The potential of computing technology is so vast, we will continuously realize new possibilities that change the way we work, travel, play, and create. But we should not lose sight of the computer’s most central purpose—to make life easier.
So why do they often make life so much harder? One area that’s especially pernicious is document management. Computers are great at creating documents—so much so, in fact, you probably have hundreds-of-thousands if not millions of documents on your hard drive right now.
But processing the data in those documents can take enough manual clicks to give you carpal tunnel. Ike Kavas, the founder and CEO of Ephesoft, believes we should be replacing manual with automated wherever possible.
On this edition of UpTech Report, Ike tells us how his startup is automating document processing, and walks us through some of the major technological challenges the solution faces.
More information: https://ephesoft.com/
TRANSCRIPTION
DISCLAIMER: Below is an AI generated transcript. There could be a few typos but it should be at least 90% accurate. Watch video or listen to the podcast for the full experience!
Ike Kavas 0:00
I am a big AI machine learning advocate. Right? I think it is the future. I absolutely believe in it. I think it’s the new, you know, programming language.
Alexander Ferguson 0:16
Welcome to UpTech Report. This is our applied tech series UpTech Report is sponsored by TeraLeap. Learn how to leverage the power of video at teraleap.io. Today, I am very excited to be joined by my guests. Ike Kavas, who’s based in the LA area, and he’s the CEO and founder of Ephesoft. Good to have you on man.
Ike Kavas 0:34
Thanks for having me.
Alexander Ferguson 0:35
Now Ephesoft is you guys provide intelligent document processing solutions? I got that right. And actually on your site, it says transform your operations with automation, everyone loves some good automation. But how many understand in a nutshell, what was the problem that you set out to solve? And, and I’ve really, I’ve been focusing on these over these years.
Ike Kavas 0:55
Sure. You know, when you say automation, our our job, me and my, my my friends in the industry, we’re all trying to automate stuff so that we can free people from mundane tasks. And when it comes to FSR, we want to focus on business process that has documents in it, PDF files, paper files, you name it. So we said okay, what are people doing with their muscles that, you know, we can put AI in place so that, you know, they can free to do use their use their brains. So when we focused on it was all manual data entry, and looking at different screens trying to make up some ideas. For example, I’ll give you an idea. A mortgage processing in the United States might have a lot of different document types. Sure, I’m sure you’ve dealt with it. Let’s let’s have software, sorted through those documents, what’s in that in there rather than entering it. And, and then when you go to next step, you know, you might have different tabs, open your browser on your browser, because you’re getting the data from document, maybe another system from the internet from another system that you had 20 years ago, and we call swivel chair, if you’re looking different screens, we can actually combine all that and then bring the contextual data, enrich the data, they need to look at and make a decision. So that things that take hours, we can help them do within seconds. That’s
Alexander Ferguson 2:34
our job. Everyone wants to be able to not have to do manual labor where possible and save time and be more efficient. We’re going in an area where technology can help in that have mentioned that there’s a couple different terms and solutions out there. There’s also like robotic, or process automation. Yeah, so help me understand like the difference than in both the different types of automation.
Ike Kavas 2:58
Sure. So there are generally about four different categories. One of them is RPA, like you mentioned, RPA is is about automating your screen, your mouse, your your keyboard, if if the person is doing these things constantly, on on the keyboard, mouse and monitor, you can actually program it, to do it for you. So it’s like, you know, Excel macro in Excel, if you remember those days, right. So of course, it’s got sophisticated, the more and more what that you can build a bot that will that can do some of the things you do on a mundane basis. That’s RPA then there’s I pass, I pass is about connections, you know, if you can connect the, this software to the internet, or a different vendors, portal, or or, or maybe if you have five different systems that you have in the past, connect that then there, the iPad says you don’t need humans or RPA we can bypass all the connected us so there is no need to program. That’s right. And then there is the the workflow automation systems where, you know, some people still do their work on on their work desks, right? If you elevate that to an electronic environment, you can start managing your cases, electronically, you can look at them, right. It’s a workflow management system, and then you can route it approvals. So basically run your business in a digital format everywhere where possible, that’s workflow. And then as IDP, you know, we we look at wherever there’s a ton of document, and we automate all that and then we can give our data to any of those three friends So that they can do their job, whatever they do a little better. So we’re kind of helper for all of those parties.
Alexander Ferguson 5:08
It’s like there’s there’s there’s multiple ways that automation can happen. And it’s a we’re going to area where, if anything, it there needs to be interrupted interoperability, however you say that the interconnections between the services so that eventually there’s, there’s, there’s less and less that we have to manually go over here, click here over that document, enter anything, it shouldn’t be able to be automated. So let me ask you just a straight more leading and straightforward question. How much can we automate? Like, where are we headed? And how much can we truly automate? And in business,
Ike Kavas 5:40
we can automate a lot. But the technology is not there yet, as well. So I am a big a AI machine learning advocate. Right? I think it is the future. I absolutely believe in it. I think it’s the new, you know, programming language. Alright. But we have to realize we are still in the baby steps, right? We are still teaching AI, how to how humans do few things in a few seconds, maybe a few minutes, not the entire hour or day of work. And, you know, either deaf or soft. I have a theory, right? That’s what we’re working on. We believe in that bringing this contextual data from all parties will help humans start making decisions. But that’s the basis of AI’s improvement. In the coming years, we should be able to start automating a lot faster, a lot better. So that we can not only just elevate the humans right, then just be a little bit more productive. We can automate the whole processes within the company, so that, you know, the company can be more productive now, not just the interview individuals. So that’s what I’m working. I’m very optimistic about it. But I also caution my clients that when they hear AI, we’re still in the in the early stages of what we do.
