Visa's announcement of six AI tools to automate charge dispute handling matters to UK regulated financial services firms because it exposes a gap that has been widening for years. In 2025, Visa processed 106 million disputes globally — a 35% increase since 2019. Behind that number sits a mountain of manual work: back-office staff manually reviewing files, chasing evidence, applying rules inconsistently, and missing FCA Consumer Duty deadlines because the system cannot scale. Visa's tools promise to speed up triage, predict dispute outcomes, and flag high-risk cases automatically. For mid-market insurers, payment processors, and financial services providers handling their own disputes, this looks like the answer. It is not yet.

What this story reveals is that the industry has finally admitted what should have been obvious five years ago: the dispute resolution infrastructure across UK and European financial services is fundamentally broken by design. It was built for volumes that no longer exist and compliance regimes it does not understand. The PRA's operational resilience expectations (PRA SS1/23) demand that firms understand their third-party dependencies and failure modes. Visa's automated dispute tools will reduce Visa's own costs and risk. But they will not reduce yours if you plug them in without understanding how they work, what they assume about your data, or what happens when they fail. The EU AI Act is coming. The ICO is sharpening its UK GDPR enforcement. The FCA's Consumer Duty now requires you to demonstrate that your processes deliver fair outcomes. A vendor tool that uses a black box to route disputes faster does not meet that requirement.

Trovix's view is that Visa's tooling is a symptom, not a solution. Large payment processors can afford to build AI products because they have the scale, the data quality, and the regulatory cover to deploy them. Mid-market regulated firms cannot. You need governance-first AI — not vendor-first. That means: (1) mapping what your dispute process actually does and why (2) understanding where automation will genuinely reduce cost and error, not just speed things up (3) building audit trails that prove to the FCA that your AI is making decisions that are fair and proportionate, not just accurate. Products like Harvey or Legora are starting to solve parts of this for legal document review. But dispute processing is not document review. It is a workflow that must be explainable, auditable, and compliant at every step. That is why Trovix Audit was built — to give regulated firms the governance layer before they deploy AI, not after. You need to know what you are automating before you automate it.

Here is what to do Monday morning: (1) Do not buy Visa's tools yet. Wait. (2) Run a process audit of how your firm currently handles disputes. Which steps are manual? Which are rule-based? Which involve judgment? Which have failed or caused complaints? (3) Identify the two or three highest-cost, highest-error steps. (4) Then — and only then — evaluate whether automation solves them, and what governance you need to deploy it safely. If your firm handles disputes for insurance, payments, or lending, this is not optional under the FCA Consumer Duty PS22/9. You must be able to explain your dispute decisions. AI that you cannot explain is a liability, not a shortcut.

Source: CNBC