Brent and Mark chat with Babu Sivadasan, CEO of Jiffy.ai. Jiffy is an IPA (Intelligent Process Automation) platform that is meant for scale. We get a background on Babu and why he got into the space. We also dive into some of the key features needed to execute IPA. Babu talks to us about leadership qualities for successfully ushering automation through an organization. We also talk about how change affects automation and the timescale lens that Babu looks at automation through.
Interview with Babu Sivadasan CEO of JIFFY.ai
rpa, capability, build, enterprise, automation, understand, scale, technology, wave, process, platform, deploy, applications, jiffy, meaningful, documents, industry, automate, players, problem
Mark Percival, Brent Sanders, Babu Sivadasan
Brent Sanders 00:02
On this episode of the podcast, Mark and I speak with Babu Sivadasan, CEO of Jiffy AI, we talk about the genesis of Jiffy AI, how automation scales? And what sort of timeframe automation actually operates on. Thank you for tuning in. So, Babu, thanks for joining us. Why don't we start out by you know, we'd love to hear some background, tell us about yourself. How did you get into RPA?
Babu Sivadasan 00:30
Yeah, did it, I got into it, because of frustration that I was facing, you know, because we were trying to add my previous firm, we were trying to implement automation in a big way, right now. Because if you want to make meaningful progress, with respect to efficiency, especially at large scale, you know, the existing technologies tend to be lacking that scale. And, and the industry in general, as you know, evolved by automating user clicks and without a lot of intelligence in it to begin with, and then people added more torn capabilities, you know, and that that creates a problem, especially when you're trying to do this at scale, right. And so that's where I thought there was an opportunity, because to make meaningful progress, meaningful improvements to large enterprises, the backbone of the enterprises, those operations, and if you want to leverage, meaningful efficiencies, and bring true autonomous nature to those applications, and processes, it needs to be rethought re-imagined. And, no, that's what we did, we built from the ground up, sort of a next generation enterprise. process, automation, technology built from the ground up, and we had the luxury of building it with AI as the backbone, rather than AI being added on as a bolt on capability. So very excited about that. And that allowed us to solve some problems at some very large organizations and got us going that way.
Brent Sanders 02:22
Interesting. So just for our listeners sake, and for my own is, when you say scale helped me understand we get a grasp, grasp around the numbers or the processes. I mean, it's hard to understand the scale with you know, people could say, Oh, this amount of automation, this amount of documents processed, I mean, help me understand how you think about scale.
Babu Sivadasan 02:43
Yeah, I think about scale, you know, especially my previous organization that that I co founded, we were processing millions of accounts right now, and we were in financial services, you know, and and when you deal with that kind of scale, right, you need flexibility to operate in, in, in large volumes, right. And if you look at traditional RPA, technology, in general, you know, you're trying to simulate one click at a time or one action at a time on a user's desktop. That hits the wall, right. And so assume for a minute that you're trying to do an activity that takes two, three seconds, you know, that's probably, you know, and two, three seconds to complete. And you're trying to do 10 million of those, you know, that takes, and typically you get small windows in, in back office operations and things like that, when you're trying to get something done in 30 minutes, right? There is no way you could do that, you know, right, with that kind of dependency, right? And so you have to go more and more towards the server side, and try and see what you do, right, so that you're not relying too much on the client side scripting and things like that, that's one, second, you're trying to also do this in a centralized manner. So you don't want tiny bots running all over your enterprise, you know, doing things, you know, without any sort of centralized control and command and you want that security and and if you no matter, the management of those bots know that running right muscle, which required a little bit of a different reactor texture, if you will, right. And that's what we did.
Brent Sanders 04:40
Interesting. Interesting. So what I'm hearing and tell me if I have it right, when I'm understanding it, it seems like the patient said the first wave whatever wave we're sort of are cresting and are moving past of RPA. So n number of waves. What I'm hearing is it sounds like the low hanging fruit. has been captured and now kind of moving into the bigger problems is where you're trying to play.
