Rabih Ramadi

The through-line

Ramadi's recurring argument is that financial institutions were built around products rather than around people, and that every operational problem in wealth management flows from that original ordering. "The fundamental challenge in wealth management and across financial services is, uh, the businesses have grown traditionally around products, not around the client itself" [2]. A firm launches an investment product, a private market product, an insurance product, a banking product, and only then asks who the client is, so it ends up looking at that client through a product lens instead of mapping the right products to the person [2]. He is careful to say the ambition to be client-centric is not new: "people always had the idea to be client-centric. So, it's not like we created a new concept. The challenge historically was, technically it was not feasible" [2].

The second half of the through-line is what makes it feasible now, and here he is deliberately unromantic about AI. AI is only as useful as the context it is given, whether that context is a client data model inside a financial institution or an engineer's judgement about what to build. Avantos is positioned as an operating system for how a financial institution manages its clients, unifying data across fragmented systems rather than solving every problem in the sector [2], and the same logic runs through his view of engineering, where AI is a multiplier on people he has no intention of replacing [1].

On why the tech stack fights the client

The technical obstacle to a single view of the client, in his account, is fragmentation rather than ambition: "everybody wanted to have a single view of the client... but technology it was difficult to do because of the fragmentation of the tech stack" [2]. Product codes differ, custodians differ, and the core systems underneath each business line have almost nothing in common, since "there's a loan system is very different than um the insurance underwriting system, very different than the custodian system for wealth" [2]. A great many of these cores are still mainframe-based, and he does not expect that to change: waiting for the underlying systems to modernise "may not be in our lifetime" [2].

His conclusion is to leave them alone. "We're not going to replace every system that was built in financial service. That's not the point because if you do this, it would take forever and adoption would be super hard" [2]. Nor is the answer more data. "It's not about creating new data sets. There's tons of data floating around any financial service companies. It's more around like how can you put a context for that data?" [2]. The stated approach is to leave the systems where they are and "build a modern AI architecture on top of them to allow for faster distribution", giving the advisor an abstraction layer they can navigate without dropping into the antiquated cores beneath [2].

On why workflow-first automation stops working

Ramadi is blunt that the conventional attack on client servicing, building workflows faster, has a ceiling. "People typically tackle the space by tackling the workflow... The problem with this is it works till it stops working" [2]. Scaling breaks down quickly because the permutations of data variation and workflow variation "can be the infinity" [2].

What he argues you need instead is a data foundation that models three entities and the links between them: the client, who is really the household rather than the individual; each financial product, each with its own data definition for onboarding and servicing; and the agent, meaning whoever touches the client, from advisor to operations [2]. "Unless you can't find a way to build the data foundation with sophisticated data model that link these three entities, you can never scale the workflow of the servicing in the space" [2]. Avantos was built foundation-first, with AI infrastructure, orchestration and UI layered on top [2].

On why AI adoption in financial services has stalled

He offers a diagnosis of the industry's disappointing AI results that follows directly from the context argument: "without providing a context, the AI can never be effective in the space. This why you see in the industry, the AI adoption is nowhere where it should be. It has been deployed more for finite uh on the fringe activity versus core activity" [2]. Without a way to give context to the data, AI is limited both in usefulness and in accuracy [2]. The roadmap he describes is explicitly data-gated: "the more data we have, the more we can develop AI skill sets", running from recommendation engines to an unstructured data engine to a meeting engine, with enough agent use cases to fill three years of work as more data is ingested [2]. He has also discussed how AI agents apply specifically to client onboarding and servicing for financial institutions [5], where AI-native platforms change how advisors interact with data and clients [4], and where AI belongs in the advisor tech stack for RIAs [6][7].

On defining the category: client lifecycle management

Ramadi accepts that what he is building does not yet have an established slot in the buyer's mental map of CRM, onboarding and client experience tools. "I don't know there's a category yet for this space, but we we call it client management or client lifecycle management" [2]. He draws the boundary firmly at the point of sale: prospecting and lead generation are out of scope, at least for now, and everything after that is in, covering "everything you do to when you touch a client across the services you offer to them, uh all the way from the advisor with the advisor do to service the clients to operations to compliance" [2]. The focus is deliberately narrow: "We are very focused on the way uh, to solve the way basically a financial institution manage the clients" [2].

On the multi-product convergence

He believes strongly in what he calls the multi-product world, and says it is already happening: banks selling wealth management, insurers selling wealth, wealth managers selling banking and insurance [2]. The client-side logic is simple. "Nobody wants to work with 10 different institutions to uh do their financial wellness. They'd rather go with one provider" [2]. The institutional logic is that if someone is already your client for one product, there is no reason not to offer them others, even where you do not manufacture all of them [2]. Larger firms want to be able to say yes to a loan, a yacht insurance policy, an investment product or an alternative in the same conversation, and today the integration burden makes that operationally very hard [2]. The stated plan is to span banking, wealth and insurance as well as the broker-dealer space, adding insurance products, alternatives and eventually banking to a platform currently centred on wealth and investment products [2].

