Trending
The System of Record Is a Solved Problem.
Everyone in wealth management is racing to hold the client record. The record was never the prize.
Nine months ago I wrote that the CRM was built for the wrong era. Architected decades ago for a job that barely resembles modern advice, and long overdue to evolve or be replaced. You can read that piece here.
Since then the market has moved, just not evenly. Some vendors bolted AI onto what they already had. Others went deeper and rebuilt around it. And the conversation sharpened into a single question: is the notetaker the next CRM?
AI is everywhere in wealth management right now, and that question keeps splitting into smaller ones. Is Jump - Advisor AI building a CRM. Is Altruist's Hazel becoming one. Heck, this even became a running joke on my podcast AI for Advisors that I cohost with Mark Heynen
as a question often posed by Joe Moss. Will the incumbent CRMs keep bolting on the notetakers now circling their turf, or build their own and cut them loose. Those are the headlines. They also miss the story.
The more useful question is which vendors are building with intention and which are building on FOMO.
From the outside you can't always tell, because the announcements are interchangeable. Everyone ships a notetaker, an assistant, an agent, a data lake. Same press release, different logo. But underneath the sameness, two different things are happening. One group is racing to hold the client data points. The other is quietly building the thing that brings meaning to it. And the noise about who is becoming whose CRM is drowning out the only distinction that will matter in a few years.
So let me say the part the headlines skip. The race isn't for the system of record. The ingestion point doesn't matter.
The prize is the synthesis in the middle, and the experience it delivers.
The turf war everyone is watching
Pick up any coverage from the last few months and the framing is the same. InvestmentNews described it as a turf war for central control of the wealth firm's tech stack. Craig Iskowitz put it more bluntly, that "The AI notetaker is eating your CRM". Ezra Group went further and predicted the standalone notetaker category won't exist in its current form by 2028. Salesforce, whose Financial Services Cloud holds somewhere between four and seven percent of the advisor market, shipped a six-part agentic suite to defend the seat, and Michael Kitces called that defense inevitable.
Michael Kitces response to Investment News article on Salesforce
The capital flowing to these AI-enabled wealthtechs agrees. Jump is past 35,000 advisors on an $80 million Series B. Zocks counts more than 5,000 firms on a $45 million Series B. That money is chasing one idea: become the place the advisor lives, and demote everything else to just being a backup.
They are all right that a fight is happening. They just might be fighting over the wrong hill.
The record is already a solved problem
Strip away the language and the system of record is an audit trail. Every email, every note, every transcript, every document, retained for completeness and compliance. That is a database. More precisely it is a data lake, which is exactly what Orion named the thing sitting under its new Denali platform. Data lakes already won that job. Capture, storage, retrieval, all solved, by companies whose entire business is doing it well.
The CRM earned the "system of record" title honestly. For years it was where the client data lived and where the governance happened, the audit logs, the retention policies, the reporting. That combination made it the hub. But AI changes all of that! Most of what these AI tools do well does not require reading the entire record. In fact doing so is not ideal from virtually all angles. It's token heavy (that just means costly), it requires the AI to reconstruct history from data points that likely have gaps and almost definitely lack context, it's slow, and it poor architecture. The value of an assistant is not that it can recite every note. It is that it knows the one thing that matters right now.
So when a vendor races to become the new system of record, it is racing to win the part of the stack that is already finished.
It is a filing cabinet with better lighting.
The value moved up a layer
The prize sits one level above the record. Not the data. The synthesis of it.
What I need to serve a client is not every data point in their plan. It is not the internal rate of return on every position they hold. It is an understanding of what all of that means for this specific person and the goals they told me matter. Whether the picture is moving toward those goals or away from them. What changed since we last spoke, and why that change matters.
That is not integration between products. Integration moves data from one box to another and calls it progress. This is synthesis across products, the judgment that turns a pile of facts into a point of view.
And almost no one is building it, at least not on purpose. The dominant approach is to accumulate the record and hope the model sifts it at query time, on every question, forever. That works right up until the record has holes in it, and it always does.
The closest the industry has come to that judgment layer showed up well before this AI wave. Bento Engine built its whole model around surfacing "moments that matter," scanning the book for the life events and age milestones where a timely, human conversation changes the relationship. Narrow and rules-based, but synthesis all the same, deciding what is worth surfacing instead of storing everything and hoping. That work is exactly what today's AI assistants are reaching for, and the telling part is that some are plugging into it rather than rebuilding it. Zeplyn | Agentic AI for Wealth Managers wires Bento's life-event signals straight into its workflow, so a moment surfaced in a meeting comes back with the advice already attached. The direction is right. What matters next is the judgment that knows not just that a moment happened, but what it means for this particular client and the goals they set.
This isn't theory, I implemented it
I built an AI teammate called Woodhouse. He and the agents he coordinates operate on exactly this principle, and watching it work day to day is what convinced me the record is the wrong obsession.
The agents carry the context that actually matters to my business. Who I can help and who I can't. What I offer and how I engage. Who I won't work with, and why. How we think about problems, our methodology, our rates. That context is what lets them judge what is important to me, my team, and the people we serve, instead of treating every input as equally relevant.
When something material changes, the context updates with the essentials. Why it changed, when, and what it now means. That is it. There is no nightly re-reading of every email and transcript to reconstruct the state of a relationship. The work is done once, at the moment it matters, and maintained forward. Each night Woodhouse updates the context for everyone I engaged with that day and sends me the deltas. If he missed something or read it wrong, I correct it by chat or voice. At this stage, that almost never happens, but it took time to dial in what context was needed about me, my firm, and our clientele.
