Swift Chat with Daniel Satchkov: AdvisorAI and 401kAI Tools for Financial Advisors

In this Swift Chat, Marie Swift of Impact Communications, Inc. speaks with Daniel Satchkov, Co-Founder and President of RiXtrema, a pioneer in developing innovative tools for financial advisors using machine learning (AI) and data science. Daniel dives into RiXtrema's two groundbreaking AI-driven productivity tools: AdvisorAI™ and 401kAI™.

401kAI is a tool for advisors to use as they court and seek to win the business of plan sponsors. It contains detailed data on 850,000 retirement plans, with full information on all current laws, regulations, and legal cases, and is integrated with compliance tools. 401kAI can summarize dozens of plans with one click, help write custom emails for the plan, and create effective calling pitches for a plan.

AdvisorAI is a tool for advisors to improve productivity with portfolio analysis and creation of custom emails and calling pitches. The tool summarizes all aspects of portfolio risk in one click, creates custom emails for each lead, learns in detail about each lead’s financial situation, knows services the advisor can offer to get a meeting, and is integrated with their compliance tools.

Learn more about Daniel, RiXtrema, 401kAI™, and AdvisorAI™ at RiXtrema.com

Listen to Podcast/Audio Version

   It (401kAI) knows about over 800,000 plans. So, when you go into a plan you can go to 401kAI and just say summarize this plan for me. The algorithm goes through all of the data, all of the filings, and gives you all the key points about the plan in one executive summary. One of our clients, when they saw that, they said, oh I get it. It's like 3 hours of work in ten minutes. This is what it does."

~ Daniel Satchkov, Co-Founder and President of RiXtrema

Transcript of Conversation

​Marie Swift: Well, hello and welcome back to another great Swift Chat. Today we're talking about AI and artificial intelligence. We know it's all the rage. Everybody's talking about it. How do we use it? How do we use it compliantly and efficiently as professionals, especially financial professionals, where there are so many concerns about security and compliance and all of that good stuff? Today I'm joined by Daniel Satchkov, and he is the president of RiXtrema. RiXtrema is a leader in the development of state-of-the-art AI solutions for the finance industry. And Daniel, you and your team recently developed two new products using AI. I watched some of the demos today and I think it's truly groundbreaking and astonishing. Welcome to the show, and tell us a little bit about what you've developed.

Daniel Satchkov: Thanks, Marie. Hi, everyone. I will say right at the outset that I use the term AI just because everybody else uses it, the word intelligence certainly doesn't apply. These algorithms are immensely powerful, but they're not intelligent. That's a separate kind of rant. I do it in a lot of videos. Not today, probably, but it's certainly not intelligence.
 
What these algorithms are, is they're extremely powerful equation solvers for those who remember their math. If you remember the equations you solved, let's say in high school, they could have had two unknowns. Or it could be one unknown, but different orders, maybe XX squared, quadratic equation. Well, a neural network like ChatGPT, for example, has 180 billion parameters or so. You can think of it as a machine that can solve an equation with 180 billion unknowns, which is pretty unfathomable.

This means that it can solve many tasks that require any kind of analysis, pattern recognition, or any kind of repetitive work. And people don't understand, but there's a lot of repetitive work, especially in finance. Whenever you're sitting there reading a document, like a filing or a prospectus or something, you're trying to make sense of it. Or you're refreshing yourself on the conversations you had with somebody over the last year, and you're re-reading the emails or something, all those things involved, they're not creative. Those are not intelligent tasks. Those are all pattern recognition summarization tasks. If you're writing communications that are more or less standard, but tailored to each individual or a firm, but they're still not personal. They're personalized. They can use personal, individual data for example, you're writing to a 401(k) plan sponsor and you're using plan data, but it's not individual. There's still enough repetition. AI can do that for you very well.
 
The two products we released are 401kAI and AdvisorAI. Those have been game changers. Since release in May, 401kAI has become the fastest selling product out of the gate. We did a live session in May, we were kind of blown away. There were close to just under 1000 advisors who signed up for it and there were, I want to say, 340 attendees live. For those who put together live events for financial advisors, you probably realize that this is not a typical event.
 
Right now, everybody's feeling some excitement but some uncertainty. But also, there's a lot of fear mongering and hype and all those things make for a really bad combination. Because on one hand, people believe that AI will solve all humanity's problems. On another side, they believe that it will enslave and destroy humanity because it is much smarter and evil. And on a practical side, people are afraid of what it’s going to do to my business, what are my competitors doing? All those things, there's so much noise that it's very hard to get through. That's why I think people flocked to those presentations we did. They really wanted to understand and cut through that noise.
 
