[Emerging Tech] The Future of Financial Advice Automation With Eden Ovadia
- Quik! News Team
- 17 hours ago
- 30 min read
Eden Ovadia is the Co-Founder and CEO of FINNY, an AI-powered platform helping financial advisors identify prospects, automate outreach, and grow their client base more efficiently. She leads the company’s mission to solve the organic growth challenge in wealth management through data, machine learning, and automation. Eden has a background in software engineering and machine learning, and previously worked at Boston Consulting Group, focusing on financial institutions and private equity.
Here’s a glimpse of what you’ll learn:
[2:22] Eden Ovadia discusses how FINNY helps financial advisors find and serve clients using AI and data
[3:53] How FINNY tracks data across North America to identify ideal prospects
[5:56] Using real-time data to deepen existing client relationships
[7:00] Why proprietary data and model training create a competitive advantage
[8:49] How to solve organic growth challenges in wealth management
[11:57] Eden’s approach to building internal AI productivity tools
[18:47] Semi-agentic AI and compliance constraints in financial services
[21:08] Why the future belongs to specialists, not generalists
In this episode…
Financial advisors often struggle to grow their businesses efficiently, relying on outdated methods like referrals, manual prospecting, and fragmented data systems. This makes it difficult to consistently find high-quality clients and build meaningful relationships at scale. With increasing competition and rising expectations, how can advisors modernize their approach to growth and client engagement?
Eden Ovadia, an expert in AI, data, and financial services, explains that solving this challenge starts with leveraging large-scale data and automation to identify high-potential prospects. She highlights the importance of combining public and proprietary data sources, continuously updating client insights, and using AI to surface timely opportunities. Eden also emphasizes automating repetitive tasks like outreach and follow-ups, while preserving human connection for relationship-building. Additionally, she encourages teams to adopt AI tools internally, break complex workflows into smaller tasks, and rethink growth strategies with a data-driven mindset.
In this episode of The Customer Wins, Richard Walker interviews Eden Ovadia, Co-Founder and CEO of FINNY, about using AI to transform client acquisition and growth. Eden also discusses agentic workflows, the future of user interfaces, and how AI enables leaner teams.
Resources Mentioned in this episode
Quotable Moments:
“So my company, FINNY, helps financial advisors better find and serve their clients.”
“We have a database of everyone in North America. So Rich, for example, you’re in it.”
“You could probably build yourself a version of FINNY that gets you 20 to 30% of the way there.”
“What we’re building is so in the realm of a new category that it actually is quite hard.”
“It’s evolving more and more into peers, I would say, within FINNY specifically.”
Action Steps:
Leverage data to identify high-potential prospects: Using large-scale data helps pinpoint the right clients at the right time, improving efficiency and increasing the likelihood of meaningful engagement.
Automate repetitive outreach tasks: Automating emails, follow-ups, and prospecting saves significant time for advisors, allowing them to focus on building real relationships instead of manual work.
Continuously update client data and insights: Keeping client information current reveals new opportunities to connect and provide value, leading to stronger relationships and better long-term outcomes.
Break complex workflows into smaller AI-driven tasks: Simplifying tasks makes AI outputs more accurate and reliable, while also reducing costs and improving consistency in execution.
Combine AI efficiency with human connection: AI can handle scale, but human interaction builds trust and loyalty, ensuring better client experiences and sustainable growth.
Sponsor for this episode...
This is brought to you by Quik!
At Quik!, we provide forms automation and management solutions for companies seeking to maximize their potential productivity.
Using our FormXtract API, you can submit your completed forms and get clean, context-rich data that is 99.9% accurate.
Our vision is to become the leading forms automation company by making paperwork the easiest part of every transaction.
Meanwhile, our mission is to help the top firms in the financial industry raise their bottom line by streamlining the customer experience with automated, convenient solutions.
Go to www.quickforms.com to learn more, or contact us with questions at support@quikforms.com.
Powered by Rise25 Podcast Production Company
Episode Transcript:
Intro: 00:02
Welcome to The Customer Wins podcast, where business leaders discuss their secrets and techniques for helping their customers succeed and in turn, grow their business.
