The Software Agents is a new podcast series sponsored by CloudBees. Each week we bring you leaders from many fields applying software to reimagine them for the new world under construction.
Shelter-in-place orders drove a rush of new signups for automated investment tool AgentRisk, as professional advisors proved suddenly unreachable, or unable to conduct transactions from home. Founder Jon Vlachogiannis, a Greek-born entrepreneur who went from teen coder to serial Silicon Valley startup founder, says many human advisors were discreetly using AgentRisk already. Jon explains how AI is transforming personal wealth management, and why he thinks Silicon Valley culture still matters to drive innovation.
Announcer: Welcome to the Software Agents, a podcast series that brings you the people using software to help change the world in this time of transformation. With the help of some of the brightest technology minds, we’ll explore how almost every possible area of life, society and business is being reimagined through the power of software.
Christina Noren: Welcome to one of the first of many podcasts on how software is helping the world survive and evolve right now, as told by the people who are making it happen. I'm Christina Noren, and by co-host is Paul Boutin
Paul: Hello, everyone, thanks for tuning in.
Christina Noren: This podcast is sponsored by CloudBees, the leader in software delivery automation and software delivery management. With over 40 percent of the Fortune 500 relying on its leading continuous integration, continuous delivery, feature flagging, and cloud native development products.
Today we have a good friend of mine, Jon Vlachogiannis, who also can go by Jon V, and we connected because we’re both ex-Splunkers, but he was the bootstrap founder of a company that became Splunk’s first acquisition post-IPO, which was a wonderful outcome for him and his team. And since then he’s moved to the United States, moved to Southern California.
He’s become a real hub of the Santa Monica startup ecosystem and a real connector, and he has bootstrap-founded a new company, not so new now, AgentRisk, and that’s what we’re going to focus on a bit today. But before we go there, I’d love to have Jon tell us… Jon, tell us a little bit about your journey from being an engineer in Greece to being a two-times successful startup founder in the United States.
Jon Vlachogiannis: Of course. Thank you, guys, for having me on the show. I'm always super happy talking with people about tech and entrepreneurship and startups. Actually, the company that Splunk acquired, Bugsense, was my third startup in the United States. The previous one had very, not similar, but it had a very interesting path to success, I would say. It’s a very long story, we’d have to talk for hours.
So I would focus mostly on how I managed to move to the United States and especially California. As you probably know, the dream for every single person in Europe is to make it to the Silicon Valley. Because the smartest people when it comes to technology, they are there. You can talk to them at the coffee place. The funding is there.
Sadly, most of the funding for crazy ideas happens in the US. And the environment. The environment is something that, you know, it forces you to become better in business, in creating new ideas, in managing risk, in taking foolish risks. But, yeah, generally this is the place that I wanted to be. I thought it’s going to be a good opportunity for me. But it’s difficult to be among the best.
There are millions of people out there that are doing something similar to what everybody does, so you have to be literally the best to come and work, especially for a public company like Splunk. When we started Bugsense, we were in Greece. We had a very small office in Greece. But our customers were everywhere in the world. We were the second biggest mobile platform in the world after Google.
So we thought that if you are the best, you don't need to be in the United States. And at least you can grow outside of the United States. And that was happening for a long time. But as you understand, it’s a little bit difficult. It’s not that easy growing that fast when you’re outside of the United States, outside of California. And I kept saying, it’s like building a racing car and you’re never going to the track.
Because funding from – very small funding, like $100,000, from a seed fund here in the United States, and we started, a center office actually also in California. And we tried to attend every single event that was happening in San Francisco.
Christina Noren: I actually had wanted you to back up to even, how does an individual software developer in – I think you were in Athens – come to start a company like Bugsense, and then that journey.
Jon Vlachogiannis: To tell you the truth, it’s actually hard work. For me, I remember programming for like 12, 15 hours, consuming every single manual I had in my hands. I remember myself programming all day long. It was literally going out with friends and programming, no TV, no Netflix, nothing at all, and I really loved it. So for me it wasn’t like hard work, it’s boring, I don't want to do it – it’s like I love the thing that I can create a whole universe using algorithms.
And then I am the king of this universe. So it wasn’t really difficult for me, but I could see that I knew way more than my colleagues in Greece. And surprisingly, way more than my colleagues in the US. That’s why Bugsense became such a huge success, because we literally had the best product. Even until now, I Google our brand, and it’s still there.
