humans to do the actual work. Meanwhile, volumes keep climbing. The PRA's SS1/23 framework is tightening operational risk expectations. The EU AI Act now treats high-risk AI systems—which include those handling financial disputes—as a compliance obligation, not an option. UK firms that have not built AI governance by now are falling behind.

Trovix's view is blunt: most dispute AI fails because it tries to be clever instead of being complete. Visa's approach—automating the intake, classification, evidence gathering, and decision logic for disputes—is the right direction. But if those tools sit on top of your existing siloed systems, you have just added another layer of technical debt. What UK firms actually need is integrated automation that connects dispute intake through to settlement, with governance built in from the start. That means Trovix Watch monitoring the specific rules Visa and your other processors update, and Trovix Audit tracking every decision the AI makes so you can explain it to the FCA. Tools like Luminance, which focus on document analysis, are useful for reading dispute evidence. But they do not orchestrate the entire process. Trovix's difference is that we build systems that work as end-to-end workflows, not as point solutions.

If you are a mid-market law firm, insurer, or financial services business processing more than 5,000 disputes annually, you have 90 days to audit your current process. Map how many disputes require manual review, where the delays happen, and which ones generate complaints or regulatory queries. Then benchmark that against what Visa's new tools can do—and against what your system governance actually covers. You will likely find that your AI governance dashboard (if you have one) does not track dispute decisions, your intake system has no automation, and your team is still arguing about whether a transaction is a friendly fraud or a chargeback. Start there. Not with buying new vendors, but with understanding what you currently lose to manual handling.

Source: CNBC