#420 Hunting the Invisible Man: Roy Daya on AI, Fraud, and the Truth Hidden in Data

Welcome to the Awakening Podcast. You'll find the episodes on Awakeningpodcast.org. What if the biggest threat to your business wasn't a hacker breaking in from outside, but a problem hiding right inside your own data? My guest today has spent 20 years hunting fraud across Turkey countries from governments, corporations, and private equity firms. He's seen it all, and now he's built an AI tool, Prova.ai, I hope I'm pronouncing it properly, to catch what humans miss.

Roy Daya, please welcome to the show. Hi, how are you? Good. Happy to be on the show.

So I heard Mr. Wolf, and I was like, what's that? So how did you get the name Mr. Wolf? It's one of the interviews, the interviewer decided I'm like Mr. Wolf. And then I went back to the fiction again, and to actually see how it, how it handles. So it's a compliment, because I think he really knows what he's doing.

But you know, I've been doing this for many years as well. So maybe it fits. So I suppose maybe before we delve into it, you might just kind of talk about your career to date.

I started pretty early in cybersecurity. I was what you could call an independent cybersecurity researcher, which basically means like a hacker. And so I needed to find like a legal, legal way to use my talents.

So I started working in different cyber startups very early. And, and, yeah, and then as I did to expand my toolbox more, I started to realize it's not just about playing the threat game. Because a lot of threats also today in cyber, they're not clear cut, we're going to break in, steal something and run out.

But it's much more embedded in operations. A lot of times, the hackers will actually go into your systems and stay in the systems for a long time to monitor and see who's the CFO, who's responsible for the money, who's responsible for account. And when the time is right, they'll start sending emails internally, asking for money to be transferred to them.

And it may take quite a while until, until people realize that and actually stop that. So today, sometimes it's not very clear. It's not like a robbery and somebody, you know, pulls a gun and tells you, give me your money or your life, you know, it's somebody internal that you may be aware of or not, that is misappropriating all your assets.

It's like having a criminal organization inside your organization. And these things, they're, they're compounding with time. So as time goes by, it really scales up.

And, and yeah, a lot of times it gets found out, but usually a year and a half too late. And usually when there is a lot of damage and a lot of reputation damage, and, and these things can actually be caught much, much sooner, if you see how the trends start to evolve. And this, this is basically, you know, what, what I do is basically to find the small signs that something is wrong, and then you can put a stop to it.

So you don't have to fire, you know, the person that does something bad, but if you know that something is wrong, you can at least lock that door or make sure that it's not that easy to, to get to that asset or to do these kinds of things. You know, whenever you have people and opportunity and some assets or money or anything, some people will start checking, will start seeing, can I just take something? If I take what happened, they start giving themselves like an excuse, I deserve it. Or if they put it like this, they expect anybody to take it or, you know, and then they start to get to do it more and more and more, and then it, when it starts to explode.

So I, you know, I can go ahead and tell you more about how I do that or what's. Yeah, well, I mean, I suppose one thing with your career, what's the one thing that would kind of keep you up at night going, whoa, can't believe somebody did that. You know, some, like one employee changes status, you know, from employee to corporation.

And actually, instead of getting a salary, started getting like a salary of 150 people, like outsourced, you know, we just change his ERP credential. And, you know, IT people, they have a lot of access to a lot of different things. So, you know, a manager might need to go through, you know, five different approvals, but the IT guy can just, you know, go to backup files, change something and then restore a backup.

And, you know, everything is, all his changes are in without getting any approvals and with no audit trail. And, you know, so there's a lot of things that people just don't think about. And, you know, with data, the reason you can hide things in data is because there's so much of it.

People told organizations, you must keep data, you must analyze data. So they started keeping a lot of data and now they have billions of records and they're just, they're drowning in data. They have no idea what to do with that data.

And it's just too much. They can't really, you know, if you have 50,000 records, you can put them in a cell and start pivoting them and try to figure out what's going on there. But if you have, you know, 1 billion records a day generated in your ERP system, what are you going to do with it? You're going to bring data scientists and start building a whole data science department, paying them millions of dollars just to try to explore and look for something interesting, don't find anything.

And this is basically where POVA comes in. POVA is just the result of me doing that for like 20 years. I actually took all these tools and put them together in 2022.

And I was just operating with them, you know, silently and just not because I started to sign with some more partnerships. Internationally, I'm in Japan now, so we have a Japanese operation now. So I had to put a name into it and actually put it out.

And POVA is very simple. It's just point of view analytics, which basically means it takes one of the problems with billions and billions of lines of records is just anything that somebody does is so small in comparison to the rest that it disappears. It's like noise.

