Mark Wickersham, Host 0:05

Hi, and welcome back to the Wealth Tech Podcast. I'm your host, Mark Workersham. In this episode, we talk about the importance of what it means to own your own data and how AI, data likes, and vibe coding are adding to the modern family office tech stack. We sit down with Tumai and Sinan from Coalation AI, and we discuss what the truly modern tech stack looks like for family offices and RAAs in 2026. In this episode, we talk about the impact AI is having on the traditional SaaS model, what it means to own your own data, vibe coding, best debris versus all in one, and how that's impacted by AI and why this may be the golden aid for family offices, as this technology revolution is just getting started. Let's get started. Would you guys mind giving a brief introduction of yourself?

Sinan Biren 1:08

First of all, thanks for the invitation. Very happy to be here. Uh Sinan here, CRO at Collation AI, um, a wealth tech out of New York. This is our second venture with uh my colleague Ton Mai. We've spent more time, I mean, I've spent more time with him than my family. That's a fact. As a matter of fact, I used to use his voice for calming my kids down when they were babies. You know, this white noise kind of thing. So that was basically Tom Mai for my kids. So uh yeah, yeah.

Tanmai Sharma 1:37

I don't know if that's a compliment, but thank you.

Sinan Biren 1:39

It is, it is, it is.

Tanmai Sharma 1:40

So I'm actually a banker pretending to be uh fintech guy, so I spent the bulk of my career, 20 years, on various uh trading floors, buy, sell, crossbow arbitrage, that kind of stuff. Um then about 12 years ago, uh uh got bored of banking, started a company called Canopy out of Singapore, which was a competitor to Adipar. Uh, but I wanted to then come back to the US and attack the biggest wealth market there is. So the management team of Canopy actually quit and we started Coalition just over two and a half years ago.

Mark Wickersham, Host 2:10

What is Coalition AI's ideal client profile?

Tanmai Sharma 2:13

Family offices, and we work a lot with um RIAs. Um our tech stack and our setup is exactly the same, but these two client segments use it very differently. Um so when we're talking about family uh offices, um we are dealing with the typical uh family office setup. So there is a family office, there are various family members, uh uh almost invariably there is partnership accounting, there is partial ownership, partial ownership changes the time. There are multiple generations in the family, uh different people are looking for different information. Uh the IC wants something else, the family office people want something else, um, and individual family members want something else from the information that they're looking at. So the ability to first um gather all this disparate data into one place, uh, and then be able to kind of uh configure different users to see the information that they need to see. So that is kind of the product, that's our profile.

Sinan Biren 3:09

Control HAI unique, Mar. So our current business model is based on the best of breed approach where we set up a centralized data warehouse or data lake, sounds kind of fancier, for customers, and we let the customers host their data uh by themselves. This is quite unique by itself because customers can not only access their data, but their own their data from day one. And into that data warehouse, we're bringing data from uh, sorry, actually, into that data warehouse, we bring in their entire data set from, for example, from third-party software vendors, banks, brokers, market data, private asset valuations, and things like this. Um, and then probably something Tanmai is gonna um deep uh deeper is that we make sure that the data sitting in the warehouse is accurate. Um, since next phase is to help the customer to make sense of their data and derive actionable insights on it. And we don't have high-quality data to report on, basically getting garbage uh data out. Um finally, and uh this is what we really would like to talk about today, is to set up a regulatory-friendly AI-based reporting platform. Uh, the fancy name is Vibecoding, like vibing, where uh the client can use AI for building custom applications, dashboards, client portals, and things like this. The AI basically is writing the code for you so that you can uh build these custom uh pieces, and this essentially becomes clients IP as well.

Mark Wickersham, Host 4:36

So we want to talk to whenever I get a chance to talk to a founder, especially a serial entrepreneur uh such as yourself with Canopy and now coalition AI. Talk to me a little bit about your founder's journey. What what was the impetus for creating Coalition AI?

Tanmai Sharma 4:57

Um, so coalition was born from three or four things. One is we wanted to attack the biggest wealth market. Second was um technology map had changed significantly, and we could now do things that we could not do in Canopy, and you could we could only dream about in in Canopy. And then there was this whole domain knowledge and uh you know learning from what you had done and what worked and what didn't work. So all these three things kind of came together into collision.