There are no companies that still have managed to do what we did. So it’s literally: be the best, keep consuming stuff, keep consuming I mean information for your art, for what you want to do. Not consume like sit in front of Maverick’s. That’s the only way that I know to make it surely but steady in the United States.
Christina Noren: So let’s maybe drill a little bit on the tidbit of how you came to sell Bugsense to Splunk.
and then I'm interested in, you know, you stayed for awhile, as you had to, and then there was an impetus to go start something else and move again, you know, and let’s take it at the start of AgentRisk and what AgentRisk is about.
Jon Vlachogiannis: Yeah, the interesting thing is that when we got the offer from Splunk, the company was going pretty well. We were making a lot of money.
We had some VCs reaching out to us and they were interested to invest, and we decided Splunk because we loved Splunk way before they approached us. I remember we had some presentations, and Thanos, my cofounder, had on our deck, “we are Splunk from about data.” Literally, without ever talking to Splunk. So we love the culture, we love the product.
And when they approached us we were like, okay, guys, we excel at collecting about data, Splunk picks us, we should definitely join forces. It was a very easy decision. And I'm not saying that it was an easy decision because everybody made a lot of money. I really felt like this is an opportunity to build something bigger with another public company in San Francisco. And of course the whole thing managed to move permanently in the United States and do their own thing.
Christina Noren: So you stayed with Splunk for a few years, and then something caused you to go out and start AgentRisk. So tell us that story and lead us into what AgentRisk is about.
Jon Vlachogiannis: Yeah, I had the idea of AgentRisk literally like I would say a couple of weeks after the acquisition. And not because I wanted to leave Splunk, or whatever. But when we got acquired, we were still in Greece and we got the money from the acquisition.
And I went to a wealth manager, like a pretty big company in Europe, and I hated the experience. I didn't like the way they managed, the way of creating my portfolio. They couldn't explain to me how my portfolio was created, which data they used, why these specific assets were used, how much are the fees exactly. And if the person that sits across from me actually has the same portfolio as I am going to have.
You know, it’s like going to buy a Tesla, and Elon Musk is not driving a Tesla, he’s driving a BMW. It’s strange. For me, it didn't resonate. Also on top of that, they tried to make me feel a little bit stupid by using all the jargon. But as I said, I'm an uber geek so I know all the jargon. So I didn't like that I was talking to the salesperson and not with somebody that actually knows what he’s talking about.
And I started writing a very small version of AgentRisk back then, and I talked with my VP that I'm doing this on the side to make sure there is no conflict of interest. Obviously, there was no conflict of interest. One is wealth management and the other was big data. And 3.5 years later our previous investors came and told me, “Come on, man, this actually is really good, you have a lot of people using it. I think it’s time to move to something else.”
And they were right. There is a huge need out there for financial products big, but data scientists that trust models more than opinions, and they are fully transparent in what they’re doing. Literally like having been in this market for almost like 4.5 years, I really wouldn't suggest somebody going randomly and picking a financial advisor.
You need to do a really good due diligence because, you know, you will give them actually your whole wealth. Your future will be in the hands of the guy, and you definitely need to hope that this person is not a salesperson. You need to make sure that he knows what he’s doing.
Christina Noren: Tell us what AgentRisk had grown to be, say, by February when you and I last saw each other before all this, and then tell me what’s happened since.
Jon Vlachogiannis: Ah, my favorite question, my favorite question.
Literally in January we decided to start a new product, a B-to-B product, to offer our services to advisors. We would white label some stuff, they could use our machine learning to optimize the portfolios, we had some huge companies using the product already. So I'm like, guys, that’s it. By February it’s going to be a huge acquisition definitely, easily the biggest acquisition in four years that Silicon Valley has seen, like by far.
We have the best companies, the biggest investment companies out there, like…
Christina Noren: So let me pause you for clarity. At that point you had a B-to-C product that people like me could sign up to and use personally, and you were going well with that. And then I think we were talking about this as well at the time. And then suddenly – well, not so suddenly –
but you said, okay, there’s an opportunity to put these algorithms behind the work of these big wealth advisor retail salespeople and their traders and so forth, and put this intelligence behind there. And if that takes off, then one of these guys is going to try to buy us. That was your logic, right? Okay.
Jon Vlachogiannis: I didn't jump into a B-to-B product because, you know, I love B-to-B products. We had customers from our B-to-C product talking to their advisors and telling them,
“Hey, guys, there is this startup, they have these algorithms, they manage the portfolio, I love the communication, why you are not doing this?” So that’s how we started getting into the B-to-B business without actually doing any sales at all. We literally had a good amount of B-to-B customers using the product without having a B-to-B strategy at all. So we’re like, okay, it seems that they’re willing to pay a lot of money, everything works, maybe we are looking for an acquisition in the future.