To cut through the noise, 401kAI is a tool that does some very important tasks. It's for retirement plan sponsors. As a retirement plan sponsor, let's say you need to analyze a plan. Just start with, you open a plan filing or documents and we have a software that does that called Larkspur Executive, it has a lot of real time daily update data about the plan. But when you analyze the plan, it will take you some time. I mean the filings themselves could be 50 pages long.

It could take you, even to summarize, it could take you an hour, it could take you two hours, it could take you longer to do a good summary of what's going on with this plan. Now imagine you have hundreds of plans that you prospect with. Obviously, your clients, you probably would know them already, but if you prospect, you would not. Our tool allows you to go in and 401kAI actually knows all the data about qualified plans in the US. It knows about over 800,000 plans. When you go into a plan, you can go to 401kAI and just say summarize this plan for me. The algorithm goes through all of the data, all of the filings, and gives you all the key points about the plan in one executive summary.
 
One of our clients, when they saw that, they said, oh I get it. It's like 3 hours of work in ten minutes. This is what it does. And that's almost like our slogan, really. I think that advisors who don't adopt this technology are going to find themselves way behind very quickly. This is not going to take years, just because the technology itself is not new. At RiXtrema, we have used, “AI” for eight years now at least. And this is our background, in machine learning and data science. That's kind of who we are. But we used it behind the scenes. We used it to gather data, to produce reports, to create various data sets that are powering the app. On the front end, we did not, because people still didn't understand how it would be used, and technology was not there. Technology was there for us to use it internally. And our whole business is based on these algorithms, really.

But to put it on the front end, to put it in front of people, there had to be kind of a ChatGPT moment, right? People realized, oh, it can do some things. Now, ChatGPT is very powerful, but it's also very problematic. The reason it's problematic, for one thing, is that it does many things well. Like it can write rap lyrics, it can write Shakespearean sonnets, it can write koala stories, all those things that 401kAI or AdvisorAI don't really care about. We don't care about doing that for our clients. So therefore, ChatGPT wants to be powerful everywhere, but it's really lacking in some very specific areas. Also, ChatGPT knows the world only up to early 2021. Imagine trying to make sense of a retirement plan and not having access to 22 or 23 data. I mean, you're lost. It doesn't matter that you have hundreds of millions of dollars of computing power that went into it. You're just done. You're not going to be able to use it. So, 401kAI and AdvisorAI are both built as financially trained models that also have access to the latest data.
 
My prediction for those that follow this space is that this whole AI technology will move into that area, where instead of building chatGPT like massive behemoth models, people will build focused models that have topic knowledge, expertise, domain expertise. Also, those models will be able to command data that are maybe that just came out minutes ago. That is the power of the model. When you have a model that knows the world up to 2021, that's more of a curiosity. Yes, it can help write you maybe some emails or something, but it is not really actionable because it's outdated.

Our goal for a long time was to create a model that will be timely and that will produce real work. As an advisor, what we believe is that if you look at your workflow, you can easily find maybe half, maybe more, maybe less in some cases, but a very significant amount of tasks that are essentially summarization, reading, repetition, and communication that is not necessarily highly personal. By not highly personal, I mean you write an update to your client, it still talks about the client's portfolio or client's retirement plan, but it's just not a personal communication. It's similar in some sense that you would write to others. Therefore, all those tasks that you're doing right now, let's say it's 40% of your workflow to be conservative, all those things could actually take minutes instead of tens of hours a month or a week.

Therefore, these tools, what they're going to do for you, they're going to free up all your time. We believe that the goal of these tools is to free up your time for you to be more human. Because there are some tasks that those machines are not good at. They're not good at answering questions. I mean, if your clients see that you're answering questions with the help of this technology, they're not going to like you. You're better off focusing on the human part of it, and on the part that requires creativity, as opposed to doing some human work, but also doing a lot of mundane work that's really dragging you down. Therefore, the 401kAI and AdvisorAI, AdvisorAI is a more general tool for all advisors who work with individual clients that can help you analyze things. I'll give you one example. For example, you have a statement, a document you need to read or summarize, maybe a statement from the client.

You can actually upload it in RiXtrema in AdvisorAI or 401kAI and say, summarize this document for me. Give me all the mutual funds that I mentioned in this document, give me their expense ratios. So instead of working with these client statements or filings, you really have the data that will be given to you in minutes and then you can do the human part of the work. Same thing with communication, same thing with analysis of any kind of document. I think that where things are going, is that very soon you'll find that if you're not using this type of technology, you're going to be wasting an incredible amount of time that will really put you out of business.