Mark Wickersham, Host 5:24

Yeah, that kind of uh a gen two, right? If I had a complete whiteboard to do it all over again, how would I do it?

Tanmai Sharma 5:31

Yep.

Mark Wickersham, Host 5:31

Um, it's certainly uh interesting. So what are some of the common problems that that people are coming to you guys with? And also, can you tell me if like the the problems and the expectations of the buyer have shifted over time?

Sinan Biren 5:50

So the the problems vary depending on the client profile. Um, so for example, if the client is a family office, the typical problems we see are hey, I need better investment reporting. I do have a general ledger system in place, but I'm doing the performance reporting on Excel spreadsheets. Um, I looked at the market for an all-in-one performance reporting software, but A, they're too expensive, B, I don't want to run two copies of my data, two separate applications, and all that. Um, another common problem we see across the family offices is that the data is sitting in different software applications, and you know, a bunch of folks is trying to uh tie up the data. Um, so you know, this tends to be a rather expensive fix. Um and the third and the most recent topic that we see across family offices is that hey, everyone is talking about AI. We are not. Are we missing out on this, right? So uh that's what we definitely see across the family offices. Whereas with the wealth managers like RAAs and EAMs, as we call them in Europe, Asia, the common problems tend to be more on the operation side of things. So, for example, how can I be more efficient with my current tech stack, with my staff that I have in place? Where can I cut uh costs? How can I improve my uh performance? How can I be more lean, faster, smarter? You know, into this you can throw in KPIs like sales velocity, advisor productivity, client retention, um, and and things like this, right? Um, or you know, oftentimes the wealth managers tell us that, hey, I have um, you know, a lot of data sitting in disparate uh data in disparate silos. I wish I could have all that data sitting in somewhere centrally so that I can get most out of it. And we live in the AI times, so how come I can't use AI to get most out of the data that I have? Um, and then of course you have the other side of the spectrum, which is hey, we have a data lake in place, all our data lives in there, but the quality of the data is so bad that when we bought uh you know a bolted large language model on top, we get garbage data out, right? So and uh you know, I have my data, it is clean, but reporting is too expensive, right? Because the system I'm using is anticated and it's clunky. Um, and I'm personally used to having custom things like custom movies, uh, you know, custom uh music playlists. So I can't have my custom reports, isn't that the whole promise of AI, right? So vibe code anything I want. Like for example, I see a report somewhere and or a calculation. I think it's pretty cool. So I just want to use a plain English chatbot to describe what um the AI should do, and it should go and and just replicate that report for me for a few pennies in a matter of uh seconds, right? So those are the things that we see across the wealth managers and family offices, the common things we see.

Mark Wickersham, Host 8:47

Yeah, I think that's still that age-old problem that um two things that that kind of stand out that that have been kind of universal over time is one is that family offices wealth managers have a lot of data that's siloed and and it's difficult for them to pull together in a in a comprehensive way. And this it has it has gotten better over time. I mean, there's pillars now versus it used to be back in the day that you had a on the investment side, you had an alternative investment reporting system and a marketable one, and then you manually pull that together, and at least those systems seem to handle both the the wide range of asset classes, but still they don't seem to talk to their general ledger. And then the the age-old uh problem that that will never go away is you know, garbage in, garbage out. And and AI is not is not maybe AI can help clean up your garbage, but if you still, if you're putting garbage in, AI is going to give you a lot of garbage on the way, right? 100%.

Tanmai Sharma 9:41

Maybe if I can just add to that, right? So if you zoom out a little bit uh and you think about it philosophically, uh the client requirements have not changed as such, right? The family office wants to know basically three things, right? Uh what am I worth? Uh how am I doing, and do I need to change anything? Right? These are the three questions you always wanted to know. Last hundred years, nothing has changed. Um now, uh but the problem that the family office right now has is I have a expensive tech stack. So I have a journal ledger, I might have a portfolio system, I might have other uh systems. And to run this expensive tech stack, I need specialists who are also expensive people. So I have two or three tech stacks, I have this expensive software running, and I'm spending easily half a million, seven fifty K purely on uh, you know, the software and people who run the software, and I'm still not getting the report I want. Right? So philosophically, that's the problem. And what has changed now is you get a database, get all your data into one place, have AI solve uh bulk of your reporting requirements. So you can now achieve that requirement at a fraction of the cost, at a fraction of the time. And more importantly, the requirements evolve with time, right? So everybody who's doing the analysis, their requirements keep changing. And the AI, because it's just text-driven, saying, Can you now make me this dashboard? Can you make me this workflow, blah, blah, blah, evolves with you and evolves rapidly. And you're not dependent on a big software vendor making their next release, which happens to also cover your requirement. So that is the big cataclysmic change that we see.