Alexander Ferguson 7:09
Be careful of taking science fiction, and assuming it’s reality right away. We’re getting closer. We’re getting closer to that. Yes, but but it’s not, not quite there. So give me some a use case examples for specifically for you guys for when it comes to documents, being able to process what are some case studies of customers that you’re working with ever in play?
Ike Kavas 7:31
I’ll, I’ll tell you a few examples. You know, you’ll be amazed how, how many enterprises need this. Our software. As I mentioned before, we have a lot of mortgage companies, maybe a third of the mortgage companies, we probably our software probably saw saw the the mortgage documents in some form or fashion,
Alexander Ferguson 7:55
that that much of the industry is relying on you as a backbone.
Ike Kavas 7:58
Yes. And there are so many things that there’s originators there to refinance, but we’re generally in the insurance industry insurance side of it. So all of those guys, even though they opened the mortgage, they created, they got an insurance, and then we’re very strongly mortgage insurance side. In the United States mentality, this is tax season, IRS handles all of their correspondence with us. So if you have back and forth relationships, you know, we’re the ones who look at the documents and decide which department which agents should this ship, they should
Alexander Ferguson 8:35
go by, if I were to email the IRS a document, it would route it to the right person to have that document.
Ike Kavas 8:41
You’re almost correct. IRS doesn’t do emails much to you in the mail. But
Alexander Ferguson 8:45
yes, that is correct. Good old mail. All right. Yes.
Ike Kavas 8:50
In in Europe, you know, we help EU citizens to get passports. So so when you want to renew your passport, or get your passport, you have to provide a bunch of other information, not only your old IDs and passwords, but also you know, utility bills and where you live. And so our software can actually detect all of those documents and sort them, and then help them verify. And I’ll tell you one more information come back in in Singapore, you know, if you’re a large manufacturer trying to ship some containers through the right canal, so of course, not the wrong ones that we help the largest bank in in Southeast Asia with their trade financing, because all the documents that needs to be accompanied the certificate of origins, the insurance, the purchase orders has to be validated before the money can be transferred from us to Singapore. Right.
Alexander Ferguson 9:48
So it’s what’s what’s interesting is in one way, it’s very simple what you do. It’s like very core, it’s like Alright, here’s a document and then you’re able to take out the document. What’s the information Then give it to the right people. And that’s it. That’s all you do. But at the same time, there’s a lot to that I can imagine, because they’re throwing at you all types of documents that you don’t, you don’t have to you don’t make people say, Alright, here’s the form, and the layout that you have to submit your documents in. And that’s how we’ll be able to read it. Is that where your your sweet sauce comes into why people rely on us? Because we can send all sorts of doc types of documents? So
Ike Kavas 10:26
you’re, you’re 50%? Correct. Okay, I’m the 50% of the challenge. There’s all kinds of customers, all kinds of documents, all kinds of variations, and the different use cases, and everybody runs it differently, right? That their businesses, so we developed our software in a way that we can use AI to extract data from the common document types, we can build more if there if we can collect the data on those. And, and also, like, make make it easy for our software to go live so that it doesn’t take months to get the value. Right? So you’re correct, that’s, that’s 50% of it. The other 50% just extracting the data, and giving to another system is not enough. And because because those systems or or people needs to make a decision, and just the data coming from document is not is not the whole 360 view of documents. So our job also tried to find tangential data that is related to the document,
Alexander Ferguson 11:36
understand what is what is 10 to 10 gentle data.
Ike Kavas 11:39
Sure. So I’ll give you an example. Somebody sent your company an invoice, and there’s, you know, 10 pages of line items, right? And they were supposed to, you know, one of them is the, say a pen, and they were supposed to charge you $1. But they charge one and a half dollars. So how do you catch that, because it’s our job to go to the purchase order system, the agreements, and find that what all those items were what we ordered, and put that side by side to the to the user so that they can say, wait a minute, I’m being cheated here. You know, this is supposed to be $1. Right? So and there are also other cases that you can go to internet get the interest rate security information on that company sending you the invoice, maybe I never asked for that invoice. Right, who is this company and you know that this tangential information can can go on based on what customers want to do. But without it, it will be more error prone to run the business. You know, it did so getting the data accurately from the document is only 50%. To be able to run a productive company, if that makes sense.
Alexander Ferguson 12:59
For people to be able to make the right decision and for business to flow properly. It’s all about having the right data, the right information all together in one place. So someone can make the right decision because I find interesting you say it doesn’t necessarily tell the user, hey, this line item is wrong. It’s just supplying the right information that is so the user can make the distinction and say, it still still requires critical thinking from the user.