Babu Sivadasan 05:05
Yeah, that's exactly right. And, and it to an extent, that may also make the whatever that first wave or second wave sort of irrelevant as well. Right. And so if you look at how the industry evolved, by the way, I want to highlight that we are not just a simple RPA player, we are, we are a truly end to end next generation. Business Process Automation provider, right, and which is centralized now, which is enterprise wide. It's a truly enterprise class platform that way. So, if you look at the evolution of RPA, for example, like, you know, it started with, you know, users are doing some clicks, repetitive actions, you know, people figured out, hey, there is a bear, maybe I can record that and run that and know that that was that that's how the industry got going right now, where you're looking at your daily activities on on an individual's desktop. And, and you recorded them and played them. And there wasn't a lot of intelligence in that initial stages, you know, he didn't, all it recognized was a bunch of clicks, and at various places on your screen. And then it evolved, started evolving there from there, because you got to have intelligence in other ways in a URI, you're prone to simple changes, and that breaks all the all the scripting that you've done, and things like that, or you're so sensitive to a pig cell, right? position and things like that. And that doesn't have a lot of intelligence. And so that next wave started recognizing UI components, if you will, right, and has not started recognizing what fields are where still without a lot of understanding of the business function that it is performing. And that's the, you know, next wave, you know, which is all about really understanding what the business process is, what is it doing right now, you have an account opening form that you're trying to automatically do data entry through and now go to the next step. You have to understand what that is doing right now that it is an account opening form, and that level of understanding so that you can truly automate functions. And as you take it to the next step, you know, so and, and then if you look at RPA, in general, it's really extending the life of legacy technology, right. And so it doesn't really advance technology and capability to the next next generation. Right now, it is only extending that life is not making any meaningful improvements, meaningful transformation of the process, no, so it's not building for the future. Right. So that's when some of the additional next wave started happening. And looking at maybe adding a little bit of AI component on top of it, in terms of recognizing documents. And, understanding what is in documents. And because if you look at all the data entry and all the user decision making that you do on applications, it's all a function, it's all the result of no machines not being able to understand certain things that we all take for granted. Sure, yeah, we look at a document that we look at a screen and we instantly understand it, right. And whereas the machines haven't matured enough to get there. So that is, there was a tremendous amount of additional investment that went into understanding documents and things like that, especially reasonably structured documents. And next wave, and that trend continues, you know, where you're looking at more unstructured documents and are trying to extract meaning from them and trying to process them effectively. So that more meaningful automation can be done. That's, that's sort of where things are right now. And we felt there was a need for building the platform from the ground up and with core machine learning capabilities, you know, so that you are understanding every bit of the action, and it can scale and then scale is extremely important. You have to be able to combine one activity at a time and get to a point where you want to then do certain activities across a million objects, right. So that's how you get to that scale. So which requires a very different architecture. And that's, that's what nobody had the luxury because we came in pretty late, you know, we had the luxury of observing 15 years of what was going on, right and and learn from that and create the next generation platform.
Brent Sanders 10:16
Interesting. So when you talk about this, this sort of RPA, there's Gen three of scaling across millions. And I'd love to talk about going back to, you know, structured document processing, or, you know, there's going back to the invoice, right, so yeah, it's such a popular thing, it's like if we could just automate getting invoices in and put it into a database, or put it into a common format. And it's sort of this holy grail problem that, you know, Azure, they have some services out that I was able to tinker with, and they were interesting, they, they were helpful, sort of the, the method, before that, I think a lot of people were, you know, will will just kind of understand our vendors. And we'll map a template out of what we expect to get from vendors, I mean, so even before that, completely sub optimal solutions. And then, you know, again, going back for past some of these other services, where you would talk about, okay, we're gonna, we're going to build a bot. And we're going to use a bunch of external services in order to understand our structured data and do something with it and have the context, you're quickly getting into machine learning and training data, and you're using external services and their models and your internal data and building your own models, potentially, in training, what you can, and then you talk about scale. And to me, when I hear scale, I think, especially when you talk about, you know, millions of documents, I think about the amount of time and the cost per page of invoice. And, you know, for example, I'm thinking of like the document processor, or the service Abbey, which is a popular processor, where you can send files, and they're going to charge you per page. And that's an API request. And really, the challenge seems like it becomes how do I run a bunch of complex processes against multiple third party API's? And by better and to me, Mark and I were talking about this recently, it's like, data storage is one problem. It's like you to just use a SQL database. I mean, how do you, you manage that and keep that centralized? And as you said, control, it just opens up a whole bunch of technical questions that the solutions that you I think what I'm saying, what I'm hearing you say is that the solutions that are currently out there are still on point and click and, you know, process capture, and, you know, getting this first wave, and when you start to level up, it quickly escalates to a much different set of problems. And so it sounds like that's what you're going after?