Richard Walker: 00:16
Hi, I'm Rich Walker, the host of The Customer Wins, where I talk to business leaders about how they help their customers win and how their focus on customer experience leads to growth. Some of my past guests have included Dr. Jon Randall of XFA Coaching, Andy Schwartz of OnePoint BFG Wealth Partners and Katie DeFeo of SunWest Federal Credit Union. Today is a special episode in my series on new and emerging solutions, and today's guest is Eden Ovadia, the CEO and co-founder of FINNY. And today's episode is brought to you by Quik!, the leader in enterprise forms processing. When your business relies upon processing forms, don't waste your team's valuable time manually reviewing the forms.
Instead, get Quik! using Quik!. You'll be able to generate completed forms and get back clean, context-rich data that reduces manual reviews to only one out of 1000 submissions. Visit quickforms.com to get started. All right. I'm excited to talk to today's guest.
Eden launched alongside her co-founders in May of 24 to tackle the organic growth challenge facing Aria's. FINNY's AI technology automates lead identification, prioritizes high-potential prospects, and streamlines how advisors connect with clients. Under Eden's leadership, FINNY has seen early success, earning acceptance into the Y Combinator startup accelerator program and raising 4.3 million in a seed round in December of 24. She brings a deep background in AI, software engineering, and entrepreneurship to her work, combining technical expertise with a strategic lens. Before co-founding FINNY, Eden was an associate at Boston Consulting Group, where she primarily worked in tech, financial institutions, and private equity practices.
Eden, welcome to The Customer Wins.
Eden Ovadia: 02:01
Thank you. I am super excited to be doing this.
Richard Walker: 02:05
Oh, I'm happy you're here. For those who haven't heard my podcast before, I love to talk to business leaders about what they're doing to help their customers win, how they build and deliver a great customer experience, and the challenges to growing their own company. So, Eden, let's understand your business a lot better. How does your company really help people?
Eden Ovadia: 02:22
So my company, FINNY, helps financial advisors better find and serve their clients. We do this with data AI and automations. And the idea is to have better data on automations. Financial advisors can use that to match with potential clients in a more efficient way. And then once they do have their clients use data to serve them better.
And so our altruistic mission has not only been to create better financial advisors, but to deliver better outcomes for the average consumer in America. Who needs financial advice? That's around 40 million households that might need advice.
Richard Walker: 03:00
I love it. So if I heard you right, there's kind of a two pronged approach here. How do you find and attract these clients? Then? How do you serve them with the data?
So does this mean lead generation? Does this mean building relationships? I mean, because look, I was an advisor over 20 years ago and it was always about dinners and meetings. And in person there was none of the social stuff. So how do you do it?
Eden Ovadia: 03:22
Yeah, it's all of the above. If, if anything, with the rise of all this new technology and AI, the in-person connection you're alluding to is actually so much more important to build relationships. And so the idea is with our software, you can actually automate away all the manual, repetitive nature of what you have to do to actually get in front of the right prospects with the face time that you need.
Richard Walker: 03:44
Okay. So does that mean like just scheduling meetings, or does it mean finding people in the, the, the, the needle in the haystack in a group of people you belong to or something?
Eden Ovadia: 03:53
So it helps advisors. So super tactically, what we do is we have a database of everyone in North America. So rich, for example. You're in it. I'm in it.
Everyone that you know is in it. I know a little creepy, and we're tracking things from what you're looking up online based off your cookie data to your money in motion events, to your employee employment history and educational history. So we have all this massive amount of unstructured, unstructured data about everyone in North America, and we're tracking it every single day. And so when a financial advisor comes to us, they can say, hey, Eden, I want to find people who went to my high school who are now business owners who are looking up exit planning keywords online, because I'd love to reach out to them and help them think through the exit of their business, for example. And so Finn can actually be that tool or that platform that helps them identify people that can actually help reach out to them at the right time, craft the right message and automate and sequence all that kind of like autonomously on behalf of the advisor.
Richard Walker: 04:50
Oh my gosh. So right. I mean, advisors have their, their niches, they have their ideal customer profile, hopefully. And they're running that with you. So what is the success rate?
I mean, if I said I just want to talk to people who are willing to sell their business or looking to sell, don't they all have advisors already know?
Eden Ovadia: 05:08
You'd be surprised. Actually, I think the Wall Street Journal article last week about this like massive resurgence of this, the ultra high net worth class. And there's something like over 400,000 households that have over $30 million in assets. And most of these are kind of like the unknown millionaires next door. They're not, you know, in Manhattan or Palo Alto, and they typically don't have access to the financial service or level of financial services they need that you would expect them to have.
