“We are taking on AI-based innovation with a deep sense of social responsibility, but also combining a for-profit and a nonprofit at the same time. We recognize that what we do disrupts traditional jobs, or it has the potential to. Every AI company is, in a way, dabbling in disruption of the workforce. ”
— Babu Sivadasan, CEO, Jiffy.ai
After 20 years at Envestnet, Babu Sivadasan was ready to start a new company and was looking for the right idea. He realized that there was a need for intelligence software that could help stitch together other applications in datasets and act as a bridge between humans and algorithms. And that’s how his new firm Jiffy.ai was born. I spoke with Babu about the difference between real AI and fake AI, how HyperApps can change the world and a whole lot more on this episode of the Wealth Management Today podcast.
Now hit the Play button!
Topics Covered in this Episode
- How Jiffy.ai Got Started [3:30]
- Real AI vs Alleged AI [7:21]
- Automated vs Autonomous [15:05]
- The Rise of HyperApps [17:30]
- How to Innovate [19:55]
- Fund Raising is Not Fun [23:25]
Complete Episode Transcript:
Craig: It’s had a fantastic day in the wonderful world of wealthtech. Welcome to episode 70 of the wealth management podcast. I’m your host Craig Iskowitz, and I run a consulting firm called Ezra Group. We’re experts in everything related to wealthtech. We deliver growth oriented solutions to banks, broker dealers, asset managers, RIA aggregators, as well as their wealthtech providers, through our premium advice and targeted market research. On this podcast I speak with some of the smartest people in the industry who are on the leading edge of technology and innovation.
I’d like to welcome our guest for this episode of the wealth management, a podcast. It is none other than Babu Sivadasan, co-founder and CEO of Jiffy.ai. Hey Babu. How are you?
Babu: Great. Thank you, Craig. Thank you for having me.
Craig: I’m happy you’re here. We haven’t spoken in a while, I was getting worried about you your head is down, working so hard and building a new company, it’s great to see.
Babu: Yeah. It’s been a great experience just hunkered down and that’s what everybody’s doing, and in the process, creating something great again. So, really excited about doing that.
Craig: Oh, I’m excited to see it. I mean, you and I go way back. I first met you, I think in 2005, 2006, so it’s been impressive to see the growth.
Babu: Quite a while ago, isn’t it?
Craig: Quite a while ago, things have changed a lot since then.
Babu: No, I remember we were such a small company and during those days so, we’ve come a long way at Envestnet up there and now I’m back to being small again.
How Jiffy.ai Got Started
Craig: Back to square one. That’s how you get everything started. So, speaking of square one, let’s talk about Jiffy.ai. Love the concept, love what you guys are doing there. Can you give me the 30 second elevator pitch?
Babu: Yes. So we are a hyper automation platform, with HyperApps. If you look at enterprise applications that were built over the last 30 years they’ve been built with a fundamental assumption that humans are going to operate them, with that when spread in AI technology and things like that. So we are delivering next generation enterprise applications so that our HyperApps that are extremely resilient and extremely autonomous applications for enterprise. So you look up and down the stack and you’re looking at possibilities of making those applications autonomous with the evolution of AI.
Craig: Yeah. I want to get into that, and how you’re doing that and are they really autonomous. But first, you and I met before even Envestnet, I think Envestnet had just been founded or it was merged with the two companies. Envestnet and Net Asset Management back in the day. So, you were in the wealth management space and trading and building applications for advisors. What gave you the idea to build this company, which is a completely different take on things?
Babu: Yeah, so we’ve built great capability for the advisors, but if you look within the enterprise, and if you look at the back office within the enterprise, there is quite a bit of activity that is extremely manual, or things that require a lot of stitching things together. A lot of things that at scale, it tends to be very challenging, right? So you need application of technology, to a lot of that. And, we went on that, I looked around, looked at all the automation technologies that were out there, things that you could do at scale where you handle a lot of documents, a lot of applications so that you have to stitch together to perform end to end business processes.
And thought there was a need for something from the ground up with a foundation built on machine learning that allows you to make a lot of those positions. And so truly bridge the gap that divide between humans and machines, where traditionally you’ve always had to employ mechanisms like data entry, like you have to stitch the applications together where you have to hand it over to people to kind of get to the next step and things like that. Whereas, using machine learning, today’s technology allows you to bridge that gap very effectively. So we built a lot of machine learning capability and things like that that can understand any type of documents that you have to deal with.
And so you had to look at it, read it and then make decisions based on that, and the machines can do that. So our machine learning models do all that, they have cognitive abilities that allow you to understand that any kind of documents or any kind of applications, just like humans. We look at things and we immediately understand things, and we listen to things and we immediately comprehend what it is, and we are bringing that level of comprehension to the machines.
Real AI vs. Alleged AI
Craig: So I always hear a lot of firms say they have AI, there was a big boom in firms adding an appending “AI” to their names. I’m assuming you’re really AI. So what is the difference between firms that say they have AI versus firms that really using AI?
