This is part 1 of a three-part series from Future Proof.
Walk the floor at Future Proof this year and one thing was impossible to miss: AI was everywhere. Nearly every booth had a demo, nearly every keynote had a prediction, and the energy around what these tools can do for advisors was real.
But in session after session, the people actually running firms kept returning to a less glamorous subject. Before AI in wealth management can deliver on its promise, the industry has to fix its data.
The integration layer nobody asked for
The co-founder of Envestnet, who is now building a new company called Field Wealth Intelligence, described the problem plainly. A typical advisory firm runs somewhere between 8 and 12 platforms. Each has its own data structures, its own naming conventions, and its own billing mechanics. The technology in each one is impressive. The problem is that none of them talk to each other.
In his view, the industry built excellent point solutions and then left advisory firms to become the integration layer. He argued they are the wrong people to solve that problem. The cost shows up in two places:
-
Headcount. Every new capability requires more people to administer it, which squeezes margins and caps growth.
-
Time. Advisors can spend six or seven hours preparing for a single client meeting because the information they need is scattered across systems.
A separate panel of wealth management executives put the number even higher. One panelist estimated that firms subscribe to 18 different software products on average, the result of decades of point solutions filling gaps that custodians, broker dealers, and banks left open.
Why AI in wealth management inherits the mess
Speakers kept coming back to one principle: AI is only as effective as the data it sits on.
Kelly Wagman, Head of AI Strategy at Cresset, and Jordan Campbell, Chief Data Officer at Impruve, made this concrete. Your CRM might define a household one way while your portfolio management or planning system defines it another. When an AI tool has to reconcile those differences on the fly every time someone asks it a question, three things happen:
-
It is slow.
-
It is error prone, because there is no guarantee the reconciliation is done the same way twice.
-
It is expensive, because the firm pays for that reconciliation in compute every single time.
Their recommendation was to do that work upstream. Reconcile the data once in a data warehouse or data lake, treat that as the source of truth, and build AI workflows on top of it. (For how a semantic layer gives AI consistent definitions of terms like “household,” see How Does a Semantic Layer Unlock Your Data for Agentic AI.)
What leading firms are doing about it
Two COOs showed what this looks like in practice, from very different starting points.
Elevation Point grows by bringing multiple advisory firms onto its platform. COO Caitlin Douglas described a firm that owns all of its own data. Its data lake, called Base Camp, is built on Milemarker and Snowflake. Every system the firm uses has to feed into it, and all of its AI and workflow automation is built on top of that foundation.
Halbert Hargrove grows organically and standardizes its tools. COO Cecilia Williams described a firm where every advisor uses the same systems. Its operational focus is integration: making sure systems talk to each other, data stays clean, and advisors know exactly where to find reliable information.
Different growth models led to the same conclusion. Own your data, and make it trustworthy before you build on it.
Why the stakes are rising
This matters more each year for two reasons.
First, clients want coordinated advice across their whole household, not a collection of disconnected accounts. Serving a household well requires seeing the full picture: plans, accounts, tax situation, estate considerations. Fragmented systems make that nearly impossible.
Second, the industry is consolidating fast. Panelists predicted trillion dollar RIAs, more mega deals, and a coming shakeout among wealthtech vendors. Every acquisition brings another firm’s technology stack into the fold, so integration debt compounds with each deal.
Where the work begins
At Cheesecake Labs, this is the pattern we see determine whether AI initiatives succeed or stall, in wealth management and across financial services. The model is rarely the bottleneck. The foundation is.
We are an engineering firm that builds that foundation. That means connecting fragmented systems, reconciling conflicting data, and building pipelines into modern platforms like Snowflake and Databricks, so AI has something reliable to stand on. We cover how we approach these choices in our data architecture decisions guide.
In the next article in this series, we look at what firms are building on top of that foundation, and why the AI chat window is only the beginning.