Like, literally, this seems that it’s generating a lot of money. February, when February came and Covid-19 was a little bit harsh, I could say we lost every single customer immediately. Like they disappeared. Whenever we’re sending an email, they were out of office. We were calling their offices and they were telling us –
Christina Noren: These were the B-to-B informal wealth advisor customers.
Jon Vlachogiannis: Yes.
Christina Noren: They were going dark on you, is what you’re saying.
Jon Vlachogiannis: Oh yeah, and I'm like, “What happened? How did this happen?” A week later, we started having a lot of new B-to-C customers, because they couldn't reach out to their advisor, because either their advisor was missing in action, I don't know what was happening, or the advisor didn't have access on his home to do trades or whatever, they haven’t set up anything. So literally they had no financial advice.
So we started getting B-to-C customers, all of a sudden like a lot of them. And then as a CEO I had to figure out, what’s happening here? We have the B-to-C that is running and it’s getting a lot of people, but we invested a lot of money in building a B-to-B product. How long this is going to take? Is it going to take a month or is it going to take a couple of years to figure out should we keep investing in this B-to-B product.
We did a meeting with the whole team and my cofounder Alexis, and we decided that it’s better to pause the B-to-B product because we think that this situation could take more than a year to figure out exactly what’s happening. So we’re like, okay, let’s go back to our B-to-C product. Let’s pause any marketing, any outreach we’re doing for the B-to-B, and let’s return back to what we know and we do best.
And this might seem like, it’s okay, nothing happened. But for a small startup, this could have been the final blow. You divert all engineering sources to a B-to-B product, and all of a sudden the product that you’ve been working for some time, goes dark. No customers, and you lost every single customer. It was so bad that if we didn't have enough revenue from our B-to-C it could have shot us down immediately. Oh yeah, good times, good times, this February was very good.
Christina Noren: Let’s get a little bit into the software side of that. So as a small SaaS company, B-to-C SaaS company, suddenly seeing the spike in demand, and needing to retreat from a pivot, how did you need to respond in terms of features and scale and team and the software delivery. What did a day in the life of your startup look like in early mid-March?
Jon Vlachogiannis: We haven’t changed a lot. Our stack is pretty much the same stack I had for the last three startups. We have everything on Google Cloud.
We try to have no DevOps infrastructure. Currently we are using Docker and Kubernetes. Obviously, we have staging environment development and production. Whenever we deploy…whenever we commit something on our GitHub, actually on a big budget, automatically we run the tests, and if the tests pass, they load automatically on staging, and then they’re promoted on production. So everything is running auto-magically, as we say.
Christina Noren: Did you find that you had to accelerate the pace or change the features that were in development, or accelerate the pace of change in some way?
Jon Vlachogiannis: When Covid-19 hit us, after a little bit, after a couple of weeks, we decided that we should stop all new development that’s happening on the B-to-B product. But the B-to-B product was built exactly the same infrastructure as our B-to-C. We are selling a lot of code.
But when it came to data privacy, new web pages, new dashboards, new functionality, all this had to be scrapped and be moved, and no, they are still in a big batch of grants and nobody is starting it. So, yeah, we didn't have to change our infrastructure, but it was a huge blow nonetheless.
Christina Noren: So on the B-to-C side, I'm assuming that all this additional traffic and demand involved some scaling.
Do you feel like just because you’re using such a modern stack that scaling is just not the scary thing it would have been for us in a similar service 15 years ago?
Jon Vlachogiannis: Bugsense was madness. Like we’re managing 110,000 requests per second. We’re managing data coming from 500 million devices. And we could never shut down any servers. It was real time, aggregational data, dashboards, literally like we were powering them by internet. So it was pretty stressful doing that. And AgentRisk where the stock market shuts down at 1:00 PM Los Angeles time, doesn’t work on weekends, starts early in the morning, so it seems way slower, like super slow.
Christina Noren: It’s funny, because when you work with people in financial software and IT, they often act like it’s the most demanding segment because of the high-speed nature of the trading platforms and the exchanges. But to go from there, here you’re happily making more money, serving lots more customers, and your customers are happy and they’ve left the big boys.