Mark Wickersham, Host 11:22

What talk to me about vibe coding? What is it? Obviously, you know, while we've mentioned that it's regulatory friendly, what does that mean? Um, and why should family officers care about vibe coding?

Tanmai Sharma 11:36

So, vibe coding is simply an AI interface where you can tell the AI model, I want a dashboard which does XYZ123, and the AI makes that dashboard for you on top of your data. So it just it's literally replacing um uh replacing the person who was making your Excel spreadsheets with a much more sophisticated person who does a better job, uh more sophisticated dashboards, deeper analysis, um, etc. etc. So it's not a new requirement, really. It's just that there is a new tool in the market which meets your requirement way faster and at a fraction of the cost of anything else. That's basically what it is. So if I was to summarize this in a single sentence, um you use Chat GPT where you're giving it text inputs and it's giving you text output back. Then you started using things like Midjourney, where you give it a text input, please draw me a picture of a cat drinking a banana milkshake, and it'll give you the image. And now it the next step is a logical next step, is you say, please give me a dashboard which shows me where is my exposure geographically in my private equity portfolio. That's a question. And the AI will go in and it'll draw the dashboard, bring it to you, and you can say, uh, I want this change, I want that change, whatever. So vibe coding is simply a plain English conversation you're having with an AI agent who is very smart and is able to make the dashboard that you want, and is also able to take your feedback and keep improving it as time goes on.

Mark Wickersham, Host 13:11

And what about the regulatory requirement or the regulatory?

Tanmai Sharma 13:15

That's a very, very key and interesting question. When we are dealing with any data, any family, any wealth manager's data, there's invariably PII involved, right? We have the names of entities, we have your positions, we have your sometimes even social security numbers, but definitely addresses, names of individuals. So there's a lot of very sensitive data that should not get leaked. And uh because vibe coding platforms have an AI engine at the back of it, uh, you really don't want to expose your real data to that AI engine. So effectively, the way we do it is we create a sandbox which has the same data structure but completely dummy data. The wipe coding dashboards are made actually on the sandbox. Then the code that is generated is vetted and then brought over to the secure infrastructure to run only as a code. So we actually never let AI directly touch clients' PII uh purely as a matter of abundant uh caution. Um and that's why it's regulatory, uh regulatorily compliant. When we say regulatory compliant, what we are trying to say is you know exactly where your data is at all points in time, and nothing, no PII touches an AI uh model unless you, of course, explicitly uh wanted to.

Mark Wickersham, Host 14:35

What about owning your own data? I mean, I think some of that goes with the best of breed type of discussion, but why is owning your own data important and how do the mechanics of that work?

Sinan Biren 14:51

I mean, essentially you don't want to have that single vendor lock-in scenario, and you want to have a feature-proof uh tech stack or a platform. Um if you think about the transition as Tamay was describing, right? So moving away from Excel spreadsheets to online software applications, and now being able to actually build these by yourself by the help of AI, essentially you want to have a feature-proof um tech stack where you can actually swap any solution that you want. So, for example, uh, you don't like the way how your CRM vendor is treating you, or you don't like the way how much you're paying for a portfolio system, or you don't like the way how um the general ledger system is clunky. So if you want to be in a position where you're able to negotiate better terms or sunset that system and replace with another one, you need to have a platform that supports that, right? And especially with the AI. I mean, we can give from our own experience. We started the vibe coding journey with the most hyped uh vendor out there, which was called Lovable back then. I think it took it two weeks, Tommy, that we then started exploring Bolt.new, that was a new tiny sh uh tiny um shiny toy in the market. And then we tried out cloud code. These things are so volatile. Um, and the advantage of having an open platform, future-proof platform is that you can replace these things like that, and that's exactly what a family office should should aim for because the way how things are moving, one thing is never going to change the fact that you need to have all your data in one place and you need to make sure that the data is intact and incorrect. Then however you want to visualize that data or get most out of it, you can bolt whatever LLM vibe coding tool you want. The thing is that you should not be locked into a single vendor. That is the message, basically. And that's why we think it's quite crucial.