Babu: So AI is broadly used or misused. People look at rules based logic that you use to write an application and then they just call it AI. AI is tied to machine learning in general, right, and is trying to replicate how our brains work. So you’re looking for patterns and you’re trying to make decisions, you’re training yourself based on data and training the system. And then you encounter things that you haven’t encountered before. A rules based system will immediately fail because you haven’t coded for that. Whereas a machine learning system will approximate and try to make a decision based on that in just the same way our brain works. And a lot of our capability when we say we understand the screen, or we understand the document, what we really mean is that we have trained the system for understanding or familiarized the platform with our technology with a lot of datasets.
And then you’re able to understand that, to make decisions based on something that you haven’t seen before. Right. So for example, we have an invoice processing HyperApp where we deal with thousands of vendors, but we haven’t trained the system for thousands of vendors, we have trained the system for a select number of vendors and then our machine learning algorithms and the understanding that it gets from that training is applied for future things that you haven’t seen before. So it is core capability, core applications of machine learning in those scenarios. So similarly, like when you look at applications, and you see screens, you’ve trained yourself to recognize those screens and that leads you to understand using a certain set of screens. And now you see a new screen and we are able to understand it. So what that allows us to do is integrate with applications more seamlessly. Even when APIs are not available, we are able to go in and out of applications.
Craig: So let’s move on to the HyperApp. Again, another buzz word hyper, right. A lot of people append the word “hyper” to things. So when you say HyperApp, what do you mean by that?
Babu: Okay, so you, you have standard enterprise applications that exists today, right. You know, so they perform functions whether it’s accounting, whether it is invoice processing or purchase order processing, a lot of functional things that happen within enterprises that our enterprise applications of today. The only problem with those applications is those applications are built with the fundamental assumption that humans are going to operate them. So, in other words to get an invoice, for example, into the system, a human has to key that in. Let’s say a vendor sends an invoice to somebody supposed to go in and input all that into the system. So on an organization spends anywhere between $5 – $20 for an invoice, for example, and then the same can be applied to what I do, whether it is opening accounts or whether it is a lot of other activities that are fundamental to keeping your business running.
So that’s what is there today. When we talk about a HyperApp it’s autonomous application. So what that means is the same function map, same underlying enterprise applications maybe, but you put a layer of technology on top it where you eliminate that. So where the invoices or any type of documents that are coming in, you’re automatically able to feed that in so that there is no human involvement in the process. And if you look at HyperApps and those are typically applications we only see when it gets to the enterprise back office, and there are functions that happened there. But then there is a trend and component to it, which is engaging with the consumer. And when we talk about a HyperApp, that whole stack, extremely engaging client experience, consumer experience and then autonomous behavior on the backend, middle office, back office where all of that information comes right in, and it goes right through all the way to the backend platforms.
Craig: So now we know what a HyperApp is, and we know one of the most popular apps that Jiffy.ai sells is the invoice processing. Can we talk about something that’s a little more related to financial services? Although of course, financial services has to invoicing, but it’s more of a backend service. On the front end, or maybe this is more middle office, you have automate KYC process or HyperApp. That seems like every bank and every broker dealer could use that. Is that true?
Babu: Absolutely. So there what we’re talking about is every account that you open right, so there is a KYC requirement. You need to collect a lot of information from the consumer, from the investor. So make that process extremely engaging where you don’t necessarily have to get it all in one shot, you can get it in multiple phases but, get that process to be extremely engaging. It can be multi omni-channel right, multiple ways of engaging with the consumer to get that information. And all the supporting documents, whether it is your supporting financial documents and things like that, all of that gets sent over and it is automatically processed. So that’s what we refer to when we say HyperApp. So where all those documents come in and there’s very little human touch involved in processing them. And then an account gets established all the way to the custodian.
Craig: Excellent. I was reading that there’s a validation engine, is that shared amongst the HyperApps or is it only for certain ones?
Babu: No, there are validation engines so for example, if you have an account opening form, and then you write on it, let’s say you make corrections on it you fill it, but then, you met with the client and they made corrections on it or had written on it. And so being able to read all that, automatically extract and make sure all of that gets fed back into the core systems and all that happens automatically.
Automated vs. Autonomous
Craig: So there’s a difference between automated and autonomous. How do you make these apps autonomous meaning they don’t need any human direction?
Babu: Yeah. So initially there is human direction, right? But then the system is constantly learning, constantly understanding. So we have something called “human in the loop” and where the system isn’t comfortable enough to make a certain decision and then human gets involved and takes that action. But the system behind the scenes is learning that so next time say a similar action happens, it’ll automatically learn.
Now what we mean by resiliency is being able to understand and apply certain, logic like for example, if you’re talking about documents and things like that, sometimes you’re doing an OCR, and you’re not able to distinguish between a letter “I”, and the number “1”. It looks very similar, right? But being able to apply context, Hey, there is this field, this could possibly have this these set of values, and being able to apply that kind of understanding and self-correct mistakes in the underlying technology, like an OCR or things like that, that makes it resilient.