Swift: I watched your demos this morning for both the AdvisorAI and the 401kAI, and it's quite remarkable. I would invite all of the listeners and viewers here today to please go to the websites, and we're going to put those links right here in the video and in the show notes as well so that you can go and take a look. I've been experimenting with ChatGPT and also with Reword and beautiful.ai and some of the things that writers and content creators like me like to use. Not every advisor needs to be a content creator to write that creative work. Some do, some outsource it, some do it internally. But I was struck by looking at your products, how it does the serious work. As you said, it's not going to write something in the voice of Hemingway because you're not trying to provide that kind of creativity.

Could you just talk a little bit more about some of the use cases? For instance, writing an email to the client, or pushing information back and forth to the CRM? And I think you also have a free CRM, right?

Satchkov: We do, but we don't necessarily view ourselves as a CRM provider. It's more for an example, and for those of our clients who don't have needs for a major CRM. We prefer to partner with CRMs. But the CRM is a perfect use case for this technology. Imagine that you were on vacation. You come back and somebody calls you or emails you, you need to get back to them. I mean, that's the moment I dread, I forgot. This conversation was not top of mind for me. I come back from a vacation, now I have to go in there and reread the whole email chain. That is one of the most painful things that can be done, because the conversation could have gone on for a while, and you have to really refresh all that, and only then you can follow up.
 
Now imagine that instead you just say, summarize this for me, and AdvisorAI will actually go through all of this communication. We'll summarize it for you. We'll give you all of the outstanding unresolved issues. We'll give you all of the issues that were resolved, all the positive things, negative things, like all the points of emphasis that you need, and that's it. This is like having a personal assistant, but a personal assistant with very good comprehension. They have good reading comprehension and domain knowledge. I think that in itself is huge.
 
Most people, when they think of ChatGPT, they think of writing because it's called a generative model. But I really believe that generative part comes a little bit later. I think that you can generate good emails with it. I have, and I will adjust them later. You can generate communications with it, and certainly our tools do that. But it would be a big mistake for people to think only about that. And that's the unfortunate part of ChatGPT. They kind of think about it as the machine that talks. The machine that speaks. It's like a parrot, like a very well-read parrot.
 
I don't think that's right. I think you need to step back and think about what this technology really does. Before communicating, what it does, is it's able to summarize, it is able to convert a lot of words into vectors. The way these things work is not particularly magical. Somebody long ago, a data scientist figured out that you can convert words into numbers, into vectors of data, and those words then, amazingly, you can do math with them. Believe it or not, you can actually take words for foundation, walls, windows, a roof, a ceiling. Add them up and you're going to get a number that's close to a word house. You can take a word vector for man, subtract a word vector woman, and you can take a word vector for king and subtract vector for queen. The difference between them will be almost exactly the same. You can take the word for United States and subtract the word vector Washington and you can do the same thing with France and Paris, and you're going to get almost the same thing. It's pretty amazing.
 
At first it just struck me when I saw that back in 2013, I was like, this is unbelievable. But then if you think about it, you realize that the words, that's how they behave. Essentially, a language is a very large mathematical model. I don't mean mathematical in a reductionist sense, where it's mathematical, therefore it's boring. Mathematics is the most beautiful poetic thing in the world. There was one German mathematician, he was asked, where's that promising pupil of yours? I don't see him around. And he said, yeah, he didn't have imagination, so he went off to become a poet.

Math is very beautiful, and language is very beautiful, and the beauty between them is that they share it. It is not math, as in two plus two, four. And it's not very simple math. It's something very elaborate. It's like mathematical models of language. So therefore, when you're looking at a document, however complicated it is, it can be converted into numbers. And those numbers, then you can do very specific things with them to understand how these words are related to each other, what they really mean. In linguistics, there is a saying that you shall know a word by the company that it keeps. That's how we learn language as children, think about how you know the meaning of words. Parents don't come to you and point to everything in the world and say this is this. Yes, you get that for some basic things. You probably get some images and words next to them. But beyond this basic word set vocabulary, the rest you're going to find by the company it keeps. You're getting it from the context.
 
That already suggests to you that this is a mathematical sense in the finance, and all of our listeners are finance people. You know the word correlation. Correlation is ubiquitous. Usually when you ask people to continue this phrase they say correlation is not causation, which is totally correct. But in the word sense, correlation is what drives these models. It's like what is the word France correlated to in a certain way? I use correlation in a simplistic sense, it's more complicated, but it's correlated to France, and so on. There are many words that will be correlated.
 
