RegTech Summit New York returns on 19 November for its tenth year, bringing regulators and practitioners together to examine compliance, data and artificial intelligence (AI). Across the agenda runs a question with consequences for compliance innovation: can firms demonstrate that their controls work and explain the decisions they produce?
When a regulator challenges a reporting exception or a decision to close a surveillance alert, firms need to reconstruct the evidence. That means establishing which information was available, how systems processed it and why reviewers reached their conclusions.Recent supervisory findings show where that chain-of-evidence is deficient. The UK Financial Conduct Authority’s Market Watch 86 found that brokers using automated surveillance could not always explain how their systems worked. Some settings excluded significant proportions of trading activity, while weaknesses in senior oversight limited firms’ ability to identify and address problems. For US firms, the Securities and Exchange Commission’s fiscal 2027 examination priorities put supervision and emerging technologies among the areas facing scrutiny.
AI in Compliance Workflows
As firms give AI a greater role in selecting and reviewing compliance information, gaps in the evidence-chain could hinder significant risks reaching human investigators.
The opening panel – Future of regulatory compliance, AI and RegTech: Navigating the next phase of transformation will examine how firms moving AI from pilot projects into day-to-day compliance can judge whether the investment is paying off.
For Abigail Bitner, Product Marketing Strategy, Compliance, NICE Actimize, that judgement starts with the data the system captures. “The biggest risk in AI-driven compliance isn’t a model that gets it wrong, it’s the data it never saw. As firms move from GenAI sandboxes to production, regulators aren’t just asking how smart your AI is. They’re asking whether you can prove every communication and trade was captured, explain why every alert fired, and show that automation is freeing your experts to apply human judgement where the risk is greatest. That’s what regulatory certainty means, and that’s where AI earns its ROI,” she says.
A firm that approves AI to review trader communications in New York may need to revisit that approval before extending it to a desk overseas. The same model may draw on a shared archive, yet local privacy requirements and the firm’s own access policies can limit which conversations it may retrieve.Robert Cruz, VP, Information Governance at Smarsh, sees firms making these decisions while regulatory expectations continue to differ across jurisdictions. ”Firms are navigating a complex web of AI regulation at US state, Federal, and international levels. Some are principles based, while a few are prescriptive, potentially creating sources of policy conflict. But few are pausing AI initiatives in pursuit of clarity, because the promises of AI are too compelling, even in spite of risks that span regulation, privacy, infosec and shadow AI.”
Connecting The Evidence
Reconstructing a compliance decision can require information held across trading systems and unstructured data from communications archives voice, e-comms/email and case records. A reporting exception, for example, may involve the original transaction, subsequent amendments and communications explaining how the firm resolved the discrepancy.
Connecting those records helps reviewers establish where an error arose, whether the correction addressed its cause and an evidentiary log for subsequent audit or regulatory examination.
The next panel – Opportunities for harnessing structured and unstructured data for compliance insight – will share experiences with bringing unstructured data into compliance workflows. As Vall Herard, CEO and Co-Founder at Saifr notes: “As AI capabilities evolve, it’s clear that AI-based tools are increasingly becoming the “first reviewer” in risk, fraud, and compliance workflows,” he says.
Panelist Richard Ziegenbalg, Executive Advisor, Surveillance & Conduct Risk at Behavox offers these insights: “Compliance risk rarely sits in one system. The signal is usually split between what was traded and what was said, so firms that bring trades, communications and voice onto one data layer catch what siloed tools miss. The test is whether every insight can be traced back to its evidence, because an insight you can’t defend to a regulator is just another alert.”
The accountability for any compliance decisions cannot be outsourced to a model. Before extending controls to AI, compliance leaders must decide which records the system may access, who checks for omissions and who signs off on any recommendation.
Stephanie Rubizhevsky, Chief Solutions Officer at Novatus, describes what this means for reporting oversight: “I’ll be speaking at A-Team Group’s RegTech Insight’s summit in New York on how firms combine structured and unstructured data for compliance insight. At Novatus, our focus is transaction reporting oversight, where the test is simple: can you show a regulator that every report was accurate, complete, and timely, and explain any exception? This is dependent on connecting reporting data with the wider evidence around it, including data lineage and quality controls you can stand behind. I’m looking forward to comparing notes with the other panellists on where this is working in practice, and on how firms should treat the growing volume of AI-generated content.”
Hidden Remediation Overhead
Data errors resulting from incorrectly interpreted rules can be propagated across several reports resulting in pre-submission remediation effort. Reviewers need to establish when the rule took effect and which trades it covered. The effort then shifts to identifying the reports that need correcting. Until the rule is fixed, the remediation cost continues to compound.
Andrew Pinnington, Chief Customer Officer at Novatus, will examine that recurring cost in his keynote – Breaking the remediation & back reporting cycle. He puts the industry’s annual bill in the hundreds of millions of dollars. “It’s no exaggeration to say that firms are still spending hundreds of millions of dollars each year on fixing and ‘back reporting’ issues in their regulatory reporting. Firms can find huge savings and drastically mitigate regulatory risk by improving the ways they can understand, identify and resolve reporting issues. With the pace of regulatory change slowing slightly, there is no better time to do it,” he says.
Be a Part of the Conversation
What evidence would you want an investigator to see before signing off on an alert that AI has helped review? The summit’s closing roundtables give delegates an hour to compare answers with peers. The surveillance discussion will explore where investigators need to exercise judgement, while the AI governance table will consider who sets those limits and oversees their use. Two further tables will examine how firms adapt reporting systems across jurisdictions and decide which changes in a customer’s circumstances should trigger a fresh review.
Join us at @Ease, 7th Floor, 605 Third Avenue, New York on 19 November and bring a question from your own compliance programme to the table.
Check out the agenda HERE, and register to be a part of the conversation HERE.
Subscribe to our newsletter
