Artificial intelligence is getting better at identifying problems across wealth management, but according to Surge Ventures CEO Sid Yenamandra, that may not be the hardest part of putting AI to work.
Some of the most valuable AI applications today are happening behind the scenes – from compliance to cybersecurity to data management – where the tech can sift through vast amounts of information and surface what deserves the attention of a human.
As firms across the industry move toward more proactive AI monitoring, Yenamandra argues that the industry could face an unexpected consequence, as detecting more potential problems could create greater supervisory obligations. In this edition of Digital Wealth News AI Reality Check, he explains why firms should focus less on how many signals their tech can surface and more on whether they can defend the decisions made in response to those signals.
DWN: What types of AI applications are delivering the most value to the wealth management industry today?
Sid Yenamandra: Anywhere a human reviews the output before it matters. That’s the whole pattern. In compliance, it’s e-comms review — models triage the volume, a principal makes the call. In cyber, it’s posture monitoring and vendor due diligence, where AI reads a 300-question DDQ response or a firm’s endpoint telemetry and surfaces what’s actually anomalous. In data, it’s the least glamorous and most valuable: normalizing client and account records across custodians, CRMs, and portfolio systems that were never designed to agree with each other. None of it makes a press release. All of it is real. The common thread is that AI narrows the field and a person still decides — which is exactly why these use cases shipped while client-facing ones stalled.
DWN: What is the most overrated application of AI within the wealth management industry today?
SY: Detection. Firms are buying alert generation and calling it supervision. Every vendor in the category can flag a message or a misconfigured endpoint, and that capability is commoditizing toward zero — the models are cheap and the signal is table stakes. But FINRA Rule 3110 doesn’t ask whether you generated an alert. It asks what you did with it, who reviewed it, on what basis, and whether you can reconstruct that judgment years later. Same with Reg S-P and a vendor incident: nobody asks whether the scanner caught it. They ask what happened in the 72 hours after. The alert is the cheap part. The adjudication — the reasoning, the disposition, the record — is the obligation, and it’s badly underpriced right now.
DWN: What are the most critical pitfalls wealth management firms must avoid when adopting AI?
SY: The one nobody’s watching: firms are documenting that they deployed AI, not why a given judgment was made. When the exam comes, no one asks whether you had a tool. They ask you to reconstruct a specific decision from eighteen months ago — this flagged message, this vendor exception, this policy deviation — and most firms can’t. Everything else follows from that. Bolting AI onto a broken supervisory process produces broken decisions faster. Skipping the compliance review because a tool “feels low risk” creates exposure that surfaces at the worst possible moment. Buying before anyone maps the workflow gets you a shiny tool nobody opens six months in. And the data problem is the foundation under all of it — AI doesn’t fix inconsistent client records across systems; it moves the mess faster with more confidence attached.
DWN: What AI development will have the greatest impact on the wealth management industry in 12 to 24 months from now, and why?
SY: Proactive monitoring — but not for the reason it gets pitched. Everyone’s focused on the detection side: catching policy drift, vendor risk changes, or unapproved AI use before anyone thinks to look. That’s already shipping. The consequence nobody’s modeled is what it does to supervisory obligation. Every signal a firm surfaces and doesn’t act on becomes a documented instance of knowing and not acting. Detection at scale manufactures liability at scale. So, the firms that win the next two years won’t be the ones detecting more — they’ll be the ones who can defend their disposition of what got detected, at volume, with a rationale attached to each one. That’s the bottleneck forming right now.
