Most AI pilots in financial crime compliance are not failing on technology. According to Vivox AI, they falter in the space between an impressive demo and a compliance officer prepared to sign off on what the system produces.
That tension dominated a private dinner hosted by Vivox AI at The Shelbourne in Dublin on 5 October. The event was co-hosted by Prof. Axel Weber, former president of the Deutsche Bundesbank and former chairman of UBS Group, alongside Vivox AI founder and CEO Tim Khamzin. Senior banking and payments executives, former regulators and AI specialists gathered under the Chatham House Rule. Vivox AI has distilled the discussion into six uncomfortable truths.
The first challenges a widely held assumption: manual processes are not the gold standard. Parallel testing is often presented as measuring AI against human analysts, but Vivox AI argues it really tests both. Where outcomes differ, the AI may have missed something, or it may have uncovered issues analysts overlooked for years. Consistency matters as much as accuracy, since a system reaching two different conclusions on the same case is a problem even if both are defensible.
Second, AI generates work before it reduces it. Teams spend months running dual processes and investigating discrepancies, and each new release resets part of that effort. Vivox AI warns that business cases promising instant headcount savings are set up to disappoint. The more credible pitch is greater capacity and stronger detection first, with cost savings later, if at all.
Third, incentives are misaligned. Commercial teams book revenue from new customers while compliance absorbs the cost of onboarding and monitoring them. Until that cost is visible to the business generating it, there is little pressure to improve upstream processes, and AI cannot fix that.
Fourth, budget cycles move more slowly than the technology. Capabilities that did not exist when annual budgets were agreed can be production-ready within six months. Firms that only assess new tools once a year will always lag, so the question becomes how to create room for controlled testing between budget rounds.
Fifth, there is the question of who will supervise the AI in five years. If machines take over first-pass review, junior analysts may never build the judgement needed to challenge them. Vivox AI stresses that preserving that expertise must be deliberate, and that accountability cannot be outsourced to a model or a vendor.
Finally, more data does not guarantee a clearer picture. Aggregating registries and subscription databases saves time but does not confirm that information is current, relevant or sufficient, and quality varies widely by jurisdiction. Consolidating onto a single provider also concentrates dependence.
The conclusion from Vivox AI is clear: standing still is not neutral. Instant payments are shrinking the window to stop fraud and supervisory expectations continue to rise. The firms that move beyond pilots will be those able to prove, case by case, that their controls work, supported by clear processes, auditability, early second-line involvement, strong data governance and preserved human judgement.
For more, read the full story here
Copyright © 2026 FinTech Global
Copyright © 2026 RegTech Analyst