The full record still exists in a database. Every detail is retained for audit and completeness. But neither I nor the model needs to wade through it to know where things stand. The context stays small because small is the point. We took this same model and applied it to our project management platform that helps us run all our engagements based on each specific client and scope. We track every note, comment, status change, meeting, etc. but it's virtually all supporting data. At the time of ingestion we look at current state and our AI project manager makes suggestions on what to update and what to tackle next. Current state is updated with the new insights and the rest is stored in the database.
Two things fall out of this
The first is that the "AI needs clean data" pitch collapses, at least the version of it that sends you down a path of cleansing fifteen years of history. You've heard it, on stage at conferences, and on webinars all from "wealthtech experts".
They say things like,"Before you can do anything with AI, you must convert and scrub fifteen years of messy history."
Clean data matters, no argument there. But a multi-month project to scrub the archive isn't the same thing, and it's not what makes AI work. Even if you finish it, you haven't fixed the actual problem: how the data is stored and the fact that most of it carries no context about why it matters.
Clean data was never the ask. Accurate context is.
While you may end up with cleaner data, it's data that no one will ever look at, and it's still not better positioned for AI. Because the model works from current context, that cleanup was never the requirement. What you need is an accurate snapshot of what is true now and the governance to keep it accurate as you go. The archive can stay exactly as messy as it is, sitting quietly in the data lake for the day compliance asks.
The second is that this is the actual engine behind the return of the all-in-one platform, and it was never about integration. For fifteen years the trade-off was brutal. Accept a monolithic platform that is mediocre at everything, or assemble best-of-breed tools and pay the integration tax forever. An AI interface dissolves that trade-off. When synthesis is the surface you work from, the number of applications underneath it stops mattering to the person using it. You do not pay down the integration tax. You stop paying it, because you are no longer working in the seams between products.
When synthesis is the surface you work from, the number of applications underneath it stops mattering to the person using it.
What still has not happened
One piece of this remains unbuilt, and it is the same piece I pointed at nine months ago. Communication.
Everyone captures communication now. The meeting gets transcribed, the email gets logged, the note gets filed. What almost no one has done is own the channel itself, and fewer still own it on both sides, advisor and client. That gap is not cosmetic. Every channel you do not own is context you cannot see. And the moment you hand the analysis to an AI, a channel you cannot see stops being a minor inconvenience. It becomes a wrong answer delivered with confidence.
When a human advisor was the interface, that advisor's memory quietly patched whatever the systems missed. Move the reasoning to AI and the patch is gone. The model can only work with what it can see. Which is why owning communication end to end, both sides of the relationship, is the last layer genuinely up for grabs. It is also where the client experience finally gets built instead of deferred to whatever sits downstream.
Stop racing to hold the file. Build the thing that reads it, and the channel that keeps it true.
No one actually owns the record
Here is the part the whole turf war overlooks. No one actually owns the client record. Not all of it.
Custodial data sits with the custodian. Performance lives in the portfolio system. The plan is in the planning tool, the estate documents are somewhere else again, and held-away and outside assets are scattered across statements the firm may never see in full. Every platform that calls itself the system of record is really the system of part of the record, stitched to the rest through integrations that break and feeds that lag.
So the race to own the record only means something if you can truly own it, and right now no one can. You can watch firms attack that from opposite ends. Altruist is closing the gap from the custody side, pulling more of the data back to its source. Orion with Denali, Collation.AI, and Milemarker, are going the other way, staying agnostic to where the data originates and trying to own the lake it all pools into, along with the experience and the judgment layered on top. Advyzon is betting an all-in-one can simply own more of it outright. Different strategies, one shared admission underneath them: whoever holds the most complete and most current picture wins, and holding it is the whole difficulty.
Which is the tell. If owning the record were the prize, the winner would be whoever piles up the most of it. It isn't, and they won't be.
The moat becomes the liability
Follow this to its logical end and the vendors racing to lock down the client record are not just chasing a solved problem, they are choosing the losing side of it.
The client record was never the prize, so a vendor that walls it off is defending territory nobody actually wanted. Worse, that decisions comes with some serious obligations and expectations. If you will not let a firm connect the planning tool, the portfolio system, or choose the AI agent it actually wants, you have just volunteered to be all of those things yourself, and to be the best at every one of them. No exceptions, no weak links, forever.
This used to be a survivable strategy. The integration tax was high enough that "good enough at everything" beat "best of breed, but poorly connected." That math has flipped. The tax has come down fast, and AI is pushing them lower, because advisors no longer have to manually connect the dots between the systems. The synthesis layer does that now.
Wall off the platform and you have not protected your position. You have signed up to win every category at once, with no help from anyone.
The vendors who will hold up are the ones who went the other direction: open, secure sharing (Developer API, User API, MCP), and partners with whom they share that same client data with whomever else the advisor already trusts. Not because they are generous, but because it is the only strategy that does not require being the best financial planning tool, the best CRM, the best rebalancer, the best everything else, simultaneously. An AI layer that can only see what happens inside its own walls is half-blind by design. One that can see across the advisor's actual stack, including the tools it does not own, is the one that can tell you what really changed and why it matters.
The whole game
The system of record is a solved problem. The context on top of it is the whole game. The winners of this cycle will not be the vendors who fought hardest to hold the client's data. They will be the ones who understood that holding it was never worth much, and that the work worth doing was one layer up all along.
Related: AI-Native vs. “AI-Powered” CRMs: The Architectural Gap Most Advisors Are Missing