Compound Planning runs $5 billion on software its own engineers write

Compound Planning sells a modern family office at scale, and the machinery behind the pitch is a ten-person engineering team on the firm's own payroll. The RIA runs roughly 50 advisors and $5 billion in assets, up from $2 billion less than two years ago, and co-founder and CEO Alex Farman-Farmaian spent most of a WealthTech Today podcast hour on the software his advisors and clients actually touch — the layer most practices of this size license rather than build.

Farman-Farmaian came up on the sales side of B2B software and spent years inside software companies before starting Compound with two Y Combinator co-founders, and that background shows in what the firm chose to own. Advisor HQ is the platform its advisors work in; Compound AI is the agent built on top of it. The episode discusses what sits underneath — notetaker transcriptions, model selection between OpenAI and Anthropic, and open-weight models still an open question — and walks through Compound's build-versus-buy line, with Advisor HQ and Orion as the two examples. "Building the technology and building software is in service of the client-advisor relationship," he says, which frames the engineering as a means rather than the product.

Renting is the default at this size, and it is not cheap in the ways that matter. A $5 billion RIA with 50 advisors can staff a complete technology stack without employing an engineer: a CRM, a planning tool, a custodian's models, and increasingly an AI seat billed by the head. Schwab's Claude integration costs an advisor $240 a year per seat, and the work that makes a seat useful — the rewritten processes, the added review steps — never shows up on the license line. Compound inverts the arrangement, converting software bills into payroll so that the workflow its advisors live inside becomes an asset the firm owns outright. The episode still walks through the custodians it runs on and the specialized partners it plugs into, a reminder that building your own desktop does not remove the decision about where client assets sit.

Compound AI is where Farman-Farmaian puts the time savings, and he is specific about the unit: "With Compound AI, we can cut that time down by hours for each meeting." The division of labor he describes is deliberately narrow: as models absorb technical work, the human relationship becomes more important rather than less — "I believe that human relationship's going to be more and more important, but AI will be doing a lot of the technical work in the future" — which is the claim every AI-forward RIA makes and the one hardest to falsify from the outside.

The CRM thread is the one most likely to travel to other practices. "The CRMs are only as good as the context that lives in them," Farman-Farmaian says, and the episode takes up what happens to that context when an advisor team arrives from another firm: the notes, the history, the workflows, all of it belonging to people who ran their practice somewhere else until now. That is a recruiting question as much as a data-migration one, because a team brings a book and a way of working, and only the book moves by paperwork. The show also puts the harder version to him — whether he would build advisor onboarding in-house if he were starting Compound today.

Compliance is where the build decision stops being a matter of taste. The episode examines which parts of the business AI will and will not touch, and how a compliance function handles security, SOC 2 reviews, and vendor oversight when so much of the technology is built in-house. The logic that follows, though the episode does not spell it out, is that oversight work does not disappear when the vendor is you: a firm that writes its own software still has to produce the security evidence and review discipline it would otherwise demand from a third party, for its own examiners and for any custodian that asks.

Farman-Farmaian has a line for the alternative. "There is so much slop out there, and there is an elegant way to use AI and then there is a sloppy way to use it." The difference usually comes down to whether the process existed before the model arrived. A firm should fix the process before it buys the tool — the same conclusion from the opposite direction — and Compound's version is to own the workflow so it can be rewritten whenever the firm likes.

The episode's last question has the widest application: whether a ten-person engineering team is something other RIAs should copy. The arithmetic that would settle it is not in the pitch. Five billion dollars spread across 50 advisors is about $100 million of assets per advisor on the firm's own numbers, a density that carries overhead most practices cannot, and doubling the book in under two years is what makes an engineering budget read as growth investment rather than a cost center. Compound sits at the far end of a split this publication has described before — firms that own their workflow versus firms that buy a seat by the head — and the show closes by asking how Compound measures the return on that payroll against its growth. The comparison of what ten engineers produce per dollar against what they cost is the piece of the model a practice needs in hand before it trades a software subscription for a salary.