Mark Wickersham, Host 16:46

Yeah, I think that makes sense. I think regardless of whether you're going with the all-in-one approach or the best of breed, um, you know, vendor risk is is you have even more vendor risk, obviously, with it with an all-in-one approach. Um, one of the benefits of a best of breed approach is that it kind of minimizes uh your vendor risk to a certain extent, but the ability to kind of swap out the nodes on your stack there are important because vendors change, like their ownership structure changes, uh their situation changes, your situation as a as a firm may change, um, and and that vendor may no longer be suitable. Um and to be locked in with all that historical data makes makes that a harder proposition, right?

Sinan Biren 17:30

And you just want to have this this you know future-proof infrastructure, and that's essentially what we're providing you that future-proof infrastructure.

Mark Wickersham, Host 17:37

Talk to me a little bit about the relationship between AI and data. Obviously, I think we've touched on this a little bit about why having clean data is so important.

Tanmai Sharma 17:46

It's it's a case of garbage in, garbage out, right? So um AI is a very smart person, um, but um, you know, uh unless you give it the right data, unless you give it the wrong data, unless you give it the right data, unless you give it the complete data, you will run the risk of results being wrong. Um, so in fact, it's actually goes a little bit beyond just correct and complete data. So you have to give it context, you have to give it perspective, you have to give it solved examples, then you have to give it the data, you have to build in guardrails around uh you know the AI doing something really random and stupid. So so it there is a whole setup that goes into it, uh, not just clean and complete data. But if you don't have clean and complete data, you're completely wasting your time. So uh it's as simple as that.

Mark Wickersham, Host 18:34

Modernizing the family office. Uh obviously, AI is a key component of that. Will AI force uh family offices to modernize? I mean, there's a number of factors at influence here. Obviously, there's a a huge wealth transfer that's going on. Um, so you have a next gen that's more um technology, gotta you know, cloud native, uh, at the very least. What are the what do you see happening with family offices? Is the productivity and the enhancements so great with AI that it's gonna force them to modernize? Are you still gonna see you know this kind of bell curve where family offices still are on the other hand?

Tanmai Sharma 19:16

Uh, which is that just because you're rich doesn't mean you don't care about money. Uh and if you go into a family office and say, Mr. Family Office, your current cost of tech plus people running your tech is 750k. And if you now change to our setup, you will be inside 100k and you get flexible reporting and you know, blah, blah, blah. Um, I I I it doesn't matter how much of a billionaire you are, right? If if we walked into your office and said, AI can save you half a million six hundredk per annum, um it is simple economics. Economics will blow it away. Uh, and of course, you know, there's flexibility and you can have a plain English conversation with it, um, etc. etc. Um, do you agree with that?

Sinan Biren 20:05

Yeah, I'm just yeah, I'm I'm just I 100% agree. I'm just gonna add to that. Uh what we also see, Mark, is that the the barrier to entry for forming a family office will eventually go down. Um actually it has started, right? So historically uh you needed around 100 to 50 mil, give or take, in assets to justify the cost of running a traditional family office. Um with the advancement of AI, we see the you know reduction in headcount, um, as Tamay was saying, you know, improvements in automation. So we predict, and we're seeing this already in the market, that a lean family office uh would become viable at around $25-50 million dollars uh ass under management. Um there's also, by the way, today a concept called virtual family offices. You can go and Google or search on LLM, and these things start around five to ten million dollars uh AUM. So our prediction is that the line between a family office and an advanced AI supported advisory practice will blur uh over time. And and this is this is this is it. So 100% there's a correlation there.

Mark Wickersham, Host 21:16

Yeah, I think um single family offices are always trying to do more with less, they run lean. Yeah, and this just helps them run leaner or or do more with less, right? Um and then the multi-family offices are looking to scale, right? They're adding more clients all the time. Scalability is important, um and and this helps you know, technology is is is revolutionizing the way in which FERPs can do that.

