The 95% problem: why banking's biggest AML cost has nothing to do with catching criminals

The single most expensive number in compliance today is the false-positive rate - up to 95% on traditional, rules-based AML systems.

That is the share of alerts generated by these systems that turn out to be false positives. Not suspicious. Not borderline. Just noise, flagged because a static threshold from years ago happened to trip on a perfectly lawful transaction.

The economic weight of that noise is hard to overstate. Global AML compliance now costs the industry over $206 billion annually, and the majority of that spend goes not to catching criminals but to investigating transactions that were never criminal to begin with. While US regulators officially estimate it takes about two hours to file a SAR, independent studies put the true end-to-end burden at up to 22 hours per alert once investigation, documentation, and review cycles are included.

For decision-makers, this is no longer a compliance line item. It is a structural drag on the bank.

What AI actually changes

The interesting shift in 2025 to 2026 is not that AI exists in AML. It is that AI is finally being measured against the false-positive line, and the numbers are starting to land:

• AI-powered AML systems are demonstrating 70 to 95% reductions in false positives in production deployments.
• AI-assisted due diligence has cut KYC review times by 50%.
• ING has reported a 90% reduction in onboarding time and a 30% cut in compliance staff workload using an agentic AI stack (per McKinsey).

These are not incremental gains. They are the difference between a compliance function that scales with growth and one that taxes it.

Most AML budgets were not built to fight financial crime. They were built to absorb the inefficiency of the systems hunting for it.

When 95 out of every 100 alerts are wrong, the bank is not really running an AML program, it is running a very expensive triage operation. AI does not make compliance optional; it changes what compliance actually does. Investigators stop chasing noise and start working the cases that matter. Onboarding accelerates. Customer friction drops. And, quietly but importantly, true positives start surfacing faster, because they are no longer hiding inside a haystack of false ones.

Why this matters for banking leadership

The conversation at the executive level needs to move from “how much should we spend on AML?” to “how much of our AML spend is actually buying us risk reduction?”

For most banks today, the honest answer is: not enough.

Auditale’s AML capabilities are built around this exact gap, applying AI where rules-based systems fail, so compliance teams stop drowning in alerts and start doing the work they were hired to do. The age of AI in banking will not be defined by chatbots or marketing personalisation. It will be defined by which institutions used this moment to rebuild the unglamorous, expensive, mission-critical functions, AML first among them, into something genuinely intelligent.

The banks that move now will not just save money. They will redefine what “good” compliance looks like for the next decade.

Auditale’s AML and AML QA solutions help banks turn compliance from a cost centre into a competitive advantage. Talk to us.