And so that's the idea behind kind of like the matching and optimization between the financial advisors that exist today and people who need financial advice.
Richard Walker: 05:45
Wow. Okay. And the second part, I mean, this is actually once you've won the customer, I guess you're using this data to help them do more. Is it like expanding their business or what?
Eden Ovadia: 05:56
Yeah, one of the so since we launched about two years ago, one of the craziest learnings is that oftentimes FINNY knows more and has more accurate data about your clients than you do, because you might have updated your CRM once, 20 years ago when they became a client, but then haven't refreshed it over time. And so the idea is we can constantly refresh the information and help surface opportunities to deepen the relationship with your client. And so maybe they're looking up for one K planning online. And this is a great time to reach out and kind of answer their questions that they're going online to answer. Maybe they got promoted or they bought a house or are going through all these money in motion events.
And so it's uncovering opportunities to reach out to your clients too.
Richard Walker: 06:36
Okay. I don't know if this will be a hard question, but one of the things I've learned in talking to people over the last three years is that we all have access to the same AI tools. I can get ChatGPT just like you can, right? So the thing I've figured out that makes us different is our data. That's right.
You just said you have all this data. So is this your data you sourced or is this everybody could get access to this data and then build what you've built?
Eden Ovadia: 07:00
Yeah, such a good question. Probably one of my favorite questions. There is a massive amount of public, publicly available data that the average person can go online and aggregate. The secret and the really hard part of what we do is the data aggregation and reconciliation, as well as having a data science team actually go out and source proprietary data sets. And so not only are we looking at all the information that's available publicly online, we also have licensing agreements with over a dozen different vendors that specifically bring data points that would help financial advisors.
And so you could think of things like donation records, 503 C records, intense signals based off people's cookie data. We have agreements with web publishers, thousands of web publishers, to give us back their cookie data, essentially. And then we can de-identify that. And so you can probably build yourself a version of FINNY that gets you 20 to 30% of the way there, but it's all that extra data that's really, really hard to get that we're quite good at getting. And then when you think about like, what data really makes us different, it's the fact that we've been around for over two years now and have literally hundreds of thousands of interactions that have taken place on our platform that we're able to train our model on in an anonymized way.
And so our models actually are getting better and better at drafting content, figuring out who is going to have a high likelihood of converting, and matching those, the advisors to the right prospect. That's the data. That's the biggest difference.
Richard Walker: 08:28
Okay, so now I'm super curious. I don't always go into origin stories with people because you tell your origin story all the time, I'm sure, but I'm just super curious. Now, the data that you're buying must be very expensive when you put it all together. You had to learn what data you needed. So that took time.
Where did you start? Like how did you guys come up with this idea of like, hey, let's get a bunch of data and go build this thing?
Eden Ovadia: 08:49
Yeah. So my background is in machine learning and software engineering. So it is somewhat relevant. But then after college, I went to work in private equity, as you mentioned before, at Boston Consulting Group. And what struck me and what I got the chance to work with was the wealth management M&A division.
And so was helping a lot of these wealth management firms and aggregators on the inorganic growth side of things. And so everything from advisor recruiting to buying up practices, to rolling them up and aggregating them. And it became so clear that there was a huge gap in the industry from an organic growth side of things. And most firms just couldn't figure out how to actually grow organically through client acquisition. And so the more cost inefficient way of growth was M&A and buying other books of businesses.
And so it was clear to us very, very early that organic growth and helping advisors with like growing their businesses was really important. And then when we actually spoke to advisors and asked them, like, how are you growing today? Like what's working? There was just such a lack of good technology and good tools available to them. Most advisors I was talking to were either sitting back waiting for referrals, or they were going out and trying to aggregate data on their own using like PitchBook plus Zoominfo plus LinkedIn.
And it was taking them 60 hours of business development to convert one new client. So just like massively inefficient. And so the secret was not only bringing them the data, but bringing them the automations on top of the data that they can actually set and forget in the background.
Richard Walker: 10:22
Wow. Does that make FINNY unbelievable? Is it hard to sell? FINNY sometimes, because people like you can't do that. It's not possible.
We've tried that type of stuff.
Eden Ovadia: 10:32
What we're building is so in the realm of a new category that it actually is quite hard in a 15 minute demo call or pitch to kind of comprehensively explain everything we do in detail. So yeah, it's spot on. It's just there's really no one else doing what we're doing.
