AI for RIAs & Financial Advisors

The Build AI Podcast is a builder's show, hosted by Amit Goel of gAI Ventures, and this episode is unusual because guest and host are on the same team. Vineet Mohan is part of gAI Ventures and runs its international business, but he's spent the last year deep in one vertical: US wealth management and the RIA space, meeting dozens of advisors — some of it, Goel notes with a grin, on the beaches of Miami. The conversation is less interview than two operators comparing notes on how AI actually gets adopted in a hard, regulated industry.

That framing pays off. Because neither man is selling the other, the episode gets unusually candid about the mechanics — where the money is, where the tech is stuck, and why the winning approach in vertical AI is almost the opposite of the standard startup playbook.

An industry under pressure

Goel opens with the eye-watering numbers — asset-under-management figures he calls "unbelievable" — and asks what's actually happening in the US RIA industry. Mohan gives the tour. There are roughly 30,000 RIA firms and 300,000 financial advisors, spanning Morgan Stanley's thousands of advisors down to sole shops. Three forces are reshaping it. First, a massive generational wealth transfer — estimates range from $50 to $80 trillion — moving from baby boomers to Gen X and millennials, who invest differently and want to be served differently. Second, because investors skewed boomer, so do advisors; with a median age in the 50s, there's a projected shortage of about 100,000 advisors over the next decade. Third, consolidation: 200-300 M&A deals in 2024, 70-75% of them private-equity-backed, a trend continuing into 2025.

Then Goel raises margins, and Mohan sketches the squeeze from both sides. On the revenue side, the fee-based model that once charged over 100 basis points on AUM is compressing toward 70-80, pressured by robo-advisors at the low end and more efficient consolidators. On the cost side, an advisor shortage means paying more for fewer people everyone wants to hire.

"You've got fee compression on one side and rising talent costs on the other. That's what's creating the margin pressure — and it's why more firms are starting to realize tech could solve for some of it."

Fragmentation, and the Kitces map

When Goel asks how the industry sees technology, Mohan reaches for the Kitces map — the landscape chart of fintech tools across every vertical, from financial planning to CRM to portfolio management. In 2018 it had 180 entries; today it has 500-plus. That density is the story. A typical firm has to manage client relationships (a CRM), do planning (financial-planning software), connect to custodians, and run portfolio management — so at minimum three or four tools, and often six or seven as firms add insurance, tax, and estate services to fight margin pressure. None of them talk to each other, which means data entry and a support team burning hours reconciling records to prep for a meeting or open an account.

There's a fourth force worth naming that Mohan folds in: changing investor preferences. Money that once sat almost entirely in public markets is increasingly curious about private markets — a convergence the industry files under "alts," from PE and VC funds to private credit — driven partly by the newer demographic of investors coming in. New platforms are springing up to organize those investments, adding yet more tools to an already crowded stack. And compliance sits over all of it, with regulators focused on recordkeeping, investor communication, and now the AI tools firms adopt — another cost, another source of margin pressure.

On AI awareness, both are watching the same signal. Morgan Stanley's rollout of an OpenAI-built meeting-notes assistant, "Debrief," to all 16,000 of its advisors is, Mohan says, a great sign — when the largest player adopts AI, it tells the rest of the industry this is worth taking seriously. Below that, smaller firms range from DIY-ers using ChatGPT or Gemini for newsletters to skeptics worried, fairly, about sensitive client data — everyone's first question is whether it runs on central models or is exposed publicly. Mohan sorts the market into early adopters, cautious testers of the waters, and wait-and-see firms that jump once the industry moves. The clear trend: given the margin and talent pressures, the pro-tech firms will see benefits earlier. Goel adds the industry read — wirehouses like Morgan Stanley have taken the leap, regional firms mostly stop at meeting notes, and independent firms are the interesting ones, trying a lot of different things.

From point tools to an AI-managed system of record

The most forward-looking stretch is about consolidation of the stack itself. Goel, drawing on his time around a company that ran 75-plus SaaS tools, floats the idea that AI might finally produce a one-stop solution — stitching the systems of record together, doing the work of half the applications, and automating things that were never automated in the SaaS world. Mohan thinks it's becoming real. You could have an AI-generated, AI-managed overlay system of record: the AI listens to a prospect conversation, pulls the information, helps build the plan, and then — if the conversation goes well — reuses that captured data to pre-fill account-opening forms and client-service agreements.

"Some of those end-to-end flows can really be managed by AI now — which wasn't possible before, when you had multiple systems and different people looking at different parts of the workflow."

He's clear-eyed that others tried the one-stop-shop route — Orion, Envestnet — without covering everything. What's different now is that AI can bridge the gaps rule-based RPA never could.

Why SMBs move faster, and where the moat is

Goel offers a sharp observation from across industries: the US is at least 10x ahead of any other market on general AI adoption, yet its large enterprises are still stuck in proofs-of-concept while SMBs are the surprise adopters. Decisions run straight through a founder or owner, US labor is expensive enough that automating non-core work makes obvious sense, and there's less self-inflicted regulatory and IT drag — smaller firms, he argues, spend less time inventing problems that don't exist. He points to an RIA managing about $25 million that's already using AI tools, the kind of adoption you wouldn't see elsewhere in the world. Mohan agrees it tracks with his conversations — smaller firms decide quickly and love to experiment; larger firms move slower through bureaucracy and longer processes.

The two also get into the builder's weeds. Mohan's co-founder and CTO, Kushal, started on LangChain for the agentic middleware, then built their own to avoid disruption from the constant API churn — something breaks or changes every few weeks — which means also doing "horizontal" AI work to keep that layer best-in-class, on top of the vertical product. It's not a one-time cost, either: Mohan notes that an AI system solving a problem cleanly today will start throwing up issues on its own if you leave it untouched for two months, so continuous evals and testing are part of the job. The document-processing engine and a supervisor-plus-agents architecture — he mentions building something like 23 agents under a supervisor — exist because the bar on hallucination and edge cases is now unforgivingly high. That leads to the episode's real thesis, and it's Goel's model for gAI Ventures stated plainly: diametrically opposite to Y Combinator. Rather than backing a fearless 25-year-old to disrupt an industry from outside, they partner with an entrepreneurial operator who has lived the workflows for years and give them the startup system, the AI tooling, and the capital.

"Industry-first, problem-first beats AI-first. We've seen this movie before in fintech and blockchain — once you have a hammer, everything looks like a nail. Better to be slow, even boring, and build the painkiller for the most important problem."

Mohan closes on where the moat lives once technology itself is available to everyone. Vibe coding means a product manager can spin up an MVP in hours; soon anyone could clone a piece of software by feeding it into a coding tool. So the durable advantage, he argues almost old-fashionedly, is relationships and deep problem understanding — knowing, like a CFO he'd spoken to the day before, that the first version of an accounting AI has to be an Excel plug-in because you cannot pry CPAs off their spreadsheets on day zero. Technology becomes the level playing field; what you bring on top of it becomes the moat. For a podcast aimed at builders, it's a bracingly unglamorous conclusion — and probably the right one.