When you think about the mundane tasks you do, you have to read a PDF document, you have to read an email exchange, you have to read notes in the CRM. All those things can be automated. Not because AI is super smart and smarter than you. There's no way. AI is not smarter than humans. This is all a lie. This is a separate topic. It is not. Just like calculator, is not smarter than humans just because it can do cosine and sine functions that we cannot in our head. But a calculator is a very useful thing. Those people who did not use a calculator for certain tasks, they went out of business. So is a computer, right? And so is “AI,” let's make use of the acronym.
 
If you think about your work, really, I know most people are just used to what they do. I just do this, I have my routine and it's fine. But if you think about your work, there are tons of things that you do, even if you take out email writing. When people think about ChatGPT, oh, it's for writing emails. Then they evaluate based on whether you can write a good email or a bad email. But that is not the main thing. That is actually secondary. The primary thing is that it can digest and analyze and summarize large amounts of data for you. If you really think about it, all those email reading, CRM card reading, prospectus reading, plan document statements, or personal statements, retirement statements from individuals reading, all that stuff is actually things that you should not be doing, that humans should not be doing anymore. And this is already here. That's why I think you'll have some people who are early pioneers, and then there are some people who are early enough to save their business and prosper, and then there are going to be some people who are going to wait too long.

Swift: One of the things you said on your website is the 10X Advisor, and that's what you're talking about is doing the things that matter most and using technology as a springboard to get this other work done. For instance, I've been using Fireflies to record my meetings and to do summaries, and it's remarkable as a time saver and pretty darn good. We use them for our call notes. Now a human being goes in and touches them up because there are nuances that we understand about our clients that Fireflies Ai never will. But I love what you're talking about.

Satchkov: Well, it will. When there's a time of personalized models, which we're coming to rapidly, it will. But I will say that in Fireflies, I'm fascinated by that product, but there's one thing that we're also building is that ability with a tool like Fireflies, which does live transcripts. We're speaking, it transcribes it. Imagine if we're in a call and I'm your client, you're my client, and there's a question that a client asks that you don't know the answer to. Financial question. Imagine you had a button that you would press next to that phrase, and you would get a full analysis and things you need to know about it right away, right there. That's what it's kind of coming to, because that's not creative. People think of knowing a lot of facts as being creative, but it is not. It's just clutter in your head.
 
I don't know who likes Sherlock Holmes here, but if you remember, Dr. Watson famously asked him about if he's ever heard of Copernicus. And Sherlock Holmes said, no, I don't know about it. He said, how can you be an educated man without knowing about Copernicus and all this stuff? And he said, I don't need it. When I need to, I can look it up. I can't put all this garbage into my head. This was an extreme point, but it actually is a valid point. Knowing things, it is not creative work. It's good to know a lot of things, but beyond certain points, it is much better to free up your capacity to do the empathy, the empathic things, the humane things, or to do analysis. Analysis is very different from all those things. It's not summarization, it's creative work. But if you have somebody else that can give you the fuel for analysis, then you can analyze and do very well. That's what I think these models will do.

Swift: Well, we're in a brave new world. If I think back about my career, I've been in the business for 30 years, we've come a long way. We could have a whole conversation about that. But I want to go back to what some of your additional solutions or resources that your tools do. There was something that I saw on your site about a mobile app. Could you tell us a little bit about that?

Satchkov: Well, all our apps come with the mobile app. The main purpose of the mobile app, by the way, is not to sit there and try to actually do work on the phone. That's not very healthy for your eyes, or your attention, or your social life. No, the purpose of a mobile phone is when you're on the road, to be able to get notifications quickly. To me, that's the main thing.
 
Let's say that you're working with somebody and they decided to open an email from you, or they decided to go through a questionnaire. In all those cases on our RiXtrema app, you're going to get a notification, and then these tools, 401kAI or AdvisorAI become paramount. Think about it; I'm on the road, somebody opened my email. But who is this? I send emails to maybe 100 people, maybe 200, maybe more. Who is that person? You can quickly click and say, summarize this lead for me.
 
There's a tool in 401kAI and AdvisorAI, say create a calling pitch for me, tell me what to say on the phone to get started. It will give you what to say to this particular person with their history, with their numbers and everything. Then that will give you enough information to have that quick conversation to maybe set something up later. In the business world, of course, if you get back to people quickly, you have much higher chance of winning business. That's what we try to capitalize on. That is the only reason for existence of the mobile app.

Swift: That's remarkable. Then something about campaigns. How can people use campaigns in your tools?