Richard Walker: 10:48
Yeah. How do you overcome that?
Eden Ovadia: 10:51
Have really good sales team members that are really good at explaining what to do in simple ways.
Richard Walker: 10:56
Do you actually, like, give people free trials and give them free leads and things like that? I mean, where do you build the trust?
Eden Ovadia: 11:03
We try to have a lot of content online. So we have a YouTube channel that has tons and tons of platform demos and content. We write a lot of articles and thought leadership. My team is always on the road meeting people in person and doing demos live at conferences. And then beyond that, during the actual demo, when advisors come to us, we share our screen and walk them through the platform and let them kind of drive wherever they find it to be most interesting, because the way we build the platform is there really is something for everyone.
And so we kind of let our clients drive and tell us what part of the demo is most interesting to them.
Richard Walker: 11:40
Yeah, that makes sense. So how do you spend your time? I mean, me, I joke about golf, some people garden. My spare time is all building out AI stuff because I'm so fascinated with AI. And you have this AI background and you're the CEO.
So where do you like to spend your time right now?
Eden Ovadia: 11:57
Probably similar to you. I've been obsessing with building my own internal AI productivity tools. And so when I'm not, you know, working specifically on FINNY, I'm just trying to automate as much of the rest of my job or my life as possible.
Richard Walker: 12:13
Yeah. So do you get into software development or are you a cloud code coder or something like that?
Eden Ovadia: 12:19
I, I have to admit, graduated from college, I don't know how many years ago and I haven't actually coded since I left college because I went right into private equity until last month. And so last month was the first time in years that I actually opened up cloud code and started writing my own code and launching apps. It was really cool. It was quite daunting. As someone I have that technical background, but I haven't actually developed or written a line of code in years.
So it's a bit scary to fire it up, but cloud code makes it so easy to get started. It really is. You know, even if I didn't have a technical background, I would have been able to figure it out. It's really super accessible.
Richard Walker: 13:01
So I have a question in here, but I want to tell you my experience because I haven't coded for a decade. And I mean, I originally built it quickly by myself by hand, etc., but I haven't done coding for a really long time. And last summer I said, I've got to understand this coding capability. What is it like? And I literally said to Claude, please explain this to me like I'm a kindergartener.
I don't even know what to do. And he said, oh, go get VS code. I'm like, what is that? And then I said, tell me the file menu. Like, what do I click on to start something new?
Eden Ovadia: 13:32
I have to admit, I even was taking screenshots of the errors I was getting and I was like, help me fix this. Like literally, what do I write? Like, what do I do? Like down to the exact, like, basically it's crazy because when I came out, it was us giving the AI orders. And now it's kind of like giving us orders.
Like, especially when we're using cloud code, I feel like I'm like, what do I do next? Like I'm asking the AI to guide me, to prompt me. So it's such a reversal of roles.
Richard Walker: 13:58
Yeah, it is. And it's so educational. If you ask it to be educational and help you understand things. And that's how I've learned. So here's my question.
I feel that there's a couple of perspectives on what AI is to the person, like what its role is, how do you treat AI? And I can give you the categories. Do you think of it as a tool set or do you think of it as like a peer, a human team member?
Eden Ovadia: 14:23
It's evolving more and more into peers, I would say. And within FINNY specifically, we're getting away with keeping a significantly leaner team than we would have because everyone on my team is literally building agents to do their job for them. So instead of having to hire a team of five business operations, I can hire one person who can just build internal apps and tooling. It's pretty powerful in that sense. It's, it's, it's actually it hasn't replaced headcount, but it's replaced the need to increase headcount.
For sure.
Richard Walker: 14:54
Yeah, yeah, that's how I look at it too, is that it is definitely accelerating and being a force multiplier and slowing down the need to hire. And frankly, that's what technology's been doing all along. One of the earliest successes I had was Quik!. We showed it to the owner of a firm, an RIA firm, and he's like, if I get this, I won't hire the two people I was about to hire because I won't need to anymore. It's great, but I want to share this with you.
I think of human interaction like it's obvious you're a human. I'm a human because we see each other. We just know it. And as humans, we will treat each other as humans without thinking about it. And we will do things like, hey, even if we're working together, what's the best way to communicate?
Do you prefer slack or text or zoom or voice, whatever. And we'll learn each other's styles and preferences and work that way. And then I'll learn. You have your quirks and behavior patterns, and I have my quirks and behavior patterns. And I started to realize the AI does too.
