Notes from Underground: AI x Wealth Management
Notes from Underground
AI x Wealth Management: What I tell every founder building in this space
Let’s start with fees. Everyone hates them and any AI native solutions will have to eliminate them, or at least reduce them. For old-school RIAs fees are genuinely hard to reduce because the economics are tighter than people assume. Many RIAs operate at roughly 25-30% operating margins. That’s before you account for growth or market cycles. The unavoidable costs of regulatory and registration overhead, expensive legacy OMS, data vendors (Addepar) add up and then you need teams of analysts to actually use it. Sure, the likes of Gamma and Hebbia will help with presentation building and analysis, but we are not there (yet).
Even in an AI native end state, the core issue remains: wealth management is fundamentally a relationship business, especially in the US.
I’m not saying AI won’t get good at this. ElevenLabs voices will be indistinguishable. Memory will get extraordinary. Agentic advisors will remember your kids’ names, schools, and last summer’s vacation. Still, there’s a human-to-human delta. People like feeling known, not just remembered. Could that delta be compensated for? Yes. If the trade-off is zero brokerage fees? Absolutely. But you aren’t just competing on intelligence. Returns matter way less than you think if you are within a standard deviation of the market. This is a game of trust, status, and reassurance.
At most firms, portfolio construction is often outsourced or centralized. At GS, teams would get in trouble for deviating from ISG’s (the centralized investment strategy group) weightings. The advisor’s real value was planning, coordination, and keeping clients on platform.
Many AI-native products are attacking portfolio construction in a market where that is not the primary USP. The rest of the job is what makes AUM sticky.
Even when you build great relationships, clients don’t always “follow their advisor”. Clients stay for access.
Bulge-bracket firms with investment banking arms can offer pre-IPO private placements, exclusive private market access, structured products, and deal flow that simply does not exist elsewhere (shout-out to my friend Chen Chowders leading art-world advisory at HSBC).
AI can’t conjure overnight. Access is institutional, reputational, and slow to build. Even superior products lose to the perception of access all the time. Founders need to be brutally honest about whether they are building around access, intentionally targeting segments where access matters less, or pretending access doesn’t matter at all.
“Wealth management” is not a single market. A HENRY and a HNWI have fundamentally different needs. Even within UHNW, the sub-segments matter. A family office often doesn’t make sense below ~$10B, and bulge-bracket coverage doesn’t make sense for a $1B+ client. A $50k-net-worth 20-year-old and a $50k-net-worth 40-year-old are not the same customer. Be specific!
My cardinal rule: people like being anxious about money. This is usually the first thing I say to anyone building in this space.
I don’t mean that people enjoy anxiety (though if you know me, you’d think I do) but the activation energy required to fix it is far higher than most entrepreneurs expect. People know they should budget, plan taxes, rebalance portfolios—and many still won’t. Even if it’s one click. Everyone believes their situation is exceptional. I’m the only expat who’s ever dealt with dual taxation. You are the only homebuyer using alternative, non-mortgage loans. Your friend is the only employee facing a massive tax bill from vested options. This isn’t true, obviously, but the result becomes a fragmented and challenging GTM, necessary white-glove onboarding, bespoke-feeling solutions.
THIS IS NOT A CALL TO STAND DOWN!
Despite all of this, I’m not bearish.
Someone will crack this.
We’re already seeing pressure points: RIA consolidation, incumbents experimenting with copilots, startups attacking adjacencies like tax, reporting, and planning rather than the core relationship.
This is genuinely hard, which is why I like it.