One would expect that there’s a lot of software teams at the big banks that are being tasked with trying to replicate what you do. And also, we all know that there’s a lot of IT teams at the big banks that are focused on making it secure and safe and possible for people to do wealth advisory from home, for their advisors to work from home,
because we’re all realizing that it’s very uncertain as to how much we’ll be able to return to any offices. So there’s going to be a lot of pressure on you. What do you see happening over the next 18 months if there’s rolling lockdown through that period? And in the wake of that, once we’ve all had the world experience of this?
Jon Vlachogiannis: I think that financial companies will start acquiring tech startups to build their infrastructure. Having worked with a lot of big investment companies, their infrastructure is like…
No comparison with a proper tech startup like we’re talking now about Docker and Kubernetes, continuous deployment like this doesn’t exist in these companies, especially when it comes to customer-facing products. But, the thing is that even if they manage to excel at building this infrastructure, my main concern is that they always have another motive.
So their incentive is not only for you to have a very good balanced portfolio – it’s to have a portfolio that uses their products, and definitely there are other products, other selections that might be better, or they want to sell you other products. So they’re not 100 percent incentivized to finding the absolute best portfolio for your needs.
So it becomes a question of is it the best technology will win, or the most transparent and most helpful advisor will win? And I show this, when we were doing our B-to-B product, that the interest was not on how to use machine learning to build better portfolio, it was how the machine can figure out the best times to send an email so the customer can deploy more assets.
Because we have exactly the same system of AgentRisk. So, for example, when the market goes down, we can provide data, how long the market will start recovering so we can notify our clients, now it’s a good time to deploy more assets. They were super interested about that because they want more assets in the platform. I understand that. But it was, I would say, 90 percent of this ____, 10 percent of building better portfolios.
And, it’s okay, I understand, it’s a business, but I don't see traditional financial companies dominating the space. I'm super happy that Betterment and Wealthfront literally dominated the retail space for people that have up to, I don't know, $20,000 investments. So I'm really happy they went to that segment and they dominated it.
But the people that have more money and they need portfolios that are a little bit more sophisticated, they’re still underserved, for all the reasons that I mentioned.
Christina Noren: So I think there’s one last question that that begs substantively. So it’s been an extremely uncertain market, and for those of us who are not professional in this world, the question of the initial dips followed by this huge spike in the market and a few corrections, it’s hard to fathom how that can be predicted or understood.
Do you feel like you can claim that your AIs did an objectively better job than the best human wealth advisors on that?
Jon Vlachogiannis: That’s my favorite question. That’s literally my favorite question. I hope we have five hours, because I would go crazy now. I would go crazy. No, I'm kidding. The thing is that – and I’ve heard this from many people saying that I'm not a professional advisor, professional investor –
I could tell you like 100 per cent that professional advisors are not that much different from you. Not specifically you, I mean from everybody. The main difference is that, of a good advisor and somebody that is doing this for a hobby, is that they know how to build a balanced portfolio, something that anybody can learn how to do.
And they are detached from emotions. Because it’s not their money, they’re not going to do something stupid. The most important thing – and that’s where AI excels – the most important thing in financial management is risk management. Because you cannot predict the future. If I could predict the future, trust me, I wouldn’t be doing any startups right now 100 percent. But no one can predict the future. If you look at hedge funds, especially, this year and last year, they got crashed.
They got destroyed, because no one can predict the future; it’s all about risk management. And what a machine could do is it could analyze millions of possible outcomes and how these outcomes fit on the risk profile of the customer. For example, let’s say you are a very risky customer, you have a very risky portfolio, you have a huge appetite for taking crazy risks.
Your portfolio in case of a crash will go down, 100 percent, will definitely go down. But how it recovers, if you can collect tax credits, if you’re going to rebound faster than anybody else, then something that only the algorithms could do it at scale.
If you have an advisor that loves you like, I don't know, like a goddess, and he spends all the time every single day and minute of his waking life managing your portfolio, he could definitely be on par with the best machine learning algos out there. But because we cannot have that, and we want to have more than one customer, you have to use machine learning and algorithms to build and manage portfolios more efficiently. So that’s the whole trick. It’s not like a machine can predict the future or figure out what’s happening, it removes 100 percent emotion, stays to listening to data and not listening to opinions, repeats every single second the same patterns, tries to speak on well proven theories. And at the end, based of course on mathematics and Nobel prizes that people won specifically for these theories, it creates better portfolios, more balanced portfolios than somebody that looks at your portfolio once and that’s it.
Christina Noren: So we’re pretty close to a wrap here, just to net out a few points, and then, Paul, if you want to do the same. A few things that stand out to me is that in so many of the fields we’re talking to what’s happening right now and what’s happened over the last several months and is threatening to happen for these next few years is accelerating changes that were already in progress.