Babu Sivadasan 12:49
Absolutely, you hit the nail right on the head, you know. So if you talked about a centralized database, you know, that's the key, right? And especially kale. What you don't want it as the industry was like, no, it started with an individual bots running on desktop, and then then thought, Okay, well, we need to bring multiple of these together. So how do you communicate among each other, right? And then you put some kind of an orchestration component in the middle. So these are all patched solutions, in my opinion, if you were to build a truly enterprise grade application, you would start from the, from the backbone, how do you build a solid backbone, which means that now you need a big data back end? You know, so we built a big data back-end as a framework. Now, that's where everything runs off of No, so so as you process documents, there are always questions about, you know, what you did, and you you need the ability to go back in time and not take a look at know why you made certain decision and things like that you need, you need all of these to be stored somewhere, right? You need that to be part of your backbone of your platform. So again, maybe, you know, if we were to start 15 years ago, we would have gone through the same paths. And, uh, but we had the luxury of re-imagining and creating from creating a foundation that and by the way, we are all solid enterprise software guys. And so we understand what it takes to build and deploy enterprise grade platforms, you know, the, where we have in the past and have built and deployed, you know, systems know, that process, trillions of dollars worth of assets and things like that. So it comes with a certain level of expertise and you know, and that's what we translated into our software.
Brent Sanders 14:52
Understood. Interesting. Yeah, this is, you know, this gets into this new era, which, you know, we talked about the The industry of RPA of automation in general. And, you know, I think the conclusion as of right now is the, you know, and it's hard to even we every conversation we have around RPA, we have to kind of re label what is RPA mean? And what does that label actually defined as, but as we think about, you know, the generation that we're in now, it's, we think the major players are going to, you know, they run off licence fees, and the fees are going to get more and more competitive and go get lower and lower, and the values me driven outwards to these providers. And so I point to, you know, maybe this is where you guys fit in where Jiffy fits, and where it's, I understand that it's a platform as it stands alone. But you know, if there are particular sort of features that you're good at, like document classification, or things like that, where you can just, you know, use a third party in order to make sort of this using hyper automation as the the key term here where it's like, you have your your bot runner, and that is going to have a license fee. And, you know, right now that costs one thing, but we think in the future, that's going to start to go away. And the real value is going to be in these sort of smarter third party systems. And so eventually, maybe our sort of prediction is, that part is going to be for free to run a bot and maybe ingest information from a webpage and DRP, or, you know, whatever system, but then the real value add is getting that data to another system and getting it back in a format in a way you understand, I think, what I'm hearing Jiffy, the differentiator there is, it's built with that in mind. And with the scale in mind, because if you try to do this, I'm going to say on UiPath, automation anywhere blue prism, Robo cloud, all the players that we've spoken to, you're going to run into this sort of data question and have to start answering pretty difficult questions or your IT team, that also I should add, probably didn't start this initiative, is going to have to step up and start building from sort of the, the outside in versus what you're saying, built from the backbone, the inside out?
Babu Sivadasan 17:05
Yeah, no, it is, it is, it has to be treated just like any other enterprise grade application that, that you're bringing to the enterprise, you know, traditionally, no, that's not how these technologies got into the enterprise, anybody could just simply download and run and have productivity help that or a period of time others got to know about it, and, you know, so you had to slowly start forcing some kind of control and governance structure on pieces of technology that was not built to be governed that way, you know, so. But these are all organizations, they have done a fabulous job in terms of tapping into that wave. And have been very successful, right, and, you know, so nothing against them, we just think, you know, we need a stand for a redesign of the technology and, and that's what we are doing. So we are not just looking at automating an existing process exits an existing set of steps. But we are also trying to combine that with the transformation, meaning, the ability to create something new, right, and also certain time, when you look at a process, end to end is extremely important to look at automation, not as simple task automation, you're just opening up an Excel spreadsheet and doing something, that's what I'm trying to automate. If you look at it that way, you're only getting certain amount of value, you have to look at your end to end processes, you know that that's very complex, you know, but you have to look at it that way. Because ultimately, your real are always going to come from looking at things in a holistic manner. When you look at that, and try and look at every component of it, and you're gonna find that there are things you know, that you really need to retire, you know, and, and so you know, if you can patch them, but you can patch them and you're just extending the life of certain things, you know, that you're better off recreating. And so you need capability for that as well. So, what we have done is created a low code, no code, enterprise application creation framework along with it. And so you can just with ease, you can automate. But you can also replace certain components on that I've seen its life through right now. It needs to be retired, you know, so that's how you transform.
