Bassam Chaptini

The through-line

Chaptini's argument returns again and again to a single claim: client management is the unsolved white space in financial services, and the reason it has stayed unsolved is that the underlying data was modelled wrongly. "For us client management has always been a white space in enterprise. It's something that big enterprises and wealth managers alike have suffered through for a very long time" [1]. Sales tooling captured the easy part of the lifecycle, and everything after the signature was left to manual work. "As soon as a client says sign me up is where the complexity starts because a client is not just a client, it's a household or a family crew. It's individuals and entities" [1]. He is explicit that this has been attempted before and failed: "It's been attempted a few times to be solved in with different platforms. We've seen it firsthand. It just doesn't work" [1].

What changed the odds, in his telling, is AI, but not as a feature bolted onto existing systems. AI "gave us ammunition" to attack the problem "from a data perspective" first [1], which is why the knowledge graph, not the model, is the foundation of everything he describes. An engineer by background, he frames the whole company as a return to fundamentals: "you go back to the fundamentals. The fundamentals is how you represent the space from a data perspective because that infinity complexity comes from how you're representing all that data and how it's related to each other" [1].

On onboarding as the hornet's nest

He treats onboarding as the most underrated hard problem in the industry. "Onboarding I will say is an underappreciated complexity and it is I refer to it as the hornets nest" [1]. The nest is combinatorial: data has to be collected in the way each firm wants it represented, in the way each custodian wants it represented, with layers of optionality and firm-specific extra forms on top, compounded again when the same intake also feeds financial planning and other services [1]. His diagnosis of why the industry defaults to manual work is unsentimental: the complexity is so high that "it's easier for them to fill out PDFs, send out emails than to create scalable solutions" [1].

The payoff he claims from solving it at the data layer is that the complexity disappears from the user's view. "When I onboard a client, I'm not onboarding them 15 different times to the different products they have. I onboard them once and then I open up their account with custodian one, custodian 2, custodian 3. It's all part of the same onboarding" [1]. He extends the same principle past the wealth boundary: life and annuity products, normally a separate onboarding entirely, fold into the one experience [1].

On modelling the human before the product

The design decision he is proudest of is inverting the traditional order of representation. Historically, he says, "people had a product view of the world," starting from an account and mapping a person to it [1]. Avantos starts from the relationships instead, because that is where the difficulty actually sits: "A given person can be a beneficiary of a different household, can own multiple products. A given person can be part of multiple family groups. A product is not just a product like a brokerage account is a different account if it's with a given custodian from the other one" [1]. Advisers are part of the same graph, since "it's not just one adviser typically onboarding a client. Sometimes it's a team of people onboarding a client" [1].

He is direct about why tables were the wrong tool: "Tables historically has been set up to represent things that are more linear" [1]. And he owns the graph as a point of view rather than a neutral schema, calling it "an opinionated way of thinking about the space and how we construct it" [1]. His summary of the philosophy is that a client "does not exist in an abstract form" but only in relationship to the products they own and the adviser servicing them, and "all those relationships are equally important to be able to model that complexity" [1].

On refusing data migration

A firm rule of deployment is that customers keep their existing system of record. "Life is too short to do any data migration. Showing up to a wealth manager and saying great migrate your data into ours is a nogo. Is a nogo for us" [1]. Instead, the client master stays in the CRM or bespoke database where it already lives, AI maps that data into the knowledge graph so it can carry the necessary complexity, and bidirectional connectivity keeps the original source of truth updated with anything changed in Avantos [1]. The practical effect he wants is that the adviser operates on "a knowledge graph that has the full context of the clients, the products and the advisers" while the firm's own data architecture is left alone [1].

On context being the scarce resource in AI

His sharpest general claim about AI is that data is abundant and context is not. "The name of the game right now in AI is context. It's not data per se. Everybody has a lot of data" [1]. He grounds this in what advisers do naturally: they hold the client's history, last conversation and holdings in their heads and stitch that context together live in conversation. "An AI will not be effective if it doesn't have that context, because it's going to miss stuff. And it's not because of lack of data. They will have access to the data" [1]. He agrees strongly with the framing that generic AI use is like firing a shotgun at the universe and that narrowing the attack vector raises effectiveness sharply, adding that narrowing alone is insufficient without supplying the context [1].

This is why he positions the knowledge graph as the foundation of the AI rather than an adjacent asset: "that knowledge graph is the foundation of our AI because it is fully contextualized. Every interaction is fully contextualized. The AI can truly be a good assistant now to an adviser. It has access to the same context that the adviser has" [1].

