Latest / Entrepreneur Odyssey / AI In Fraud detection & Finance
Transcript
- Tejneet Selhi: Hello everybody, welcome to this edition of Entrepreneur Odyssey. And today I'm delighted to be joined by the lovely Barbara Steinberg, who is the founder of eFord Services. Welcome to the podcast, Barbara. Barbara Steinberg: Thank you, Tejnees. Pleasure to be here. Tejneet Selhi: Great to have you on. And Barbara, you've got a really interesting career, a really varied career that's led you to your current business. ⁓ you mind telling me in the audience ⁓ led you to creating eFord services? Barbara Steinberg: It's kind of a long journey. I'll try to shorten it. My journey extends longer than probably everybody, longer than the lifespan of everybody on this podcast. But I started out in the computer industry many, many years ago. Briefly, I had an undergraduate degree in medical computer science from like 50 years ago. I have a master's, I was studying biomedical engineering, got a master's in electrical engineering and got an MBA along the line. ⁓ not relevant. I started out, I was a programmer and then built systems. I was a manager from there. I went into consulting and eventually into sales. So ⁓ over my career ⁓ you know, I've much done it all in terms of technology and I've seen obviously a lot of changes in technology as a segue in my ⁓ experiences and applications. I was in healthcare for a long time. I was in manufacturing. for a long time. I always had it feel like I was doing something, you know, for somebody. I, you know, doing what I was doing. Anyway, 1990 around and decided there was nothing interesting happening in the computer industry. So I went into financial services. So I did that for a while. I actually had a business. I've had a few businesses educating employees about, you know, their benefits and stock markets and things like that. Fast forward around 2000.com era, I was back in software sales. After that, and I'm not sure what the segue is, I had a business for a long time, Eldercare Financial Planning. And then I'm getting to e-fraud from this. And I helped the elderly pay for care, assisted living, nursing home, home care, ⁓ assistance from veterans benefits, as well as eventually Medicaid, which is how we get the e-fraud. When we're doing Medicaid applications in the U.S., for those of you who are not aware, it's a program for the elderly or for the impoverished who need long-term care or other health care and can't afford it. So the government pays for it or subsidizes it. But when it comes to nursing homes and things like that, they're very critical in terms of applications. and scrutinizing people, they have what's called a five-year look back, which means they look at everybody's bank accounts, opened or closed for the past five years before the application. And what they're looking for is fraud. Did you give away your money to your children and now you're turning around and saying, you know, government paid for my nursing home care at $15,000 a month. So they're scrutinizing all that. being a tech person, ⁓ looked at my employees and their copying from paper bank statements into giant Excel spreadsheets. This must be a way to automate this. So I found a company ⁓ in Scotland who ⁓ had for doing this. And I'll explain the technological challenge as well. so we were using it. And then what happened around that time ⁓ Tejneet Selhi: Good. Barbara Steinberg: 2014 or so I moved to Florida. I sold my business in New Jersey I became their US reseller. Now ⁓ explain a little bit about banks and by the way, their business was fraud detection. ⁓ was the crossover. ⁓ Tejneet Selhi: okay, that's the key thing. ⁓ brilliant. Barbara Steinberg: Yeah, so the software tool worked for bank statements. just happened to be a different industry. Now, here's the challenge. In the US, we still have about 4,000 banks. 10 years ago, we probably had 10,000 banks. Every one of them has a different format. I mean, it's very different than Europe. and very, very different. the challenge is, ⁓ when they do these cases, the same thing happens in fraud. The crime probably happened a couple of years ago. The investigator is looking back five years before that. And if you have a big fraud case, you might have 40, 50 accounts, bank accounts, credit card accounts, brokerage statements, and things like that. And there's money moving all over the place between them. So the questions are, who is the money coming from and going to? Where is it coming from and going to? amounts of data. So when it comes to the bank statements or the financial documents, you just don't download it from the bank anymore. You can download PDFs and things like that, but a lot of this is paper documents that are being scanned. The majority of it are paper documents that are being scanned. So we're not looking at data, we're looking at images. Tejneet Selhi: Yeah. Barbara Steinberg: So the