The different models have different patterns and behaviors and preferences and biases and communication styles. So I'm going to here's my challenging question. Nobody has said yes to this. Have you ever asked AI? The best way to communicate with it?
Eden Ovadia: 16:06
I have asked it to help me give it prompts. So I guess in a certain way, yes. Like I'll give it kind of structure out, almost like give it help it, it will help me plan. And so I'll give it my goal. Then I'll say, help me ask you in the best possible prompt to achieve this goal.
And so somewhat, I haven't asked for specific communication preferences. Is that something that you've done?
Richard Walker: 16:32
Yeah. And it was unreal. And this came from a problem. I gave it a prompt twice and I got two diverging answers. And I said, I don't understand.
Why did you give me two different answers when it's the same prompt? And it said, because there are assumptions in this that aren't said. I didn't know if you wanted cost optimization or speed. So one answer is for cost. The other one's for speed.
They're different answers as a result. And I thought, well, what's the best way to communicate with you? Then I literally asked it like, how's the best way to do it's, it's akin to what you're saying about prompting. It's learning how to prompt better. I don't write prompts anymore.
I let it write its own prompts now, but I asked it like, what's the best way to communicate? And its answer was Jason like, whoa, hold on.
Eden Ovadia: 17:12
Oh, wow.
Richard Walker: 17:14
What do you prefer Jason? I was like, yeah, I prefer structured content over anything else. And I was kind of blown away. And that's led to various breakthroughs that I have. But you know, I'm, I'm going down this path that AI is more like a human and we should treat it more like a human, not exactly a human and get different results than if it's just a tool.
So to go back to, to what you're doing with FIINY and how you guys think about it, are you thinking about them as actual team members for yourselves and or your clients?
Eden Ovadia: 17:42
For our client, Sophia's platform? Yes. I think it's democratized for a lot of our clients, the ability to have a sales and marketing team. And so when we, when we think about some of our clients in smaller Ras, maybe they're just getting started, maybe 1 to 5 person teams, they typically don't have someone fully dedicated to marketing or sales or an associate even doing that. And so FINNY actually replaces the need for someone like an associate and also does the work that no one really wanted to do.
Like no associate was thrilled to be reaching thousands of people and then drafting personalized messages for each of those 1000 people and then following up with them. Right? So it replaces and it democratizes access to what would have been a pretty costly hire.
Richard Walker: 18:28
Yeah. Yeah, I think about that a lot. So where do you guys sit in this world of agentic workflows and trying to I mean, the whole world this year is about agent, agent, agent, agent, and even the definition of agent is fluid. Like what does it really mean? So what's your view of this and where do you guys sit?
Eden Ovadia: 18:47
Well, we build only for financial advisors. So we're in a slightly more regulated industry than most. And right now, we've taken the stance that the most agents can do. And that's called 99% of the work, but requires a human stamp of approval at the end of the day. And that human will be the advisor.
And so semi Agentic is where I would put things. And I'll give you an example. One of our agents helps our advisors with their content creation. And so what the agent can do is research the advisor, research the value proposition online, and then draft suggested content. It's up to the advisor to then read it over, approve it and post it.
Right. And so there's that last mile for both SEC and Finra regulations. You actually need to have a human, a human oversight. And so it's semi agentic, I would say, or agentic in the sense that it's autonomously researching, autonomously drafting, following up. It doesn't, you don't need to prompt it.
It's doing all this on its own, but still requires that final approval.
Richard Walker: 19:47
So I'm wondering if you guys are talking to companies that are doing the compliance reviews. I had Patrick Hannon of Fidelity Labs and they have a product. The name's escaping me is doing compliance reviews with 40 years of history at Fidelity of Compliance management. Are you guys thinking about doing that yourself or partnering with me?
Eden Ovadia: 20:06
So right now we are almost agnostic. We will partner with any of the compliance tools or teams that our clients are already using. It's built so we can just integrate really easily with any of their teams. The tool that's been coming up a lot, and I think they're an awesome company and team is addressing their compliance Ferrara's and they, from what I hear, are absolutely crushing it. And every advisor that I've spoken to that uses them is a big fan.
So they are, I think, the up and coming compliance for financial advisor tools.
Richard Walker: 20:37
Okay, cool. All right. I don't get to do this a lot. I want to talk about the future. I'm not asking your age, but you're younger than me.
