In the last twelve months, the conversation around Al in compliance shifted from "should iwe adopt this" to "how do we prove to a regulator that we did it responsibly." A year ago, most teams were still working through whether Al belonged in their workflows at all. Today, the leading question is accountability: what happens when volumes scale, who owns the decision when an agent flags a case, and what does a defensible audit trail actually look like.
That's what we tackle in this issue. We look at how regulators from Singapore to the US are shaping what responsible Al in compliance has to mean in practice, and at what the future of the compliance desk really looks like when the volume problem fully arrives.
As our CEO Chaitanya Sarda puts it:
He expects the number of businesses on the planet to quadruple, or quintuple, in the next five years.
For compliance teams, that changes the job entirely as KYB becomes a throughput problem. The manual workflow most teams are running simply wasn't built for a world where a TikTok creator or a freelance contractor is as likely to be onboarding as a traditional business. The strain shows up in a familiar place: more signals than anyone can act on (registries, documents, sanctions lists, website content, adverse media), and no clean path from data to decision.
Al agents are starting to close that gap. Specialist systems each handle one job end to end and hand back something structured that an analyst can review and act on. In practice, the teams doing this well start with agents that handle the groundwork, like gathering signals, summarizing findings, flagging anomalies.
Importantly, these teams ensure that humans keep ownership of every final decision, making accountability an inherent trait from the start. That's particularly important because a workflow that cannot be explained and defended is a liability in compliance, regardless of how efficient it is. The goal is automation with a clear audit trail, not automation instead of one.
Regulators are arriving at the same conclusion. More on that below.
Read more about The Future of the Compliance Desk here
How MAS Is Putting Al at the Heart of Financial Crime
Singapore has long been one of the more sophisticated financial compliance environments in the world. But what makes it worth paying attention to today is not just the quality of its framework, but the direction it's moving in.
In November 2025, the Monetary Authority of Singapore published proposed Guidelines on Al Risk Management for all financial institutions, setting out formal supervisory expectations on governance, life cycle controls, and responsible Al deployment across the sector. Then, earlier this month, MAS went a step further. It announced an active collaboration with five banks, the Government Technology Agency of Singapore, and the Singapore Police Force to build Al and machine leaming models for pre-emptive financial crime detection.
The message from MAS echoes what FinCEN has been signaling in the US. Earlier this year, FinCEN proposed formal guidance recognizing Al as a legitimate tool in KYB and AML compliance, provided institutions can demonstrate governance, explainability, and human accountability in how it is used. For compliance leaders operating across both markets, that convergence matters. The expectation is no longer that you will eventually adopt Al. It is that when you do, you can demonstrate it was done responsibly.
Read more about what FinCEN's proposed rule means for KYB and AML here
On that note, this month AiPrise expanded its Global Database KYB coverage to 10 new countries, including Ghana, Cameroon, the Cayman Islands, and the British Virgin Islands, among others across Africa and the Caribbean. As the regulatory expectation to cover more markets grows, so does the infrastructure behind it.
Three Questions with Martin Lauria
The challenge today is less about whether data exists and more about whether teams can operationalize it fast enough and confidently enough. Compliance leaders are dealing with a new generation of businesses that don't always fit traditional onboarding models: creators, online sellers, remote contractors, global SMBs, all operating ecross fregmented digital footorints.
What we hear consistently is that the problem is no longer a single source of truth. Teams are trying to reconcile what a business says it does, what a registry says it is and what its actual online presence suggests. Those mismatches create friction, manuai reviews and repeated AFis that slow onboarding down significanths
The most forward looking teams are using Al agents to help triangulate those signals upfront, pulling together registry data, website intelligence, documents, sanctions screening and behaviorel indicetors into something analysts can review and act on quickly: That doesn't remove human judgment. it gives teams a cleaner and more structured starting point for making decisions.
Most compliance Jeaders are already bought into the idea that Al wit become part of the workflow The hesitation is less about whether it wit happen and more about where the boundaries are. The biggest question usually comes down to accountability. Teams went to understand where human oversight remains necessary, how decisions can be explained to regufators and whether the system can produce a defensible audit trail when something goes wrong.
What's interesting is that the conversation has evolved quickly. A year ago, many teams were asking whether Al belonged in compliance at all. Today, the discussion is much more operational; where can agents reliably handie repetitive work and where should humans stay directly involved?
My view is that human oversight will remain fundamental, particularly around euge cases, escalation paths and risk decisions. But as institutions become more sophisticated in how they define rules, risk appetite and customer segmentation, agents wi increasingly handle the groundwork that consumes compliance teams today.
It has absolutely shifted the conversation. Regulatory endorsement matters enormously in compliance because it changes Ai from something experimental into something institutions can seriously put into play. When regulators like MAS and FinCEN acknowledge that Al can play a legitimate role in compliance workflows, provided governance and accountability are in place, it gives tears confidence to move from exploration into implementation
At the same time, there's also a practical reality driving urgency, Compliance tears are already struggling with increasingly manual and fragmented workflows while onboarding volumes and complexity continue to rise. The pressure is coming from both directions: regulators are becoming more open to responsible Al adoption, while operational demands are making traditional processes harder to sustain.
That combination is what's accelerating adoption. Most teams are no longer asking whether Al will become part of compliance operations. They're trying to determine how to implement it responsibly before the volume problem becomes unmanageable.
Compliance Compass is published by AiPrise, an Al-powered global compliance platform. We write this newsletter because we needed an outlet for everything we're learning and thinking about.