challenge for technology is to extract the data from those images. And now we're looking at scanned documents. Some of them, you know, they're cockeyed. I just have one, it was the worst bank statement ever, sort of my life. I mean, it didn't have tables or anything. It things all over the place. But the other challenge is the format, because the format's different on every one. How do you write, you know, a code that says, okay, look at this over and over again? Tejneet Selhi: watch making. Barbara Steinberg: Okay, so it begins the journey of e-frog. That five, six years ago. The software that we were using was very manual. It could do any bank statement, which was the beauty of it, but the downside was it had to draw a box with your mouse on your screen around every table on every page. So the credits, the debits, and some of the statements were hundreds of pages. So was very manual. It was a lot better than doing it totally manually. And it worked, but like five, six years ago, it like, is not cutting anymore. People don't want to do a pop system. So AI. Tejneet Selhi: you Barbara Steinberg: I said, OK, I don't know anything about AI, but AI must be able to do what I'm doing manually. So we started the journey with machine learning. And years later, we could not get machine learning to give us the accuracy ⁓ that needed. Because with these these people are going to court, and they're testifying that the data is 100 % accurate. So having an accurate is very important and having to do lot of manual work to make it accurate is not a good thing. So what happened with machine learning was we couldn't get the accuracy. ⁓ Tejneet Selhi: Mm-hmm. Yeah. Barbara Steinberg: And one of the reasons is the difference in the formats. I like to make the analogy. If I give a computer 100 invoices or 200 invoices on 201, it says, yeah, I know that's an invoice. Here are the keywords. Here's the data. I give it 200 bank statements and 201, says, I don't have a clue. So that was the difference in the uniqueness of the bank statement challenge. Tejneet Selhi: Mm-hmm. Right, yeah. Barbara Steinberg: Anyway, fast forward a little over a year ago, somebody says, it'll never work. He says, Gen AI. He said, two weeks ago, you couldn't do this, but I want you to try this. Sure enough, that was the game changer. Tejneet Selhi: Amazing, amazing. So, wow. just say, first all, what a varied and interesting career that you have. ⁓ Really, interesting career. ⁓ And eFord it sounds like such an important tool because like you say, having to look at... Barbara Steinberg: So, bye. Thank you. Tejneet Selhi: different bank statements, maybe up to 50 from multiple banks in different formats is a real challenge. ⁓ And I'd love to hear a little bit more about the technology behind it. And so you leverage Gen.ai and how has that changed the game for you? Barbara Steinberg: Yes. Sure. And I'm the prompt engineer. And I hadn't programmed in a million years. We had all these developers and stuff. For five years, we went through a ton of developers. None of them could get it to work. And so I was the prompt engineer. so knowing bank statements, I'm probably the world's foremost expert in bank statement formats. I was able to get it to do what I wanted it to. So the challenge is if you think about AI, AI wants to give you everything in the world. And here I was forcing it, no, you have to do the same thing over and over again every single time, and it has to be accurate. So that was a real challenge. OK, so this forward, so now the AI is working, the prompts are working, but the developers we had couldn't get the rest of it to work, the APIs and everything else up and down. Okay, best forward to June of last year. I said, okay, I've stayed away from the technology because I don't know Java. don't know this Python. I don't know this. said, time to look at it. So I started looking at it. I said, something's wrong because what would happen is, you know, we get somebody to try it out and they put in a bank statement, got put in 10 bank statements, 10 months. Some them some failed. Why? Each statement had over 300 transactions. in one monthly statement. So it just, the system wasn't built to handle that. And one of the other challenges we have with Gen.ai is we have tons and tons of data and it's not built to handle tons of data coming in or tons of data coming out. So we had to do a lot to make that work. Okay, so I start looking at the code and I was mortified. And I'm using Gen.ai by the way, again, I don't know the code or anything, but using Gen.ai, I'm able to figure out what's going on. on and make changes and update things and stuff like that. So over the course of six, eight months, I pretty much rewrote the whole system using Gen.ai, still not knowing really how to code this. I mean, I would say my coding background obviously helped. I wouldn't expect somebody who didn't have it to be able