Satchkov: This tool predates 401kAI, but it will be integrated with it. It's a tool that allows you to send massive campaigns. They're not AI generated. Those campaigns are written by copywriters, reviewed. The AI does not send email campaigns in our software. That is not a good idea at this moment for compliance reasons and many other reasons. The campaigns that are sent out are predictable though. They're customized to each plan, but without the use of AI, with a different algorithm where certain variables based on qualified plan, let's say their admin fee or the red flags get populated in the email. This tool is very popular with our clients. It predates AI. It will eventually work with AI, but in a very compliant manner. Right now, it is a tool that you would use for email campaigns. That's why I say that when you think about 401kAI or AdvisorAI, I don't think about emails primarily because there are many other ways to create good emails.

You can pre-write them, fill them up with variables and so on, that are very well tested in marketing software. And it's called Larkspur Executive PRO. In our system. LE PRO is the email campaigns. AI I think is a different level of a game. It's not so much for writing email though, you can do that. It is much more for analysis and taking out this kind of mundane summarization reading work.

Swift: One of the things that I'm sure our listeners and viewers are thinking about is security of confidential client information. As I've been looking at ChatGPT and even working with Fireflies from my own experimentation, I worry a little bit about confidentiality. What's happening with the data, and should I be careful about that? Could you address that?

Satchkov: That, I think, is a huge concern. That's why I think that until the models are personalized or created by a very specific provider you trust, that you would trust with very personally identifiable data, it's not going to work for business. ChatGPT is not going to work for businesses, because according to the terms of service of OpenAI, whatever you put in there can become their property. Also, what we'll find is that increasingly you're going to need to fine tune your own models. Fine tuning is a term in machine learning where the model learns very specific for specific tasks. For you, it can learn nuances about your clients and about your view, and then it will use that to transcribe better. But you cannot do that right now because, first of all, in ChatGPT you cannot fine tune it, but also you can't fine tune it because if you fine-tuned it, then other people will take advantage of what you did.

Therefore, we're moving into a world of proprietary models. That's where with RiXtrema we are releasing a model that is finance aware. It's much more than 401kAI or AdvisorAI. I'm kind of getting ahead a little bit, but that is going to be completely private. That is going to be only at RiXtrema, a model with very tight safeguards where every client will get their separate, what we call adapters. In the machine learning world, it's called adapters. There's a large model that is trained by RiXtrema, that's a financial model, that's much better than ChatGPT at finance. It's housed on RiXtrema servers that are highly secure. In the agreement, it says all your data is yours, it never seeps through to anybody. Then also you can fine tune the model, but what you fine tune is not the overall model, but your own adapter. It's never applied for any other clients. This kind of infrastructure, I believe is the future. It will answer the privacy concerns and I believe it will make ChatGPT more or less obsolete.

Swift: Wow, well, that's amazing. In the last few minutes we've got, I know you've got the two products available now, so it's AdvisorAI and 401kAI, and you have a money back guarantee, right?

Satchkov: That's right. I think that many people are confused by the hype, by the hope, by the fear, and that hype clouds certain things. Then also this perception that ChatGPT is only for writing, which I think is completely wrong. These models are not just for that. That's why we give money back guarantees to people. When they get the tool, they buy it, within 30 days you can just return it and we'll support you. The reason for that is we want people to understand how exactly it is helpful. If they're not ready yet, they can come back later. But you really need to try it to understand what it can do for you. Because if you don't try it, you're going to be hostage to some of these things, like it's super 800 times smarter than humans, it will enslave us all, and all this stuff.

Now, I'm not saying that AI cannot cause harm, but causing harm and being smarter than humans is not the same thing. Like, a piano that is dropped from a 30th floor can cause a lot of harm. Doesn't mean that it's sentient. It is a very powerful tool and a weapon, even, this technology, but it is not smarter than humans. Once you remove that, you remove all this kind of noise, and separate the wheat from the chaff, you will see that this model can do things for you to really cut down on 40% of the hours that you spend on grudge work. That's incredible I think.

Swift: Are you going to be at any conferences this fall? I'd love to sit down and have a cup of coffee and talk some more about this.

Satchkov: I think my colleagues will be but honestly, I'm so hunkered down with these developments that we're releasing something even bigger. It's coming in September you will see. 401kAI and AdvisorAI are already the most successful products for us. But something even bigger is coming, and I have to do my math geek work and hold off on conferences for now.

Swift: Well, I can't wait to see what's next. Daniel, thank you for being here.

Satchkov: Thank you, Marie. Thank you, everybody.

Find more Great Swift Chats Here