Ike Kavas 13:26
Correct. So to get to your other question, how much we can automate. Today, we use all kinds of AI technology to get the right data of the document, right. And without AI that would be very difficult, especially if you have millions of different document types and variations. That’s where the AI is today. But but in the future, like if we’re gonna if we’re going to make more company more productive, right, we need this tangential information to be provided to AI so the AI can actually can learn on business aspects. And that’s that’s making us make more decisions. So we’re getting there. But you know,
Alexander Ferguson 14:07
that’s the next level up that’s the event that the autonomous car version of documentation of
Ike Kavas 14:15
documentation right now we just can see the lanes and the people
Alexander Ferguson 14:19
it’s more like a driver assistance like it’s alerting so right now in in in identity documentation processing or your that’s where you are but you’re headed towards full automation AI that’s what I find fascinates every industry is in different states of progress using AI and machine learning. And it’s good to know where we are at the moment. But it’s it’s getting closer. The thing about I’m curious to kind of move conversational a bit towards this of where where it is failing, like when can you trust the system and know alright, AI is smart enough to take a certain area and when does it when does it work and when doesn’t it and how do you actually Is it is around just like the right data? How can people have a business leader have confidence in a solution in today’s environment, when they’re not even sure how the technology works? There, there
Ike Kavas 15:10
are several ways to do this. And I’m trying to put myself in my, in my customers shoes, right? If If, if customer a has a goal of, I’m gonna bring an AI technology here, and I’m gonna automate, you know, 80% of my business, they might have a different view, they might bring a company who’s gonna collect train AI systems for the next two years, and then make that a reality, which is good, you know, that was customers his goal, and then they can accomplish that the customer B comes and say, You know what, I want AI assistant, I want to get my feet wet, I want to get started. But my primary goal is to start getting return on my investment within two months. They may not want to bring 50 data scientists to do something, they want you to come in and say, Okay, what have you built so far with your AI that can get me started tomorrow? And then we can start building on it over time. Right? So the to answer your question, you you have the my customer has to have an idea of what they want to accomplish, what’s the outcome? They want from this phase one project, right? And then and then bring the right tools, right partners to accomplish that.
Alexander Ferguson 16:37
So I wish I could tell you there is one answer, but it depends on is that it builds for your business specifically, is it you have basically a base technology and ability. But for each customer, you’re developing a more custom solution. Oh, this is what you need. This is the problem you have, here’s our basic knowledge, and then you’re building it out, or is it a standard process? They just come in? Which way do you work?
Ike Kavas 17:01
We have we have both different needs. For example, in accounts payable, I think, you know, last count, we’ve trained on 350 100,000 invoices of different vendors. And that’s check, right? Done. We don’t have invoices, that yes, we don’t have to build that for every customer. You know, we can get them up and start tomorrow. But if an energy company comes in tomorrow, and say, I have this problem, you’re never going to see anything like this on Earth. Right? That may take more than an hour, huh.
Alexander Ferguson 17:44
But you still have the right team and people that you could build that type of identity document identification process. Yes. For for the roadmap of where you guys are headed, what are you most excited about anything that you can can share as far as on features or just where you’re headed?
Ike Kavas 18:01
I think you know, along the lines of what I said earlier, we believe in if I, if we bring that the tangential information to the business, then the users are the decision makers, the knowledge workers will start to make a little bit more faster decisions. But the key is to start training the AI so as the as the humans are looking at and then making decisions, AI can listening right, and then turn those actions into into more AI based decision making so that we can actually automate more. So that’s, that’s where we’re heading. And then every technology we’re, we’re we’re looking to accomplish our goals.
Alexander Ferguson 18:45
I love the idea of building AI and then AI on top of it, to be able to run even more and that’s the future that we’re headed towards. Right, exactly. Now, for those that want to learn more about ISIS stories stick around for part two of our discussion, we’re angry asking a lot more about Thank you so much. So for sharing some of the insights for those who want to learn more about specifically your service and option what’s a good first step for them to take?
Ike Kavas 19:14
Give us a call. We’ll help you. I mean, it’s it’s you know, there’s there’s so much information on the on the internet on the internet, for sure. Right? And then we try to educate our customers as well from our channels. But even and it’s okay, we’re in this business, we’ve been doing this so give us a call what I mean that email or chat, whatever. We’ll we’ll we’ll help you guys and we’ll tell you how we see the world.
Alexander Ferguson 19:40
Apparently IRS trusts them and a lot of other places all around the world. So it’s one place to be able to check out let’s say you can get a free trial, go to Ephesoft.com, EPHESOFT. And also stick around for part two, go to Uptechreport.com and hear our next part of discussion. We’ll see you guys on the next episode of UpTech Report. That concludes the audio version of this app. Episode. To see the original and more visit our UpTech Report YouTube channel. If you know a tech company we should interview you can nominate them at UpTech report.com. Or if you just prefer to listen, make sure you subscribe to this series on Apple podcasts, Spotify or your favorite podcasting app.