to do it, but I still can't write Java or anything else, JavaScript. Tejneet Selhi: Yeah. You don't need to know that now, right? Because AI is building it all for you. Barbara Steinberg: Right. And we had all these different developers and everyone used a different, you know, language format. know, our front end is React node, know, our back end is Java. This one is JavaScript. you know, it's just great. It's a TypeScript. This is the AI. It's all over the place, but it doesn't matter because JPEG knows it all. Tejneet Selhi: Yeah, yeah. Barbara Steinberg: So anyway, so I was able to do a lot. mean, there were really a lot of shortcomings and stuff like that. couldn't, you know, scale, you know, all kinds of stuff. Security wasn't there. So, so since I've been knowing, mean, you know, tweeted and have worked a lot, you know, day and night for a lot of months and stuff like that. And along the way, hired some other people. Every time I'd hire somebody else, they would break it. What happens is one thing I didn't realize upfront was the complexity of the technology. Until you get into it and all the pieces, like everything is an AWS, we must use 20 different AWS tools within it. And then all these languages, there's a different repository for each one, each one's a different language. But I'm able to handle it all with Gen.AI. I I sit here and I'm doing stuff. I'm doing some new reports and dashboards and things like that, user interfaces now and stuff. Tejneet Selhi: Mm-hmm. Barbara Steinberg: One cool thing I'm working on, I'm very proud of this. If you think about these cases, like I said, the money goes back and forth between all those accounts and things like that. Now, I'm an investigator, where do I start? What am I looking at? So one of the things is the inter-account transactions, because they just move money around. My brother-in-law and my cousin, everybody's got accounts in these cases. Tejneet Selhi: Yeah. Barbara Steinberg: So I just did a report on a diagram that actually shows, it summarizes no matter how many accounts are in there, whatever time period or whatever, it shows the money going, the transactions, inter-account transactions with a diagram that says for each account, okay, how much is going in, how much is going out, where is it going to and stuff like that. So you can see all the linkages. So the $2 million is going between these two accounts, you're going to focus there rather than $2,000 between these two. So now it's really the analysis piece and getting into it. But the beauty, the whole thing, the beauty of e-fraud is that we have, it's near 100 % data accuracy. I will say, you know, the days of OCR and errors in manual correction and stuff, which is what most people are doing now, you know, are beyond us. And it's fast, you know. Tejneet Selhi: And the security is built in as well, you said? Barbara Steinberg: The security, we're using AWS security to take care of it. And we have a business intelligence platform that we white label that's built into it. And that gives us security, know, one customer can't see another customer's data and things like that. And we've got a few capabilities. Some of the features I'll just mention quickly, and this is based on my being in the industry. And for a while when the other software got so bad, I was the service bureau, I was doing it for Tejneet Selhi: AWS. Great. Barbara Steinberg: and stuff like that. So I was actually a user, so I get it. So it's browser-based. So you don't have to buy licenses and install them on your computer and stuff, which is always a problem because sometimes they have a lot of cases, sometimes they have no cases, sometimes they have to share a license, you know, that type of thing. Anybody can use it from anywhere, anytime. We charge based on usage. We have little people that don't need it so often, you know, it's based on how many pages they do, and so we can scale up and down as far as that goes. Tejneet Selhi: Yeah. you Barbara Steinberg: And, you know, it's the security, the scalability, the performance, and the accuracy. Tejneet Selhi: Amazing. That sounds great. And in terms of your target users or your target audience in terms of who's buying the software, is it Is it the government? ⁓ mean, it feels like it could be open to a lot of different ⁓ end users. Barbara Steinberg: All the above. market is huge. Government, ⁓ worked with every level of government in the US from the federal government, ⁓ government, and the they're all using old software. All ⁓ of are 25-year-old software and they're all struggling. So, I mean, we know the FBI. Yeah, ⁓ exactly. Tejneet Selhi: They need an upgrade. They need eFord services. Barbara Steinberg: ⁓ We know the head of the FBI who does this, the Department of Defense, the IRS, mean everybody on the federal government level. And then you've got the state governments, the attorney general's office. Every state