Tanmai Sharma 21:42

But but there is um so unless you are uh if you are a multifamily office and you're not set up to offer customization at scale, um, again, you won't have you'll have unhappy clients. So this is one problem that AI solves radically fast, which is that you will be able to Set up custom dashboards for each client, and the cost of making a new custom dashboard is ten dollars, fifteen dollars. Uh, and and that's where uh you know the interesting things uh start happening.

Mark Wickersham, Host 22:10

Yeah, there's personalization at scale, right? I mean, that's that's absolutely really why you go to a family office or why you create a family office is ultimate in customization based on your family's needs, and certainly the expectation doesn't change if you're uh an ultra-high net worth family and you're going to a multifamily office, you're gonna expect a level of personalization. Um let's talk about AI and where we're in the um AI evolution. Obviously, I think vibe coding is this key kind of component of it, large language models, the ability to have a conversation with your with your data, a genetic AI is is now on the forefront. Where are we in the AI evolution? And what's a preview of what firms should be starting to think about and see uh in the very near future?

Tanmai Sharma 23:01

We are really just starting, in my opinion. Um, and uh we are at a cataclysmic uh increase in productivity that comes from AI. Um so it has it has a bunch of implications. First, we just discussed about the cost of running the setup, it goes down drastically. The flexibility, flexibility goes up. Uh, and we are still right now talking about doing reporting, right? We're not talking about decision making and uh insights and analytics and um actually you know even executing uh based on that. So all that is going to come uh with time. Uh but it's fairly clear to us that this is going to be cataclysmic.

Sinan Biren 23:48

Um just to add to Tamai's point, um I don't think we we're not far off from having um digital chief of staffs uh in family offices, like especially AI-enabled family offices.

Mark Wickersham, Host 24:02

So uh an genetic chief of staff is uh Exactly.

Sinan Biren 24:06

So within every family office, um large or small, doesn't matter, will use an A agent for you know monitoring portfolios, banking activity, private investments in real time, and things like this. So it will essentially, the agent will essentially behave like a junior analyst uh plus an executive assistant plus an ops associate combined. And uh this you know single AI agent will probably reduce around one to three full-time roles, you know. That's typically what we predict. And anything that has already started, uh, but I think it will become more mainstream, is uh real-time consolidated reporting, replacing the quarterly, if not annual, reporting. So, you know, the gen three, gen four says, Hey, I want to see where I'm standing right now, and just you know, go to your app, which has been built by AI again, as Tamai was saying, you know, just by plain English describing what needs to be done. Um, you got that report done in in a few minutes, and then you deliver it to the Gen 2, and the Gen 2 just clicks on a button and sees in real time the reporting. So um I think these are the things that that is coming our way. But the agentic AI pieces is very, very exciting, of course, uh, because that again reduces the barrier to have your own family office, you know.

Mark Wickersham, Host 25:21

We've obviously talked a lot about the future of family offices already, the the democratization of it, the barriers lowering how technology will change the landscape. Um, I'd like to hear from both of you uh a prediction. What does the future hold for the family office industry?

Tanmai Sharma 25:42

There is a very famous quote from Yogi Berra about how future is hard to predict. Um but um if you if you see things and the trends that are already obvious, right? So one is about productivity, which is the essentially means uh smaller uh number of people doing uh more things and more uh things rapidly. Uh what uh what I expect or suspect will happen uh is is a cooperation or coordination uh between family office units. Uh it already happens. They already share uh feedback and they talk to each other a lot and they co-invest and they are you know talking to each other, but that will start taking um its own its own life. Um so uh let's let me give you an example, right? Uh there are a few thousand family offices in uh in the US, just in the US. Uh, and let's say all of them are buying private equity, uh, which is investing in real estate development, let's say, for example. AI will be able to essentially allow them to coordinate with each other in bulk uh and essentially squeeze the margins of the providers. So I would expect uh a lot of the friction out of the market to go away uh because coordinating with each other and uh making uh kind of macro investment decisions will become um easier. So let me give you another variation of the same uh requirement. There are a few thousand uh family offices uh in the US, and let's say using AI they are in sync and they have a very good sense of what each other is doing. They become a forced reckon with because they suddenly you are you know where the next move from a trillion dollar size uh investor or investor group is going and is is coming. Uh and so these things will become very important. So I see a shift of power and a shift of uh being able to uh influence the market, move away from the providers and towards the investors, uh, towards the family offices.

