Visa announced six new AI tools in April to automate credit card dispute management, processing over 106 million disputes globally annually. This matters to UK financial services firms, insurers, and accountancy practices because disputes are a hidden cost driver — and because Visa's move signals that payment processors are taking control of the client-facing automation that banks and fintechs thought was theirs to own. For regulated firms, this is a strategic inflection point: either you embed dispute intelligence into your client workflows, or you become a data supplier to platforms that do.

This story is part of a larger pattern. Over the past 18 months, we have seen infrastructure vendors (Stripe, Wise, now Visa) deploy generative AI into operational workflows that were previously manual or fragmented. The pattern is consistent: identify a high-volume, rules-based process; train a model on historical data; reduce friction; capture the data. The FCA's Consumer Duty (PS22/9) already requires firms to demonstrate they are acting in client interest and managing harm. When Visa automates dispute responses before they reach your desk, your firm loses visibility — and control — over outcomes. That is a governance problem dressed up as efficiency.

Trovix's view is straightforward: AI in dispute management must be auditable, not just fast. Visa's tools will probably reduce processing time and improve merchant satisfaction. What they will not do is eliminate the need for human judgment in contentious cases, and they risk embedding merchant bias into the decision logic if training data is not properly validated. Compare this to how Harvey or Luminance approach document review — they surface reasoning, flag conflicts, stay human-in-the-loop. Visa's approach appears more opaque. A UK regulated firm deploying AI for dispute handling must implement governance first. That means Trovix Audit, which gives you real-time visibility into how AI is making decisions across your operations, ensuring you can evidence compliance with SRA Code, FCA rules, and PRA SS1/23 expectations on AI risk management.

What should a mid-market financial services firm or insurer do now? First, audit your current dispute workflow and identify where you are vulnerable to competitive displacement. Second, if you use Visa (or any payment processor with embedded AI), map how their decisions flow into your compliance obligations — your compliance team needs to know this is happening. Third, do not treat dispute AI as a cost play. Build it as a governance asset. Implement automated triage, yes, but ensure every merchant dispute that carries reputational or regulatory risk surfaces for human review with full reasoning trails. Your FCA engagement letter probably already asks for evidence of this. Fourth, evaluate whether your current AI tools give you this auditability. Most do not.

Source: CNBC