On what is reshaping US wealth management

Two forces stand out for him. The intergenerational wealth transfer is primarily an operations problem: transfers of assets and changes of ownership are complex events that humans can absorb at low volume and cannot absorb as volume climbs, which makes proper automation unavoidable [2]. The second is consolidation. Independent wealth managers now exceed 500 or 600 billion dollars in AUM, and "these are not boutique firms anymore. These are real financial institution with big ambition with ability to go into multiple products and business line" [2]. He sees these firms professionalising and coming to resemble banking institutions in capability and technology, with the infrastructure obliged to keep up, producing an industry with "a lot of velocity, but uh a lot of focus on operational scale" [2]. Complexity compounds along the way: "The more product you have, the more regulation you have, the more complexity you bring to the space" [2].

On results, deployment and the shape of the customer base

Ramadi quantifies impact cautiously and flags the limits of his own evidence, noting it is early and moving quickly. He cites 30 to 40 percent gains in advisor productivity measured by client coverage and roughly 50 percent operational scale efficiency, while declining to put a number on retention or upsell: "I can't quantify it yet, but the expectation is client retention metrics will go up substantially and client upsell organic growth with existing clients should go up substantially, too" [2]. The current focus is large complex institutions including Guardian Life, SCI, Vanguard and Mercer, with partnerships intended to reach mid-market firms in the 10 to 30 billion dollar AUM range, and Guardian Life notable for spanning both wealth and insurance [2]. Growth plans extend internationally to London and Canada [2].

On deployment, he names integration with existing systems as the hardest part, compounded by the absence of standards, so firms customise pre-built journeys such as account opening and money movements rather than building them from scratch [2]. His answer is accumulation: connectors already exist for the big custodians, major portfolio management systems and CRMs, and each one added "drastically reduce the implementation timeline" [2].

On choosing clients as partners

Client selection was treated as a product decision. "We were very uh careful and selective on uh choosing which clients we want to work with at the beginning because our clients currently we don't call them clients. We are we are we call them partners" [2]. The reasoning is that influential firms carry insight into industry trends and can materially shape the roadmap, which is set jointly with existing and future clients alongside industry advisors who give continuous feedback [2]. The founding pattern was the same: the idea was taken to Mercer Advisor, who agreed to become the first client before the company existed, and the product was built with them before funding was raised [2].

On AI and engineers

Ramadi's position on engineering is that AI expands output rather than reduces headcount. "I don't want to lose my engineers. I just want them to get better at using AI to go faster, do more. I don't know if I want to hire more engineers, but I don't want to lose a single engineer" [1]. The payoff he is chasing is speed of response: a customer says on a call that they wish the product did something, and tomorrow it does [1]. He contrasts this with the reality of team-based delivery, remembering solo programming in 2002 when he could build form groups over a weekend and show his partners on Monday, against today's planning, sprint and release cycle [1].

He also treats AI fluency as a hiring signal, and a fairly hard one. "Even our interview process, if you have engineers interviewing with us that are not using AI, uh that's always a bad indication, because why not?" [1]. His definition of quality follows from that: "the good engineer is the one that leverages everything that currently exists and be 100 times more productive by leveraging what currently exists versus reinventing the wheel" [1]. He notes that many engineers are afraid to learn AI, and expects the industry to change drastically because the best ones already embrace it in everything they do [1].

Takeaways

• Financial institutions grew around products rather than clients, which is why they lack a holistic view of the person they serve; the desire to be client-centric is old, but until recently it was technically infeasible because of tech stack fragmentation [2]
• Do not rip out core systems, many of which are still mainframe-based and unlikely to change soon; build a modern AI architecture and abstraction layer on top so advisors never have to touch them [2]
• Workflow-led automation "works till it stops working" because data and workflow permutations tend to infinity; scaling requires a data foundation modelling client households, each product's own data definitions, and the agents servicing them [2]
• AI adoption in financial services has stayed on fringe activity rather than core activity because the data lacks context; more data is not the problem, joining it is [2]
• Reported impact so far is 30 to 40 percent advisor productivity gains in client coverage and around 50 percent operational scale efficiency, with retention and upsell effects positive but not yet quantifiable [2]
• Expect banking, wealth and insurance to converge, because clients do not want ten providers for their financial wellness and institutions want to sell across product lines they do not manufacture [2]
• Wealth transfer and industry consolidation are the two structural pressures: ownership changes are hard to handle manually at volume, and independents past 500 billion AUM now need bank-grade infrastructure [2]
• An engineer who is not using AI is a bad signal in an interview; the goal is to keep every engineer and make them three to twenty times faster, not to hire fewer of them [1]