Sinan Biren 28:08

Yeah, and then just just just to add to Tamai's point, uh, we really see and and predict uh a massive expansion of the family office universe. Um more liquidity events happening for founders, tech, AI, crypto, myatech, you name it, and uh falling operation, operational costs, due to automation, um, and lowering the barrier, the entry point for establishing a family office, all this will just funneling the massive expansion in the family office uh space, is what we say, which is a great thing for us for vendors like us.

Mark Wickersham, Host 28:42

It is great. I mean, you know, it's the beginning of the golden age for for family offices. Uh certainly could be a proliferation of them. I think uh that that saying that the the futures here is just not evenly distributed, it's it uh I think it applies pretty well that there are lots of cutting-edge capabilities that are that are coming uh the way to the family offices, and they're they're they're here today, you know. AI is very much operational, it's not uh future state thing to be thinking about it it's something to be acting on today. I think vibe coding is another good example of that. Um, I like to end the podcast on a personal note. Uh Dame, you came from uh Singapore, you're now in New York City. Um tell me, give me a little comparison, compare and contrast the the the two cities.

Tanmai Sharma 29:35

How's the it's uh that's a very interesting question. Um the two cities could not be more different, and the two cities could not be the more similar. Uh both statements are actually uh true. Um so Singapore is a very uh sort of nice, uh in a sense, uh very clean, it's squeaky clean, it's it's it's set like that, and you know, uh the infrastructure works and it's designed like that. And New York is completely controlled, but sometimes out of control chaos, right? So uh they're very different in uh in in that sense. Uh but if um and um Singaporeans definitely tend to be more polite than New Yorkers, New Yorkers like not to be polite, uh especially when they don't want to. Uh but uh if you look at it slightly uh differently, then both cities are entrepreneurial cities. Um and there is this streak of entrepreneurship, you know, I can do it, I can get this done, let me get this done, let me see this, let me fix this. And that's true in both places. So it's it's a very interesting question. They're very similar and very different at the same time.

Mark Wickersham, Host 30:45

Yeah, that's what I was saying. You can make it here, you can make it anywhere, kind of right. Um New York has a certain level of chaos to it that I I I I love. I I obviously I love going to New York. I I also love leaving New York too, so it's it can be a bit much, but it it is great. And I'm looking forward to going to Singapore one of these days.

Sinan Biren 31:14

I'm a big-time cyclist, so I love these urban paths and these scenic climbs it has to offer, like um this path from uh a local place called Glott. So from Glott to Bülah, the gravel pass, and then there's another pass called the Albis Pass Loop. So there are these fantastic little villages where you have to climb up and then have this hill descent, and it's just just magnificent. But apart from Zurich, of course, uh one place that many people don't know, uh, so you know, um, except for the interlacons and and Lutzens and things like this, there's a small spot called Kalmas Day, which is a small lake. I think it translates to Lake Siesta, you know, like this afternoon nap that is taking on a hot day. Although I've never seen a hot day in Switzerland. Uh, but you know, uh that's also a favorite family spot for us. So try it out. Kalmas Day.

Mark Wickersham, Host 32:12

When you say climbing these passes, like how what kind of elevation are are you are you gaining on the on these roads?

Sinan Biren 32:18

So so at max in Zurich, I would go 600 to 800 meters. That's how we have uh meters, right? Uh but one day I want to do the big mountains, right? The 2000s to three thousands, but I'm not at that level yet. Working on it though.

Mark Wickersham, Host 32:35

Like you know, some of these passes like Bon 2 or some of these other ones you see. Oh, yeah, yeah, yeah.

Sinan Biren 32:40

Yeah, yeah, yeah, yeah, yeah. Yeah.

Mark Wickersham, Host 32:41

Maybe maybe an e-bike uh might be there.

Sinan Biren 32:44

That's that's cheating, right?

Mark Wickersham, Host 32:46

Good stuff. Well, gentlemen, I appreciate your time. This has been a fascinating conversation. Thanks for being on the Wealth Tech Podcast.

Tanmai Sharma 32:54

Thank you. Thank you.

Mark Wickersham, Host 32:55

Thanks, Mark. Cheers. Thank you for listening to this episode of the Wealth Tech Podcast, brought to you by Seta AI. Seta is on a mission to modernize the family office. I really appreciate you listening, and don't forget to subscribe so you don't miss future episodes. Talk to you soon.