The Rise of HyperApps
Craig: I was reading part of your website and I was interested in your Innovate and your Cogeneration tools, which appear to be building blocks that anybody could use to build a code. Are you building your own HyperApps out of these building blocks, or is that a separate system totally?
Babu: No, so earlier I talked about a very engaging client experience, right. And so, when you talk about that autonomous behavior, you don’t need any kind of friend. But then, the way they define the HyperApp, there is a friend and engagement at the consumer level, at the advisor level and things like that. And you do need to build applications quickly. So being able to just simply define what you’re looking for and let the automatic technology automatically create those functional components that you’re looking for, that becomes a component of this end to end HyperApp, is what we are talking about there. So we have technology that can take in simple commands or instructions or functional specifications, and then work that into applications, right? And so you blend automatic with an engagement with the consumer or anyone that you have to engage with. So you can nicely blend these two experiences and that makes it a HyperApp.
Craig: And the other part want to talk about the HyperApp was the configurable workflows. A lot of workflow engines out where a lot of workflow tools, how are Jiffy’s configurable workflows different?
Babu: We look at workflow from an automation perspective. There are workflows out there, more capability out there, but that’s all built for again, going back to whether somebody has to act on it. There is no autonomous behavior built in. That’s where that’s what we do.
How to Innovate
Craig: Interesting. Moving on. So we talked about HyperApps. You have another product that you call Innovate, this is the one I think is really valuable, and I’m really interested in this, being a former programmer. You’re re-imagining software development and having the system generate its own code. How are you doing that, and how is that better than a real programmer doing the code?
Babu: You know, where this is really different, the value add is in every line of code that we write, we introduce the possibility of bots. So as we speak right now, there are thousands of people out there writing, introducing bots into the software. Because we bring new people in all the time and repeating the same set of mistakes and the same things we have solved before, or you don’t know that you have a piece of functionality that is already in there, but you’re still building it because you don’t know that a new person coming into the team doesn’t know about it. So software development, especially enterprise application development, is fundamentally broken that way.
Whereas if we were to let systems do that, they have that digital blueprint of the entire ecosystem and their application that digital twin, if you will, right. So it already knows. And if you are looking for a new piece of functionality, it’ll automatically bring up that that already exists, that you may have already created. Being able to write self tested code automatically. So you’re taking this to a paradigm, a much higher level where you’re talking in terms of the needs that you’re trying to solve for and let machines interpret that. So we have capability and natural language understanding so you’re able to read those instructions in plain English.
And why do we have to write code? For the machines to understand your own way of expressing things, you had to learn machine’s way of doing things, right? And now we are talking about how machines have gotten to that stage where we can with technology, enable them to think like us. Or enable them to interact and take things that are native to us and have them be interpreted rather than the other way around, which is supposed to be the way, it is what we’ve always wanted. And it’s finally getting to be a reality,
Craig: But isn’t it true that the AI is just simulating human thought? Or is it really having his own thoughts?
Babu: Yeah, no, it’s really having a series of small innovations that you put them together and that’s what contributes to that capability.
Fund Raising is Not Fun
Craig: Hmm. Let’s talk a bit about, well, this year, you guys raised $18 million in a Series A, so congratulations.
Babu: Thank you.
Craig: You’ve done a lot with a company in a short amount of time where, how has this new latest funding raise going to help the firm?
Babu: Yeah. So what we do takes a lot of capital. We are, putting out groundbreaking technology, you know, that takes a lot of effort and a lot of engineering, a lot of investment that is necessary to do that. And we have been fortunate to get that first round of capital that was essential to invest and build a team out, a really strong technology team, technology capability and things like that. So we’ve done that, we are serving clients today. We have, 30 plus clients and 10 Fortune companies as clients. And so we are just getting our technology out there and customers start leveraging the value of it, and, getting to that next step of the evolution of the organization. And we raise capital from outstanding venture capital firms and public company executives and people who believe in what we’re doing.
And so the way we are doing this we are taking on AI-based innovation, with a deep sense of social responsibility. And so we are a combination of a for-profit and a nonprofit at the same time. We understand and that recognize that what we do disrupts traditional jobs as we know it at least it has the potential to. Any AI company is in a way dabbling into disrupting the workforce as we know it. And it’s been a good way to combine the two, so on one end you can be disruptive, but then on the other end, you can also be compassionate about it. We can help invest in those people who might be potentially affected, help them retrain for the next set of jobs so it sets up a very humbling experience that way to kind of try and do both at the same time.
Craig: I want to remind people who are listening, if you are working for an enterprise wealth management firm or any enterprise, you should look up Jiffy.ai, and look into their HyperApps and innovative software development tools and testing automation, we didn’t even get to that at all. Data analytics, so much stuff here you guys are doing. We’re about out of time, Babu, thank you so much for being on the program. It was great to chat with you again. I’m glad we were able to connect after all this time away.
Babu: Yeah. Thank you for having me. And let’s get together in person next time so I can walk you through, I can show you some of the cool things that we’re doing and things like that. So it’s great to connect with you again, Craig.
Craig: We’ll definitely put that on the calendar for when we can all get together in person.
Babu: All right, sounds good. Thank you.