in the U.S. has a Medicaid fraud control unit. So they're investigating fraud. Every agency has an office of the Inspector General. I mean, there's just so many different places and agencies and government, and then all the local sheriffs and the police departments. You know, we have 3,000 sheriffs in this country. You know, so is every level of law enforcement, even like we have somebody here in one of the county schools, they had their own police force and police chief, you know, contacted me and said, said, you know, we have somebody embezzling, you know, from us, you know, we want to investigate. So you have that. Then you have forensic accounts. A lot of individual practitioners up to big, you know, giant companies worldwide. Then we've got attorneys prosecuting white collar crimes and all this kind of stuff. Bankruptcies, divorces. I mean, one of the cases you're talking about is a huge divorce case. It's crazy. Um, so that, then the businesses, mean, I once had a guy, he says, my partner, think my partner's in Besling for me. Oh, so there's two men, you know, company, hired a forensic accountant and he suspected that they were using the, the company American Express credit card for his personal expenses. So, you know, we did the analysis. People, sometimes they lose their computers, you know, their, their QuickBooks disappears and stuff like that. They've got to redo all that. Um, I mean, the uses, they're. Tejneet Selhi: Yeah. Barbara Steinberg: or amazing. And it's not just a US problem, even though we've got crazy banks and stuff here, it's a global problem. Tejneet Selhi: ⁓ definitely. I can definitely see. ⁓ does ⁓ technology translate globally in terms ⁓ all of the security protocols, legal protocols, things like that? ⁓ Barbara Steinberg: Yeah, I mean, those are the things when you ask, you know, what's coming down the pike in the next 6-12 years, know, obviously, you know, EU, you know, has the requirements, security requirements, you know, and that kind of stuff. So that's the kind of stuff that we'll be adding, you know, and building upon as we go along. But in terms of the basic technology and the software, nothing changes. We can do anything in any language. We just have to, you know, edit, you know, what language we're using. It's amazing. Tejneet Selhi: Yeah. Yeah. Yeah. Amazing, amazing. Yeah. Barbara Steinberg: We're going to talk about AI, so I'll let you lead me. Tejneet Selhi: Auntie, it's so fascinating. I hear you talk about it for hours. I wanted to talk about is, ⁓ we work and talk about industry trends, but why don't we talk about AI first? How are you ⁓ ⁓ in eFord services at the moment? ⁓ Barbara Steinberg: Yeah. Yes, AI I use for everything in my life now, literally. ⁓ And again, just I'm doing it day in, day out. I am on AI. Also, that's my tool. That's my best friend now. That's my team. Tejneet Selhi: ⁓ Good to hear. Which AI ⁓ bot are you using for Claude? Yeah. Barbara Steinberg: Yeah, and we're actually using Claude. Interesting story. We started with Claude. and switched to Gemini last year. And the reason we switched was, again, we've got this big data coming in. know, people upload PDFs. You know, I can have a 50 megabyte PDF. know, Claude couldn't handle it. know, Gemini did. We were able to handle it. It handles the PDFs better. And on the output side, it was very limited. It was like 8K. You know, we could output in Claude and we were putting out tons of stuff. So, you know, and then we're chunking it and stuff, but, you know, if you got 8K version, Tejneet Selhi: and Barbara Steinberg: versus 64K, you know, it's a big difference. So that's why we went to Gemini. So Gemini was our engine. However, for development software and code, I got very frustrated with Gemini. I'm back to Claude. And I will consider even using Claude as the engine again. And the only thing holding me back now, and I need to see it, the PDFs. Gemini handles the PDFs very well. Tejneet Selhi: Bye. Okay. light. Barbara Steinberg: And Claude didn't, I don't know. Claude was moving so fast that you don't know, but. Tejneet Selhi: Yes. Yeah. Yeah. ⁓ good to know. And what about in your, because you said you're using AI in every aspect. How about in terms of your managing your prospects and things like that? Do you have a database or ⁓ an SDR set up? Barbara Steinberg: ⁓ yeah. Yeah, we use HubSpot for our CRM. ⁓ I use Sales Navigator a lot as a tool. Our audience, I have 27 groups on LinkedIn, thousands of members. Then there's another thing with our market, professional organizations. There's an organization called the Association of Certified Fraud Examiners, 94,000 members worldwide. There's just and tons of them. was just talking about that as a founder of the local chapter several years ago and I was talking to the president the other Tejneet Selhi: Yeah. Yeah. Barbara Steinberg: the day about doing some presentations and stuff. So