You've come from this background of AI, I've learned it along the way. And I think you might have a different perspective of what the future looks like. And especially with what you're building and being a new category, new type of product. So where do you think we are going with AI? What part of our lives does it become and where do where does that relegate the rest of us?
Eden Ovadia: 21:08
Really good question. The answer that I keep reverting to is that it's going to bring out everyone's spikes. I don't know about you, but when I was so I was in college, but I, which is now in a different generation. And when I went to college, my parents told me, pick the path that will keep as many doors open for you. I'm sure you've heard this advice before.
And so I went into engineering, and then I went into consulting and private equity, and I just became the best generalist. I was a great generalist. I kind of knew a little bit about everything. And I was a jack of all trades. You can put me in a bunch of different situations and I can pretty much figure it out right now.
There's if I went back to college or if I was giving advice to someone who was, you know, kind of picking their major, I'd say like, pick the spiciest thing you can do. The thing that will differentiate you the most, the opposite of being well-rounded, like pick 2 or 3 things that make you like really weird and stand out and that none of your peers can do as well as you and that, and then let the AI kind of do the rest. Let the AI be the generalist, this general purpose AI that has all this information and can do that can do like an 80% okay job at basically everything, outsource everything to there, there, and then be better at the AI and be an expert in a handful of things that make you super different. And that's, I think that's where we're attending. We're trending towards less well-rounded people, more spiky people.
Richard Walker: 22:35
Okay. I love this term. Spike. I hadn't, I hadn't heard about that. I, I think what you're really getting at is that AI in some ways is allowing us to be the artisans again, where we can specialize and really go deep on something.
And look, I laugh about this. When I was 12, I must have raised my hand and said, when I grow up, I want to do forms because I know more about forums than probably anybody else in the world. And it's super geeky, nerdy, and boring, but I'm excited about it. And therefore I can do things nobody else can do with forms. And I think that creates the differentiation that you're talking about is having a specialty, having a focus.
That's really interesting. Do you mean, where do you see products going? And I'm going to give you an idea, by the way, I am right now designing a website that is primarily driven by AI. I'm kind of done with the whole, hey, I got to log into a dashboard, click, click, click, click, click. Why can't I just tell the AI, hey, go add a user.
Go tell me the status. Go start a project, whatever it is. And I've never seen your product. So maybe your product's already like that, but where do you think we're going to go with tech and AI as part of tech?
Eden Ovadia: 23:44
Yeah, this is something that I've said from the very beginning of AI, which is that user interfaces are going to become obsolete and AI is just going to start meeting us the way we communicate with each other. You kind of alluded to this earlier when you were talking about everyone's different communication styles. And so let's say, Rich, you're, you know, you love to kind of think out loud and you're a really great verbal communicator. Your eyes should, you should be able to prompt them and give orders and interact with all your systems in the same way that you feel comfortable, which might be a phone call. And so I think there will be a world in which you can just call the different apps that you need and communicate and give orders and prompts over the phone, and everything else should be abstracted away.
And if someone prefers written, if someone prefers email or text, that should be available to them also. And so learning, having, having to retrain yourself and relearn all these different user interfaces, I think is going to be a thing of the past. It's really clunky. You know, it's hard to remember. User interfaces are changing so much also, as technology is changing.
I just don't think that's where the future is going.
Richard Walker: 24:49
Yeah, I love this vision, but I'm also kind of stymied with how we get there because we have all these walls, right? We have different authentications and we have different devices. I have an iPhone, maybe somebody else has a Google Pixel or whatever. You know, when you're talking about this idea that we just communicate and it does what we need. Oh, the first thing that came to mind was my to do list.
I can't figure out where to keep my to-do list. It's in eight different places. So if I could just be on a walk and oh, I have an idea. Hey Siri, add this to my to do list and it goes back in and does all the work. How do we get there?
Like how do we actually link all these pieces together? And what does it really look like? Do I want to log into a ChatGPT interface every time I want to go do something with software? I don't think that's right.
Eden Ovadia: 25:33
Yeah, that's a really good question. I'm not sure. I don't have a strong opinion on whether it'll be a series of apps that are just all available to you on your phone, by phone call or by text, or it'll be some sort of super app that kind of like a super to do list or a super aggregator of like your personal model that is the layer between you and everything else, almost like an orchestrator where you're giving, you're putting a to do list in one place, and then it's assigning things to ten different apps and it's organizing all your life for you. Kind of like what a chief of staff or personal assistant would do right now. Yeah, that I'm not sure about.