Brent Sanders 19:49
Kind of segues really well with another kind of corollary and it's less to do with RPA IPA automation. And I guess I would want to back up you know, I keep using the term RPA to describe what Jiffy is. But you guys, if you look at your site, it distinctly has no mention of RPA. And it actually is around intelligent process automation. So I want to call that differentiator out to our audience is that you guys are not an RPA. Player, you're an IPA player. And so that's kind of a new term. I mean, do you want to add anything to that definition? Is that a definition that you guys feel like you pioneered? Are there some other players in the space that are also coining that term?
Babu Sivadasan 20:28
I mean there are others coining that term as well, our real innovation is in hyper apps and hyper automation, you know, which, which the industry is catching up to now. And and that's what we started with two years ago, two and a half years ago, when we started building these, we just have this view that, you know, enterprises need a new set of applications that are autonomous, you know, that will, that'll integrate very well. And with other applications, you know, that are vertical focus and a function focused, if you look in the enterprise, and what you notice, is that no, there are capabilities that are really good at, for example, finance, and accounting, and HR, you know, these these players, you know, who really do a good job on their respective areas. And, but enterprises have the need that cuts across those applications and and need need for automation across those applications and, and building new capability that takes advantage of those applications as your digital assets, if you like, no, most of those applications do provide API's and, you know, capability that they expose, that allows you to re-imagine what is your organization designed, as you mentioned earlier, right now, so, and we enable that to happen.
Brent Sanders 21:49
Interesting. One minor question about the business and how it sort of, or I should say, the software and how it's set up? You know, if there's something that for whatever reason your platform didn't do, could I can people plug it into other data sources or services? Is it pretty flexible in the sense that I could, you know, or if I'm coming from a, you know, an RPA? You know, an existing pretty large RPA implementation, we're looking to level up? Do I need to rebuild my bots? Can I run them? Like, how do you guys integrate, I guess, is the short version of the question.
Babu Sivadasan 22:28
I'm glad you asked that question. So we are very big believers in API, ecosystem and API world. And so if we look at their processes that we that's designed on our technology, and if it has, let's say, seven steps, each of those steps are an API that is callable, right? Each of those steps, you know, is something where you can plug in some capability that you're built as a micro service. So it's a micro services based architecture, and you can plug in capability as easily, and doesn't matter what the underlying technology that you've built it in, you know, as long as it can be thought of as a micro service, we can integrate the main to the framework and so pretty extensible, that way we could coexist with there are many other solutions out there. So any legacy investments that you've made in RPA, we could encapsulate that and get it ready for the next wave.
Brent Sanders 23:33
Hmm. Interesting. That's, that's good to know. I'm curious to switching gears to around AI and around the value of training models. You know, you think about the really the recent release of open a eyes GPT three, project with that playground, and it's been all over Twitter, and everyone's going to access into there's some really exciting things it's doing, but the thing that I'm taken with is that, and I think they put this on their site, as you know, they've spent millions of dollars or they say, the cost to train these models, if you were to try to do it yourself would be, you know, I think they had 4.6 million on one particular architecture architecture, to to capture that. So, you know, the the long term play is if you're in this AI world, you're training models, they're, they're, you know, they're becoming smarter, they're becoming more effective, I should say, should say smarter, but as they age, and so I'm wondering, you know, how do you think about the models that you're creating? Or is that, you know, is that thought of part of a business's assets? As you know, as they have, let's say they have a very specific invoicing system and you train a model around that. Is that something that goes into your platform is that go into the specific companies is that shared across accounts? You know, it's just a new topic for me in thinking about software that is creating an asset essentially. And so I'm wondering what your thoughts are on it.
