Why we’re building AI infrastructure that connects all facets of our firm (wealth, tax, and planning) so our team can see the whole picture.
By John Renaldi | Chief AI Officer, Leelyn Smith
Your wealth advisor and your CPA are supposed to be looking at the same picture of your life. Sometimes they are not, because they’re sitting inside two different systems, built by two different companies, for two different jobs, often with a third firm managing the investments behind the scenes. So you end up being the thing that connects them, explaining the same decision twice to two people who both work for you and are supposed to be making your life easier.
A tax decision from three years ago should shape a portfolio move today. The reason money moved a certain way in 2022 is sitting in someone’s meeting notes instead of anywhere your advisor can actually see it.
This isn’t really a people problem. The tax history, the portfolio decisions, the meeting notes, the small human details, all of it fragmented across more than ten separate software platforms at our firm and a handful of people’s memory, none of it adding up to one simple, complete picture.
Built to not talk to each other
Many will immediately believe this is simply a relationship issue. Advisors and CPAs who just need to communicate better, firms that need tighter processes. I think that’s backwards. Many of these systems were not built to share context in the first place, so of course the people stuck using them end up doing the sharing by hand, one recap and one phone call at a time. You can’t process your way out of an architecture problem.
Take a simple scenario: an advisor sets up a new account today, which subsequently produces taxable earnings for the upcoming year. When tax season arrives and your filing packet lands in your inbox, the required 1099 for that account is missing because the tax platform was never informed it was created. Neither was the CPA. Suddenly, your team is back to manually cross-checking platforms, chasing down loose ends, and digging through old notes just to confirm what was already put in motion months ago.
It’s not only the numbers that get stuck. Someone on our team recently learned, almost in passing, that a client had a favorite tea she couldn’t find anywhere anymore, and quietly sent her a box of it. Small thing, but it’s the same problem: information that exists somewhere, known by somebody, but not built to travel to where it’s actually needed.
The usual fixes are a shared spreadsheet, a weekly sync meeting, a CRM note that only the person who wrote it ever reads again. In our experience, they rarely close the gap. A shared spreadsheet here, a weekly sync meeting there, a CRM note that only the person who wrote it ever reads again. None of it actually closes the gap. It just adds another place for information to get stuck, and more importantly, outdated and not maintained. One of the key benefits of this new type of infrastructure is that as information changes over time – a new family member, a new job, a cross-country move – the context gets updated and replaced in real time.
What gets lost in the meantime is time syncing and cross checking, and not a small amount of it. What gets lost in the meantime is time, and in our experience it adds up. Hours spent syncing and cross-checking what someone else on the team already knows are hours not spent on the work that makes a client’s life easier, or that surfaces new ideas and solutions to their wants and challenges.
Leelyn Smith has tried to operate as one integrated team since the beginning, with wealth, tax, and portfolio management working from the same picture, because we think that’s the only way to actually serve someone well. It’s also genuinely hard to build.
The Infrastructure We Are Building
As we design and build this new AI infrastructure at Leelyn Smith, we are starting with each individual client. Every client gets a knowledge base, or a living picture, built from their full history, decisions, and context that exists across multiple systems and relationships.
This is not a CRM with a better UI. A CRM stores what somebody typed into it. It can maybe infer a little, if you’re generous about it, but it can’t tell you why a client made a decision three years ago if nobody wrote that reason down anywhere the system could reach. We’re building the thing that actually captures that in real time so it carries forward instead of evaporating the next time someone changes roles or just forgets.
Imagine a wealth advisor’s AI copilot that remembers a client wants aggressive growth, but missed the conversation where that client asked to avoid fossil fuel holdings because of their foundation’s climate mandate. The recommendation it produces would look reasonable but would be missing critical context. That is the problem with working from a partial picture, and it is exactly what good AI infrastructure has to solve. None of this replaces judgment. AI helps our team see more of the picture, and the people on our team review what it produces and make the decisions.
Zooming out
A complete picture of each client is half of what we are building. The other half is less about memory and more about the numbers. Every system we use, from tax software to portfolio management to planning tools, has its own way of defining and calculating things like income, cost basis, or gains, and when those definitions drift even slightly, two people looking at the same client can come away with two different answers.
A semantic data layer fixes that. Think of it as a shared dictionary and calculator that sits across every system we use, so that “income” means exactly the same thing to a CPA as it does to a wealth advisor, and every number is pulled live from the source rather than copied from last quarter’s report. The client picture remembers why decisions were made; the data layer makes sure everyone is working from the same current numbers. Together, they are what let wealth, tax, and portfolio management actually work from one picture.
Managing portfolios in house, which we do, makes this matter even more. It is one more system, one more set of numbers, and one more place where the full picture could come apart if nothing kept the definitions consistent.
Take a question like, “How does today’s move in the bond market affect Client X’s overall wealth picture?” A good answer has to draw on their portfolio, their tax situation, and past conversations at the same time, and that is why this layer of information is so powerful.
Built right, not fast
I didn’t take this job to bolt an AI feature onto the firm. This is infrastructure that touches real client financial and personal data, and we take that very seriously. Security is the first question we ask, not the last one. What gets stored, who can access it, how data actually moves between systems, all of that gets decided before a single line of product gets built.
In practice, that starts with where the work happens. Our most sensitive tax work is processed by AI models we run ourselves, on a Leelyn Smith server that sits inside our internal network and is not directly exposed to the internet. When we do use an outside model provider, strict policies apply: the data is processed in US-based data centers, it is never used to train the provider’s models, and under zero data retention, neither your data nor our prompts are stored on their side. The only record stays on our own servers, where it is kept for compliance audits.
Just as important is control over how our own team uses these tools. Every system an employee reaches through our AI tools requires authentication, grants only specific permissions, and is audit logged, and every AI interaction is logged centrally for compliance and governance. We have also built an AI policy engine that lets us enforce the rules for how employees work with AI and with your data from one place. That way, the rules are built into the tools themselves rather than depending on each person to apply them. For example, if a client has not agreed to the use of AI in their tax preparation, the system blocks any employee from using AI with that client’s tax data, so the decision does not depend on anyone remembering it.
A small firm doesn’t get a smaller compliance bar. If anything, it gets a smaller margin for error, because there’s no massive team standing behind us to catch what engineering missed. We have to get it right ourselves, the first time.
Closing
We’re not building this so a system can know a client better than the people serving them. We are building it so those people are not asking a client to repeat themselves and so that our team can do the most complex, innovative work for each client’s situation.
That’s a whole other aspect of this I’m excited to see come to life. We don’t even yet know the types of new insights and ideas that our team will be able to run with after seeing all of the context and data in one place, at one time, running on real time updates. We really are building a more connected, multi-disciplinary way of working here at Leelyn Smith. This new infrastructure will make the integrated model we believe in even more powerful, deepening relationships and the types of proactive guidance our team can provide. The best ideas are the ones we cannot even predict yet, the ones that only show up once someone finally has the full picture in front of them.
The opinions voiced in this material are for general information only and are not intended to provide specific advice or recommendations for any individual.
This information is not intended to be a substitute for specific individualized tax or legal advice. We suggest that you discuss your specific situation with a qualified tax or legal advisor.