And so your early customers were already going to using your platform because of the quality of the service that they felt they were getting from human advisors. No knock on the humans themselves but just the fact that software does this better. And then suddenly you’re having two situations that accelerate that, which is such uncertainty means that in my case it was harder for my advisor to explain anything when this was happening.
So you feel more lack of trust of a human advisor no matter how great they are. And then separately, a lot of the advisors couldn't be reached by their clients. So the very simple thing that a computer system in the cloud is accessible to clients when humans are not, brought you an influx of customers. And it feels like both of those things will continue.
So it seems like there’s a resilience in your having taken a really pure software approach to a part of the world that has tended to resist a pure software versus human approach.
Jon Vlachogiannis: Yep. And it’s not again that I predicted the future. I'm just a data geek, I love algorithms, and I think that everything that has to do with governance, either we’re talking about lives or financial products or even whole cities, should be done using algorithms.
Well-tested algorithms, of course, that is a whole new discussion. But they perform better. More predictable, easier to test, easier to scale, so yes.
Christina Noren: And then I think the other takeaway is that by building AgentRisk in an extremely modern way with extreme cloud scalability, with automated testing and continuous delivery – and we didn't get into the details of things like feature ____ -- You use that, but you’re using the most modern techniques, and that’s made your platform – basically you didn't have a lot to say about challenges of having this influx of customers, which is not a common thing for a CEO to say. So ____[Audio breaks up], Paul?
Paul: I did notice a few months ago when large wealth management organizations were hard to find, because it turned out they really weren’t set up to work from home, it was that simple.
They were so based on having human beings in the office at set hours. And the other thing is, I was struck… You answered my question, is, there are two issues with using a human wealth manager. One is, do you trust them, and the other is, do they know what they’re doing? And I find it interesting to know that human wealth managers are using AgentRisk. The implication is that at least sometimes it knows better than they do.
I wanted to throw in a story about algorithms. Back in 1987, Black Monday, a year later, Lewis Rukeyser, Wall Street Week on TV, said that he had to admit that the “elves”, as he called his very math-driven analysts, told him that if he had simply held his positions through the Crash of 1987, he would be way ahead now than everybody who was trying to figure out what to sell when for how much.
If he had sat on it, he would have mathematically done better. So that’s where I see that AI has an ability to not be passionate and to not get caught up in peer pressure or group think. And it seems like that’s what you’re offering. So my one question is, the people who are using AgentRisk, but presenting themselves as a human wealth manager, do they let the customer, the client, use AgentRisk, or are they keeping AgentRisk behind the scenes as their own advisor?
Jon Vlachogiannis: It’s behind the scenes. And this is like the standard practice for all advisors, I hope, at least. That when they create a portfolio or they manage it, they’re using an algorithm, they’re using a tool that their company has provided for them. But it’s not like they sit on their desk and they start using paper and pen. So they are using a tool and they don't have to disclose to the customer which tool they use, and it’s okay.
Obviously, everybody’s using tools. But specifically for AgentRisk, we give them the opportunity not only to stress test a portfolio to millions of combinations and finding literally the best portfolio for this customer, we also automated all the communications that needed to be done. For example, a vast majority of advisors is not sending emails to their customers telling them, like,
“Hey, Matthew, this seems like a great day for the stock market, maybe we can put some money there.” They don't connect with the customers. So the machine automates all the communication and says, “Hey, you know what? Today I did some rebalancing because this and this and this and this.” After a couple of weeks, “Hey, today we decided to do some tax loss harvesting to get you some back credits.” And the important thing is that when you have opened the terms of communication, it’s easier to build trust and easier for a customer to deposit more assets.
Because he knows that you’re doing your work, you are there, looking over the portfolio. And for an advisor without a technology, this is impossible, like literally. Even if you have one…
Christina Noren: That’s funny you said…so that’s a whole other avenue we could probably do another whole thing just to highlight that, which is, it’s not just AI for the building of portfolio, it’s intelligent software for all the communications to interface with the humans who own the portfolio.
But I think we can wrap it there. It’s wonderful to talk to you always, Jon, and thank you very much for spending the time with us today, and to our audience, I hope you learned something interesting and useful and that you tune in next week.
Paul: Sure, thank you, Jon.
Jon Vlachogiannis: Thank you, Paul and Christina. Thank you, it was a pleasure as always.
Christina Noren: Okay, bye bye.