On permissions, human in the loop and enterprise trust

Because the buyers are enterprises, he treats security as a design constraint on the AI itself, not a wrapper around it. "The AI would not have access to any data that an adviser does not. It's very important that it has the same permissions as an adviser has" [1]. He gives the concrete case: within the same firm, an adviser may not be entitled to see a client's family assets, and "the AI will literally mirror those permissions" [1].

The same conservatism governs autonomy. Agentic, in his usage, always means supervised: the system surfaces recommended actions and waits. "We have always, this is something that we've put in place from the get-go, a human in the loop to confirm a lot of the AI things" [1]. When a client mentions a newborn, the platform may propose opening a 529, updating beneficiaries or recommending a trust, but "we always have a human in the loop doing it. Once confirmed is when our agentic workflow will start" [1]. He ties this explicitly to "various compliance reasons," and accepts that even a well-targeted, well-contextualised model retains a probability of failure [1].

On relationship intelligence and client journeys

He regards client 360 as table stakes and describes taking it further by joining client data to products and to the journeys and services in flight, so recommendations are grounded in the actual relationship [1]. Crucially, the AI is not generic: "it's not an abstract AI that is a generic way of thinking of wealth management. We have configured our AI to understand those client journeys in wealth management very intimately" [1]. Life events, a newborn, a marriage, a divorce, a change of job, are detected from conversations and other signals, and the system proposes the best next action [1]. These are the moves an adviser would think of anyway; the platform's contribution is surfacing them reliably and then executing once confirmed [1].

On servicing across custodians

The second recurring pain he names is servicing, where multi-custodian firms face "15 different flows for the same service event" [1]. Avantos handles this with deep integrations into the custodians themselves, naming Schwab, Fidelity, BNY and Goldman Sachs, and insists on carrying the action through rather than stopping at its own boundary: the platform will "go all the way to the custodian and execute" [1]. Where a custodian offers no API and the process is form-based, the platform fills the form behind the scenes and routes it through the wealth manager's channel, so "for an adviser, they still have a digital way of doing it except we're filling forms behind the scene" [1]. He frames the mapping burden as something the vendor should absorb entirely: "we guarantee that mapping for our clients so that they don't have to deal with that complexity behind the scene" [1].

On building cross-product from the start

Chaptini insists the company was architected as multi-product before it was single-product in market: "we have built the company from the ground up to be a multi-product company" [1]. The rationale is the client's own frame of reference, since clients "want advice on their financial portfolio, on their financial well-being versus just a brokerage account or my taxes" [1]. Wealth was the entry point because it is a passion area, a genuine industry need, and because "the nature of the complexity of the relationship is unique to wealth" [1]. From there the expansion is into life and annuity insurance, which carries its own complexity and its own servicing workflows, into banking products, and into capital markets, where investing in alternatives involves "the same servicing, same onboarding, same type of things" [1]. He is candid that "operating system is a big surface area. It means a lot of things to different people," and describes the ambition to occupy that surface area fully as deliberate [1].

On building in stealth with a design partner

On how the product was validated, he describes roughly a year in stealth with Mercer advisers as design partner, building from the ground up around specific personas: advisers, the operations staff supporting them, and customer success people serving clients [1]. Only "once we had tremendous impact with the advisers and with the operations people is when we officially went out of stealth mode and launched the company" [1]. He flags his own bias about the product while pointing to traction across traditional RIAs and enterprise wealth managers, and to partners announced with the Series A including Vanguard, Guardian Life and SEI, whom he characterises as trusted partners wanting to deploy the technology at scale [1].

Takeaways

• Sales pipeline tooling solved the easy part of client management; the complexity begins the moment a client agrees to sign, because a client is a household of individuals and entities across multiple products and custodians [1].
• Onboarding is "the hornets nest": firm-specific representations, custodian-specific representations, optionality and extra forms make manual PDFs and email genuinely easier than building scalable flows [1].
• Model the person and their relationships first, not the product; tables were built for linear data, and the difficulty lives in the relationships between entities, so a knowledge graph is the right primitive [1].
• Never ask a wealth manager to migrate: keep their CRM or bespoke database as system of record, map into the graph with AI, and maintain bidirectional sync [1].
• "The name of the game right now in AI is context. It's not data per se." An AI without the adviser's context will miss things even with full data access [1].
• Agentic means human in the loop by design, for compliance, with the AI mirroring adviser-level permissions so it can never see data the adviser cannot [1].
• Servicing should execute all the way through to the custodian, including filling form-based processes behind the scenes where no API exists, with the mapping guaranteed to the client [1].
• Build cross-product from the start, because clients want advice on their whole financial life; wealth first, then life and annuity, banking and capital markets [1].