Babu Sivadasan 25:00
Yeah, so, um, you know, we have invested heavily as we build technology and things like that around, you know, whatever GPT has been doing, you know, we, you know, it's great work, you know, and we have, we have invested heavily in similar capability. And we look at it from an enterprise applicability perspective. And as we go deep into certain verticals in where capability that you showcase, and we bring that into the enterprise, and we make it possible within the enterprise, we are delivering for enterprises using those kind of capabilities. You know, we would have been great to leverage some of those work in but when we started, we, we were the only ones experimenting with those kind of things that that, you know, so So we, you know, we have gone too far in terms of investing and creating, what I refer to micro innovations, you know, when you look at capabilities, you know, where you make decisions around how to lay out a series of data on a on a screen, for example, how you make decisions that are out on the fly. So if you look at today's enterprise applications, they all live off of inflexible machine code, right now, that is the backbone, you create all these things, and then you deploy machine code, that doesn't give you the ability to be nimble, right. So you have to recreate if you want to change, you know, and that is very, very, very expensive to do that. And that hasn't fundamentally been changed. Whereas we've taken a slightly different view, and said, Okay, well, the that that knowledge now that we all have you and I have right around an application, how we understand it, that's what is the foundation of the application itself, so that you can easily make changes, and that will render themselves. So we've kind of separated very clearly the business knowledge that drives the applications from the rest of the foundation. So that and we made that to be extremely efficient to transfer this business knowledge or the domain expertise into that core foundation that renders and creates that may dip, render and deploys that application, so that there are a bunch of innovations that we have done around that. And that's what enables us to go deep within an enterprise and deploy capability in real time.
Brent Sanders 27:47
Sure, sure. Switching gears back to, you know, your general RPA experience, let me know, maybe less, so which if you may be more so chippy. You know, walk me through your experience with our pay when when you've been involved in initiatives or seen it work well, you know, what ingredients do you feel like are necessary to ensure success of even you know, whether it's a pilot or to your point scaling to the, you know, the, the million document?
Babu Sivadasan 28:21
Yeah, so, so, um, you know, the way that the industry gone about it is, you have some technology and, and 10 lakhs of consulting dollars to go with it, that's what you need to be successful. Right. And so you invest 100 k on technology, you need a million dollars to derive value from that, you know, so, and that doesn't last long, that kind of an approach, you know, it is it has worked in the past, you know, in the early days, you know, but these days, you know, your people are looking at the total dollars know, that is spent, you know, and and seeing, okay, am I getting value for that whole thing, right? And if you don't do that, so, you have to, you have to start going back to the drawing board and figure out now, how do I deliver real value meaningful value with the license dollars know, that you spent, you know, so that, so, which means that it needs to be user friendly, it needs to be able to, you know, deploy things like natural language as as a way of getting getting things done. Eliminate the cryptic coding, you know, that you have to do it to make things work. Right. So, improvements in a number of areas, you know, within the core RPA framework is what will enable you to address that next wave of, you know, capability being deployed within organizations and so and so that they can derive meaningful value.
Brent Sanders 29:54
Yeah, that's helpful. That kind of answers my next question which was, you know, what, what does that work, right. So that's a good perspective.
Mark Percival 30:02
It would be interesting to get your perspective on where this is all going, especially in the direction of right now a lot of the RPA vendors are focused on and I think and the case of Jiffy as well, things like finance, invoice processing, HR, but obviously, that's a, that's an area that there's been a lot of activity in RPA, where do you kind of see in the next two to five years where the next big piece is going to be?
Babu Sivadasan 30:25
Yeah, that's the next big pieces would be going deep within verticals, now, where you're trying to understand because the horizontal play that cuts across many industries, and where you're trying to make meaningful progress, I believe in it, it has worked, and it does it to a large extent, it doesn't deliver the promise, you know, because in a, you have to be able to go deep, to derive operational efficiency, operational efficiency is a function of what industry you are in what kind of processes you are in, and you have to deeply understand those processes, and you have to bring that understanding back into your platform, you know, just having a foundational technology and deploying that and, and have all kinds of people build things on top of it, that doesn't contribute back into the platform doesn't enable you to deploy success in the next implementation. So you need a lot of that capability, a lot of that work decal knowledge to be baked into the core platform.
Mark Percival 31:35
How do you build an organization like that as a leader?