there's a ton of, know, organization, professional organizations. what I found on sales navigator, and I have lots of connections, obviously, from over the years, that, you know, I can reach out to people and people, ⁓ high level people in big companies, ⁓ me. ⁓ So, but what I ⁓ AI there more as a tool, that's almost like, you know, a manual thing, you know, I write a note, something that in my mind, marketing is lot of personalization, Tejneet Selhi: Yeah. Barbara Steinberg: relationships, these mass mailing, automated things I don't do. Tejneet Selhi: in the end. Yeah. Barbara Steinberg: What I do use it for, I'll tell you, like an interesting thing. I mentioned the Medicaid fraud control units. There's 53 of them. And they have, they publish a directory, you know, for the public of every director and all their information and stuff like that. It's a PDF. So it's a PDF. And, you know, a few years ago I would have retyped it or whatever. So I gave it to Claude or whoever. And I said, you know, give me a list of, you know, Excel list or CSV list that I then uploaded in time Tejneet Selhi: Mm-mm. Barbara Steinberg: but mail murders into an email. from a time-staving standpoint and those types of things, ⁓ use it as a tool for all kinds of things. Tejneet Selhi: Yeah, nice. Very cool, very cool. Love ⁓ then what sort of industry trends are you seeing at the moment, would you say? Barbara Steinberg: Well, I'd say everybody sees that fraud is growing rapidly and rapidly. ⁓ know, the big thing we see, I don't get that much involved with, is elder abuse, elder fraud, and stuff like that. Tejneet Selhi: It is. Barbara Steinberg: people get to take advantage of. But what's happening now and AI is, you know, accelerating the rate of fraud because of what fraudsters could do with AI and, you know, countries and stuff like that. The bad actors disguising themselves and doing all kinds of stuff. it's ⁓ real. mean, I will admit, I don't write anything anymore. I tell Claude, you know, what I want to do. I give him background information and stuff like that. It writes better than I do. So anybody in the world can write Tejneet Selhi: Yeah. Barbara Steinberg: You know something that that you know looks reasonable and implausible and make sense I even see it, you know in emails, know, sometimes I mean and I'm very you know, I stood about this stuff, but you can see stuff that Anybody else, you know click on you know, you could really get into trouble very easily ⁓ Tejneet Selhi: Definitely. Yeah. Barbara Steinberg: Yeah. And it's interesting, know, some of the fraud cases that I've seen over the year, they're not anything high tech or anything. They're just people that think they can get away with things. Tejneet Selhi: anything. Barbara Steinberg: You know, we have one case, the guy raised like $8 million for a tech product that was going to revolutionize credit card processing, make it cheaper, faster, something like that. Never spent a penny on the business. He and his lawyer, he and his attorney, this was a public case. They had apartments, they had houses, their relatives had cases and stuff like that. It's just, people do these things and I get, I don't know how they think they can get away with it forever, but it's weird. Tejneet Selhi: Yeah. I know, it's incredible, it's incredible, but I do agree with you. scams getting more more sophisticated. How is E-Ford helping navigate those new and sophisticated scams? Barbara Steinberg: Yeah. Yeah, and again, it comes back to, know, we're looking at the bank statements. You know, that's what they're looking at with the money trail. Tejneet Selhi: Yeah. Barbara Steinberg: The other thing that we do that just happened, I didn't do it intentionally, although I would have, when we have a check that's listed, know, check one, two, three, four, check five, six, who is it made out to? So one of the things the investigators always had to do is go back and look at the checks. Well, guess what? If we've got check images in the PDF file, it automatically populates who the payee is. If we have it, it puts it in the description field. Even if it's handwritten, you know, or whatever, it may not, you know, recognize it, but puts it in there automatically. What is all about is saving time. Okay? Investigator. Tejneet Selhi: community. Barbara Steinberg: going to be spending months on a case, weeks and stuff like that. And if they have to spend weeks on just the data analysis, data and entry, that's not a good use of their time. And ⁓ of the problems with the growth of fraud, ⁓ this is a past problem, they can't prosecute all the cases ⁓ are there, especially law enforcement and stuff like that. So this enables them to process more cases in less time and then doing the things they should be doing, which is investigating the crime, not fixing up, ⁓ in data type thing. Yeah. Tejneet Selhi: Sure. Yeah, yeah, absolutely. They can spend time