WealthAi has now convened its Compliance Launch Group three times: a small circle of compliance and commercial leaders from across wealth management, talking openly about where AI regulation is heading and what it means for their teams day to day. Session 3 ran on 10th September. Three themes stood out. Here’s what came up, and where to focus first:
1. Clients are using AI to sound more sophisticated than they are
Attendees described a growing pattern: clients arrive at conversations having already run their questions through a chatbot, and that confidence doesn’t always match their actual risk appetite or behaviour. Someone can talk fluently about a crypto allocation while holding almost everything in cash. One observation from the room captured the shift well: the definition of knowledge is moving from knowing things to knowing how to find things.
The definition of knowledge is moving from knowing things to knowing how to find things.
Where to start: treat this as a data problem, not a client education problem. Fluency in conversation is easy to fake and expensive to verify manually, so client categorisation needs to lean harder on what a client actually does: trade history, portfolio composition, past decisions, rather than what they say in a meeting. If suitability assessments still rely mainly on self-reported knowledge, that’s the first thing to fix.
2. What compliance leaders want most isn’t more AI. It’s oversight of the AI already in use
Asked directly what their biggest need was, the room didn’t point to a new feature. They pointed to governance: visibility into what staff already put into AI tools, and confidence that sensitive client data isn’t leaking out through everyday use. As one attendee put it, junior staff are already building their workflows around AI, and nobody wants to be the one who shuts that down, but that enthusiasm needs guardrails.
Where to start: build the trail before you need it. Every AI-assisted decision in a firm should produce a reconstructable record: what data was queried, what was asked, what came back, and who signed off on it. That’s not a nice-to-have for when a regulator eventually asks. It’s what lets staff use AI freely without losing control of where client data goes.
Build the trail before you need it. Every AI-assisted decision should produce a reconstructable record.
3. An audit trail isn’t enough. Wealth managers want to open it, not just be told it exists
Attendees weren’t satisfied by being shown that a decision could be traced; they wanted to see inside it. One asked for the guardrails and decision rules behind each step, not just the workflow diagram. Another pushed back that a label like “checking for risk” is too vague, since a regulator will want to know precisely which criteria were applied. A third asked how easily a specific decision from eighteen months ago, and the client statement behind it, could actually be pulled up and interrogated.
Where to start: treat explainability as a drill-down problem, not a display problem. A diagram showing that a trace exists is not the same as a UI that lets someone open any node and see the rule, the data source, and the reasoning behind it. If the trail can’t be interrogated at that level of granularity, on demand, months later, it won’t hold up under regulatory scrutiny even if it’s technically complete.
The thread running through all three
The manual, self-reported, calendar-driven version of compliance is running out of runway. Firms already building continuous, evidence-based processes won’t need to scramble when the rules catch up.
None of this is settled, and that’s exactly why sessions like this one matter. The Mills Review, the demand for genuinely interrogable audit trails, and the pressure on firms to prove AI oversight are all moving faster than most compliance functions can absorb alone. The value of putting compliance leaders in a room together isn’t a set of answers. It’s the chance to compare notes early, pressure-test an approach before it’s locked in, and walk away with a shorter list of things to fix first. That’s the conversation WealthAi intends to keep hosting, and the next one will pick up exactly where the open questions from this session leave off.