Babu Sivadasan 31:40
Its painful, I would say that it and, and you need to have a long term view to to make that work, because that is hard work, it takes a long lot of time, because you know, and if you work with in our case, we are we are working operating with the 20 years timeframe in mind, you know, so, we will be here, we will be around. And so, we are not making short term, you know, mistakes, you know, that takes your eye away from that long term focus No. So, it is hard to do that, you know, and try to deliver, you know, explosive growth at the same time, you know, so you have to be measured in your approach of how you go after that. And if you're, if your approach is, you know, that long term eventual goal you want to be around and be a very successful company, you have to look at it with that view. And you'll know that most very, very successful companies have had that long term view and built from the file or built every bit of the foundation, the right way. There is another view, which is, you know, we want to be, you know, build and flip and make it somebody else's problem. Right. And that approach also works. But that's not what we are doing for sure.
Mark Percival 33:10
Yeah, that makes sense. I'm not sure that he sees would be happy. But you know, actually, actually, I think the long term vision for the, for anybody who's invested as obviously the bigger play, because that's what builds the billion dollar companies. Yeah, let's go, it's certainly a different mentality, sometimes on the ground.
Babu Sivadasan 33:29
Exactly, you know, so, but now, you, we all go through these cycles right now. And, you know, that our hype cycles, you know, then followed by, you know, ignored, I mean, being ignored for a while, so, but you have to go through those cycles, and, you know, and, and still not be worried about it, we have to have the long term focus and build, build the foundation build, build every step of building that company, right. And we are very focused on that. And we've been fortunate to have a group of investors, you know, who believe in that, and, you know, have been really, really supportive towards that mission. So we'll be around for a while. That's that.
Mark Percival 34:14
I think, I think the interesting thing about RPA is just how long these companies have been around some of the older ones, people kind of people, especially today, kind of think, oh, RPA is this new thing, but a lot of these companies have been at it for quite a while building these, these platforms.
Babu Sivadasan 34:26
But what I would say is to get meaningful results out of those, you know, you need another number of years before you tap into all of that value, right? And so, and that takes, especially going deep in multiple verticals at the same time and continuing to scale. And make sure you're, you're not building a bunch of custom applique made custom platforms in the process.
Mark Percival 34:57
Yeah, it reminds me sort of that with that deep expertise. Sort of SAP, right? It's that, yeah, having to have that domain expertise and building that platform that's not specific to that vertical, but you also then have to ingest all that domain expertise into your own firm.
Babu Sivadasan 35:10
Yeah, got it, got it. So, so it's a, it's a good balance of that you need to have, and, you know, if you have great enterprise, application, build experience, and things like that, you'll get that right, you know, you will, you will find that good balance between the two. And you need to be patient and continue to work with your clients and continue to take them on that journey with you. Right, and, you know, we have so far we have been fortunate on many fronts, and, and we look at it with the view that we can't, you know, ignore the social impact of all of your actions, either, right. And so, so we've tried to combine all this into an overall structure and framework, you know, that allows us to be able to get there.
Brent Sanders 36:02
You know, is there anything that we're missing anything that you wanted to mention to our listeners, before we wrap up?
Babu Sivadasan 36:10
Yeah, but we covered a lot today. But I would say this, you know, just reiterate this in the investments that we making today, you know, our our should be done with a, with an eye towards some of these capabilities being deriving, providing value for the next 20 years, you know, so that's the kind of investment especially post COVID, you know, we are we are going to be in this war, we have changed our life has changed, you know, our, the way we work, the way we live, you know, has changed. And it requires a redesign of how all of these things are done right now. So, which means to know, the investments have to be made in a very thoughtful manner. Every enterprise investment, you know, in a thoughtful manner, that takes you to not 10 years, next 20 years, you know, how can I advance and build capability that lasts for that long. And you know, that that's a very important set of decisions that enterprises are having to make and, and we were fortunate to be in that right place at the right time to partner with great companies and help them take them through that journey. For the next 20 years.
Brent Sanders 37:37
It's great. It's valuable insight. I like that 20 years scale. It's not a quick win. We're in it for the long term. That's good. Great. Well, Babu, thank you so much for joining us, and I appreciate you taking the time talking to our audience.
Mark Percival 37:50
Babu Sivadasan 37:52
Thank you. Thank you very much.