From Chaos to Clarity: Getting Data Ready for AI and Compliance
- FinScan
- May 30
- 3 min read
Updated: Jun 3
At the recent SIFMA AML Conference in Washington, DC, a breakout session explored the evolving relationship between artificial intelligence, compliance, and the data strategies necessary to support both.
Titled Untapped Potential: Building Innovation-Ready Data and sponsored by FinScan, the panel brought together three industry veterans—moderator Steve Marshall (FinScan) with Adrian Murray (Fisent Technologies) and David McLaughlin (founder of QuantaVerse)—for a candid discussion on what it takes to get data ready for AI, automation, and regulatory transformation. Here are five key takeaways.
1. AI Is advancing fast—but only as fast as your data allows
The panel opened with a bold analogy: deploying AI strategies on unstructured, unreliable data is like expecting a pro athlete to perform on a junk food diet. As Marshall noted, “We’ve limped along relying on spreadsheets and half the story. It’s been enough—until now.”
AI in financial crime prevention has progressed from simple detection use cases to autonomous decision-making in reporting and alert triage. But the panel agreed: innovation-ready data is a prerequisite. Without it, AI accelerates bad decisions.
2. False positives are a data problem, not just a tech problem
One of the most compelling discussions centered on AI’s ability to reduce false positives—a constant pain point in AML compliance. McLaughlin shared cautionary anecdotes from clients who purchased AI-driven platforms but failed to onboard usable data, sometimes years after contract signing. As he put it, “If you have good data, it keeps getting better. If you have bad data, it keeps getting worse.”
Better data quality and integration aren’t just nice to have; they make AI and AML more effective at analyzing unstructured data, minimizing noise, and improving decision-making.
3. Don’t just replicate the old process—rethink it entirely
“AI doesn’t have the same constraints as human workflows ...”
Murray introduced the DO / NEED / WANT framework, which sparked strong agreement from the panel. Too often, organizations attempt to automate what they already do rather than rethinking what’s truly needed to achieve the outcome they want. While Fisent initially focused on AI for compliance, its customers have discovered many other use cases for how AI can enable automation with the company’s Applied GenAI Process Automation solution, BizAI.
“AI doesn’t have the same constraints as human workflows,” Murray explained. “It can process in parallel, not sequentially. So, if you force it into human-style workflows, you’re modeling the limitation, not the outcome.”
4. Reframing the data problem: it’s not just about input channels
Murray also challenged conventional thinking on how to fix data problems. Historically, organizations have tried to control data quality at the front end by pushing customers and partners into structured digital channels. But that often fails due to customer friction, regulatory constraints, or outdated infrastructure.
With modern AI, Murray argued, the paradigm has shifted: “What if you don’t have to change how people engage with you? What if you can still extract unstructured, usable data from any modality—email, PDF, fax—without burdening the customer?” This shift from front-end transformation to back-end intelligence opens up new possibilities for innovation without disruption.
5. Data alone isn’t enough—rethinking both sides of the equation
“Effective AI comes from designing systems that reflect your goals ..."
Marshall emphasized that good data is necessary but not sufficient. AI success also requires a mindset shift. “We need to rethink not just what data is, but what AI is,” he said. “Effective AI comes from designing systems that reflect your goals, not just automating what you already do.”
The panel urged attendees to question their assumptions and reframe both the technology and the problem: What are we trying to solve? What does success look like? And what data, new or existing, can get us there?
Build for tomorrow, not just today
The conversation wrapped with a call to action: don’t wait until you’ve solved every data issue to pursue innovation. Instead, rethink how data is defined, collected, and used. Build agile systems that can ingest imperfect inputs and still deliver results.
In an era where AI is poised to transform compliance, the firms that will win are not the ones with the most data but the ones who best understand it.
As the room of AML leaders walked away from this session, one message rang clear: If data is the fuel, then it’s time to clean the tank before stepping on the gas.