It's a good question.
Richard Walker: 26:09
All right. Well, this leads me into something else. And this is a challenge that I have been playing with for nine months. And it really comes down to this deterministic versus probabilistic software is built so that one plus one always equals two. It always does the same thing every time.
And I already mentioned I gave the same prompt twice to an AI and it gave me two different answers and it probably would give me a third answer. The third try. How do we get AI to be more deterministic and therefore can we make it truly autonomous?
Eden Ovadia: 26:41
Yeah, it's something we think about a lot, obviously, as we're building AI Tools in an industry that's regulated, we actually need it to like force it to start being more deterministic. AI has gotten a lot better at being more deterministic with structured outputs over the last, let's call it two years. I can actually nerd out a little bit and tell you about really interesting products and development. Learning that we did that had to do with structured outputs and more deterministic outputs. So when we started two years ago, we wanted to give our advisors the ability to search in natural language similar to how you would query ChatGPT.
And so we wanted them to be able to search for a person or run a query to build an audience in English. And then the idea was that would actually then query an entire database of 300 million people and output a structured output of a list of people. The problem is that two years ago, by 2024 AI was really, really bad at deterministic structured outputs. And so it just was breaking 25% of the time, 25% of queries were breaking it. And that's obviously not production ready applications.
And so we had to come up with the idea of. Natural language and to structure outputs. At that point, we're now bringing it back for the first time in two years because the underlying models have gotten better and better and better at kind of converging towards deterministic answers. If you can structure the prompts in the context in the right way. So I think we're getting there.
Richard Walker: 28:09
I think the models play a really, really big part, but I have a different strategy.
Eden Ovadia: 28:13
Okay.
Richard Walker: 28:13
And I actually proved this out recently with my strategy. So first of all, the strategy is the same thing with humans. If I give you one task that's simple to do, I would reasonably expect you to be able to do it. If I gave you a task that was 18 steps, all encompassing with, you know, five different topics, etc., that's really hard. And we're not going to get the exact results every time.
So AI is similar way. I think the smaller the work item is, the more deterministic the AI acts as long as you give it boundaries and role and you understand trying to help it understand what you're doing. So what I've done is I've been building software that builds software. And when I break it into small enough pieces, it just works. I tested this, I gave it the higher end model opus versus sonnet.
At Claude. The opus result was worse. The sonnet result was better because the bigger model, the more advanced model, overthought it, overengineered it, and didn't just want to do the task anymore. Whereas sonnet was like, I'll do the task. Sure.
No problem. Get it done. So I think part of the strategy is to think about our workflows and our pieces of the puzzle and make them small enough so that the agent can be focused so well that it can be more constrained in how it works.
Eden Ovadia: 29:32
That's super interesting. What kind of tasks were you giving to sonnets, for example?
Richard Walker: 29:39
Let's see. This was in the form of this. Well, this was actually software development, so it had a task to do some piece of software coding. And what it ended up with sonnet would come out with a smaller footprint than opus would write. Opus was just more robust.
I didn't need to be more robust. I needed the exact thing sonnet was doing over and over again. So I was paying higher fees, taking longer to do it, and getting more outcomes than I wanted. Another one was document generation. So like, think of it like a business plan.
I have this whole conversation, turn this into a business plan. I don't need a 28 page business plan. I need the concise version. And sonnet was like, here's the concise version. It was just fascinating to see.
You know how it is.
Eden Ovadia: 30:21
Super fascinating. I'm wondering, do you have any tasks that you still would give to opus?
Richard Walker: 30:27
No, not not the way I've architected this. Frankly, I know opus is really advanced, but I think it's so advanced. I'm not doing anything that advanced. I'm not trying to calculate, you know, DNA sequences or something really crazy. I just think of software and the tasks that we're doing as bite sized tasks.
So for example, here's another thing I did. I created a voice profile of myself. I had Claude Cowork using sonnet. Just go get all my transcripts of all my podcasts. That was one step, second step, second task.
Take all those podcasts and isolate my words from it. Third different task. Read all those words and figure out what the density of statements and themes and things and how I talk. So you can replicate how I talk better and analyze it. I don't know, like maybe I could have done all that in one step with opus, but I just think if you give it one thing, it does one thing really well, then you can move on to the next thing.
