This is part 3 of a three-part series from Future Proof. Read part 1: Before AI Can Scale, Wealth Management Has a Data Problem to Solve and part 2: From Single Player to Multiplayer.
Wealth management is a trust business. That idea came up again and again at Future Proof, and it frames the two challenges that may matter most as AI moves deeper into firms: proving that AI is governed, which is the heart of AI governance in wealth management, and making sure people actually use it.
Keeping humans in the loop
Panelists were enthusiastic about AI’s potential, but clear about where the risk lies. The scenario that worried them most was an agentic tool producing a recommendation that goes straight to a client with no expert review.
Clients trust their advisors, and that trust doesn’t transfer automatically to software. Several speakers argued that human connection is AI’s most important counterweight. The more time AI frees up, the more time advisors can spend with clients.
AI governance in wealth management takes more than a policy
Jordan Campbell of Impruve raised a question that landed with the room. Most firms now have an AI policy. How well can they prove they enforce it?
He and Kelly Wagman of Cresset described the emerging idea of a control plane. That is not one product but a set of controls wrapped around a firm’s AI systems. It answers questions like these:
-
Access. Who can use which models, what can they do with them, and can access be cut off quickly?
-
Data permissions. When an agent acts on someone’s behalf, is it limited to what that person could see if they logged into the system directly?
-
Audit trails. If a regulator asks why an agent updated a record in the CRM, can the firm show that a human approved it first?
The good news is that much of this builds on tools firms already use. Identity platforms like Okta and Microsoft Entra now integrate with enterprise AI tools. Model providers offer compliance and analytics APIs. Firms that plan to use more than one model may add an AI gateway to govern them consistently.
Every new vendor is a fiduciary decision
Cecilia Williams, COO of Halbert Hargrove, noted that growth creates a steady stream of vendor decisions. Each one raises a question of what that vendor will do with client data once the firm hands it over. As fiduciaries, firms must understand the answer before they sign. The challenge is doing that diligence without slowing adoption to a crawl.
Adoption is the real test
Both COOs on the operations panel called adoption their biggest challenge. Neither considered the technology itself the hard part. What mattered was getting it used consistently.
Halbert Hargrove’s approach has held up from note taking tools through today’s AI agents:
-
Pilot new technology with a small team that understands it.
-
Build internal champions from that group.
-
Release it to the wider firm only once everyone is comfortable with it.
-
Track usage after launch, and follow up with anyone not using it to learn why.
For AI agents specifically, the firm is giving access first to a few people who truly understand how agents work. The aim is to avoid a sprawl of agents that nobody can monitor.
Caitlin Douglas of Elevation Point added a nuance for platform firms. Her affiliated firms don’t all have to use the same tools. But whichever tool a firm chooses, it should be fully adopted, not sitting unused after one person tries it once or twice.
The AI strategy session added a final point. AI programs work best when leadership across the organization is aligned and one clear owner drives the effort. When ownership is spread thin, programs stall. For more on why pilots stall and what the ones that scale do differently, read Why 95% of AI Pilots Never Reach Production.
The client stays at the center
Looking five years ahead, panelists predicted more consolidation, larger firms, and big changes in wealthtech. The hope they kept returning to was that none of it pulls the industry away from the client. One panelist even predicted that the best planning advisors will be able to raise their fees over the next decade, much as concierge doctors have, because clients will increasingly value human expertise alongside automated results.
The common thread was that AI’s job is to give advisors more capacity for the work only they can do.
Building trust into the architecture
At Cheesecake Labs, we believe governance should be designed into AI systems from the start, not added after something goes wrong. We build permissions, audit trails, and human approval steps into the workflows themselves, and we plan rollouts around adoption, beginning with small pilots and measuring real usage.
For firms deciding where to begin, our AI Readiness Assessment offers a structured starting point. It looks at four things:
-
the state of your data foundation
-
one high value workflow
-
gaps in your governance controls
-
a pilot plan built for adoption
If the conversations at Future Proof sounded familiar, we would welcome the chance to continue them.