on actually trying to catch the criminals as opposed to having to analyze thousands of of data. So ⁓ a really, really important software that you've built to be fair. Yeah. ⁓ Barbara Steinberg: And the key to it, again, is the accuracy. There's plenty of software out there. Everybody says, you know, we can convert bank statements to Excel and try to use it. You spend 80 % of your time correcting it. Tejneet Selhi: Yeah. Yeah, yeah, absolutely. Well, like you say, you've spent a long time trying to build it, right? You've gone inside out in terms of this software. And I think, as you mentioned, you're now an industry leader because of it and well respected in the field. So we're lucky to have you on the podcast for sure. What would you say, you've touched on it before, but in terms of your future goals for eFord services, and maybe adding in GDPR and things like that. But where would you like to be in terms of the business, in terms of the number of clients and things like that within the next six to 12 months? Barbara Steinberg: Yeah, I mean our first year we were projecting a million dollars of revenue and then doubling year over year after that. I mean this could be a hundred million dollar business, no problem. And I'll put on one more plug. Something came out accidentally that we now have as a function. One of our clients asked us, could we do brokerage statements? And I was like, yeah, we do it. We take the cash transactions like anything else. was like, no, no, no. We have this 97 page brokerage statement. It was a six month statement. And we want every page, all the data in Excel so we can analyze it. OK. Well, if you think about this brokerage statement, it's got 20 different tables in it. Each one has a different format, different columns, some is details, some is summary, this, that, and the other thing. So we now have the capability to upload the same way you upload your bank statements, upload the PDFs. We have PDF to Excel, upload any financial document, whether it be payroll, general ledger, brokerage, or whatever. And what we do is, and remember, we don't know what it looks like beforehand. We have no idea what's in any of these things. it figures out, it takes each table, it finds each table, it formats it correctly with the correct columns, and each table goes into a worksheet and into a workbook. So now you've got a workbook, each one is nicely labeled, you can analyze each page because it's all in a common format. So that's a big another thing that you know someday we'll get to, but it's there. Tejneet Selhi: Amazing. That's huge. That's huge. And so sorry, you mentioned this could be a hundred million dollar company. would you like it to be at the end of 2026? What are your ⁓ between now and the end of December 26? Barbara Steinberg: mean, financially, like I said, you know, the first million dollars in revenue. but, just a good, I mean, I think we'll have a good cross section, you know, of users, you know, in the U S at least, we'll probably have some other ones too. But I've met a lot of people in Canada, know, trade shows and stuff like that. The salvation army in Canada was, ⁓ was over this. so, ⁓ just. Tejneet Selhi: Yeah. Barbara Steinberg: you know, wherever it goes. And the other thing is to, you know, offload the technology. I have been very protective now because we've been burned so much and stuff like that. So get it to a point where, you know, we have trustworthy people, you know, who can do that and then take it the rest of the way. Tejneet Selhi: Amazing. So you want to get to a million dollars by the end of this year hire ⁓ some people. Is that? Barbara Steinberg: Actually, we'll probably outsource most of it. We have a company that we're working with now that really good gets us and stuff like that. They're an AWS partner. And so that suits us better. One of the things I'll say here, ⁓ what doing with technology, it would probably take six engineers to do it because that's another pit people in mind today. Everybody has their own little specialty. and do everything. And that's one of the problems because they don't see the big picture. That's why things never worked. Nobody cares about the business problem or what it is. Just tell me how to do my little thing here. that's where I come from. You did the whole thing. And obviously, was a long time ago and technology was a lot different. But that's a big shortcoming that I see today. Tejneet Selhi: Yeah. Yeah, yeah, absolutely, absolutely. Well, look, it has been an absolute delight speaking with you, Barbara, and I'm excited for this particular episode to air because I think any up-and-coming entrepreneurs can learn a lot, a lot from you ⁓ and business. And thank so much for your time and coming on our podcast. ⁓ Barbara Steinberg: Tejani, thank you very much. It's been a pleasure. Tejneet Selhi: No problem. Take care. See you all next time on Entrepreneur Odyssey.