Eden Ovadia: 31:29
Yeah, I'm gonna have to test this out because this is definitely super interesting. I hadn't thought about breaking up complex tasks into more simple tasks for lower cost, lightweight models.
Richard Walker: 31:39
Yeah, there's actually a repo out there that does this even more, and they claim they'll reduce your API costs by 75% because they can isolate lower end tasks to lower end models, even lower than sonnet, for example. I'm not thinking at that level because I'm kind of generic. Just get it done. But do you guys have and this will be the last question because we got to wrap this up. Do you guys have an AI design philosophy that is driving how you do things inside your company?
Because I've never heard anybody talk about how we approach AI? It's just like using it. Go.
Eden Ovadia: 32:11
Yes, this is so timely. Over the last two months, we've been building this internally. So for the non-technical staff on our team, what we do is we actually have an entire. Oh, we have cloud set up for everyone on the team, and we have entire internal tools available to them for them to start building on top of for their own personal productivity. And we heavily encourage the entire team to do that.
That's on the kind of like non-technical on the technical side, our engineers are spending a lot of their time researching and just staying up to date on everything that is AI versus just now, like product development, like we've actually traded off like product velocity right now in lieu of just better research, because I think you can do a lot of, you could do a lot more work if you just use the right tools. And so the opportunity cost of not using those tools is really high. So we've actually been forcing a lot of and encouraging a lot of lunch and learning, bringing in guest speakers, and hackathons. We have an AI hackathon coming up this week for our team to showcase the best internal productivity tools for the, for their, their own use.
Richard Walker: 33:18
Oh my gosh, I love that. It's so inspiring. My team is asking more and more to do stuff on their own and I'm encouraging it. I would like to get to that point. So good inspiration.
Okay, I'm running over time, so I gotta wrap this up before I get to my very last question. What is the best way for people to find and connect with you?
Eden Ovadia: 33:35
And yeah, so I'm pretty active on LinkedIn. I, so my LinkedIn is just my name. Our website is phinney.com. And so if there's any advisors that want to grow their businesses, I'll give a shameless plug and say go to finny.com and book a demo. We're available at most right now broker dealers and Ras.
And so that you can just sign up and get yourself started directly. And, and yeah, those, those would be the best ways to get in touch with me or my team.
Richard Walker: 34:04
And from what I heard, book the 30-minute demo, really get the full picture, maybe a 45-minute demo. Give her the time. Just helping you guys.
Eden Ovadia: 34:12
Exactly. And then if you're curious and you want to do async learning, you can also just look us up on YouTube. We have a ton of content there for you to just learn about any on your own time.
Richard Walker: 34:22
Awesome. That is awesome. Okay, here's my last question. I'd love to ask who has had the biggest impact on your leadership style and how you approach your role today?
Eden Ovadia: 34:32
I love that question. It's a good one. I haven't had to think about that in a while. So my answer is a little off the cuff, but I would say the leader in our industry that I think about quite a bit, and I think I've emulated a lot of learnings from him. Is Jason Wenk the CEO of Altruist?
He, over the last few years has become an angel investor in my company and also just super helpful to me. But I think the reason that I look up to him is he architected the idea of really building in public, creating an entire cult like following on LinkedIn or social media, specifically in our industry. And people really want him to win because of it. Like people are rooting for him, whether or not they're altruist clients. And I think that's just a really, really great leadership skill to have.
And so I've tried in my own version and format to emulate a lot of that just by being really public about, you know, what's going on under the surface, behind the scenes at the wins and the losses. What's it like building an AI company and just trying to really help, you know, lift the curtain on everything we're doing. And a lot of that I get from Jason.
Richard Walker: 35:46
Oh, that is awesome. That is really awesome to hear. I wish we had more time and you and I should have more conversations. This has been so much fun and I got to wrap it up. So I want to thank Eden, the CEO and co-founder of FINNY for being on this episode of The Customer Wins.
Go check out Eden's website at FINNY finny.com. And don't forget to check out Quik! at quickforms.com where we make processing forms easier. I hope you enjoyed this discussion. We'll click the like button, share this with someone, and subscribe to our channels for future episodes of The Customer Wins. And thank you so much for joining me today.
Eden Ovadia: 36:19
Thank you so much, Rich. This was a lot of fun.
Outro: 36:22
Thanks for listening to The Customer Wins podcast. We'll see you again next time and be sure to click subscribe to get future episodes.
Comments