So even if somebody has 100 lines where they're doing something very bad from 1 billion lines, you will never find it. Unless you ask a very, very specific question, you'll never find it. The problem is, how do you know which question to ask? So if you would ask what happens when that specific rep is in that specific branch and look only at the sales of this specific product lines and look at these corporate customers and look, you know, if you filter 10 different things and then you get a very small list of items and that's the items looks very, very strange.

And anybody that would look at it to say, well, this is weird. That's not how it usually works. That's that's not the pricing we usually give.

And that's that's really strange. But the problem is people have dashboards today and dashboards give you great answers to the questions you already know how to ask. But who will find the questions that are worth asking that you don't know how to ask? So the only way to find is a hundred interesting transactions is if you have a question that the result is these 100 or 200 of them of transactions.

And that's what POVA does. POVA is is just not a lazy software. It works very, very hard.

And scanning an organization can take a couple of weeks even. It's not a real time antivirus. And what it does is it takes the data, let's say, your list of asset transfers or sales or returns.

And it tries to see from all the different fields that you have there and, you know, columns and different data items, which views we can construct from them, which pivots we could make, like by product type, by date, by day of the week, like by weekend, by type of rep, by, you know, by product family, by product origin, you know, there's millions of combinations. And POVA actually runs for each one of these millions of combinations and run a set of, you know, many, many algorithms, including really cool stuff like geoprofiling and, you know, like different distribution noise and like a lot of different things to try to see is this pivot is strange or not? Is it like all the others or is it different? And when people think about anomalies, they think, you know, if all the numbers are between, let's say, 50 to 100 and then you have one which is 10,000, obviously, it's an anomaly. Yeah, but, you know, nobody's that stupid.

So, you have to find things that are much more settled than that. So, yeah, it's 50 to 100, but it's closer to 90. And it's like, it's more of them in these two weeks related to something else.

It's like it's finding a pattern which is much more subtle. And that you can find if you ask all the questions. So, basically, it's a system that takes all your data, takes all the points of use you can imagine and start running very hard to try to find anything that doesn't add up.

And then it takes that data and regenerate stories and try to see what would make something look like that. So, it finds the trail and then it thinks what kind of animal would make that trail. And a lot of times it's like an invisible man in the middle, because a lot of somebody that might do something bad will not put his name on the transactions, but he might be the manager of somebody and a friend of somebody and in the same branch of somebody else.

So, you'll find him like in the crime scene. You cannot say that he's guilty, but he's always in the crime scene. So, maybe it's worth interrogating him.

So, it's a lot of correlations. It doesn't mean causation, but let's say me and you are meeting every Tuesday under the bridge and we're throwing dice and we always get six. So, if we did it three times, it's okay.

It's lucky. If we did 200 times in a row, throwing the dice, always getting six. So, statistically, it's not probable.

So, it means either the story that we're telling about this is wrong or the dice is wrong or the world is wrong, but something is very wrong. So, POVA will bring the things that are not naturally occurring. They're so different from things around them.

It's like it lights up like a Christmas tree. Sometimes you might think that detecting fraud is hard, but when you have something, it's very easy to detect because it's wrong from every angle. It gets flagged by 20 different algorithms.

It's very hard to fake data. Very, very hard. And you see that a lot in public-released data sets.

So, it's like they have to release everything, but there's a lot of things they don't want to release. So, you either see these black holes in the data, like some generic names for some things and then no explanations or they always cut it when you don't see who actually bought and you don't see what they actually bought, like all categories and departments. It's very vague.

So, you cannot really trace it to somebody specific, but still, you can find a lot of crazy things. A lot of things are probably should be asked, the people that are leading us, like what's going on there. And you can see that a lot with health.

You can see that in almost any industry. And with blockchain technology, because I know that in the crypto world, there's a lot of things. And what I've heard recently with one of the guests is that they'd try to go on a Zoom call, even as if they were doing an interview and they plant their bug and they're inside the system then and eventually they end up emptying accounts.

Have you worked with that or will it work in systems data? Well, when you have people in organizations, it's not always who is in and who is out of the organization. So, sometimes you have a whole chain of third party affiliates. So, nobody is really in or out.

You don't have just customers and company. So, a lot of times you have people that are like hated and they get access to different things. And yeah, obviously, sometimes you have somebody totally external that gets some piece of information, let's say watching a Zoom or watching something, it sees the URL of something or it can go in and try to do something.

But a lot of times it's an internal somebody, like it can be a marketing agency sitting somewhere abroad and they have access to your dashboard and they can see different servers that you have. And then somebody there sees something and it's like they have internal access even though they're not internal. So, it's related to what you asked, but you can get escalation of abilities to do things without even knowing who is actually involved or not involved in your business.

I think insurance fraud as well. So, is this working for the insurance company on claims coming in just the way that it's working or what exactly is that? I think, first of all, insurance companies and even government, they use some of these algorithms to see, for example, if you're giving them some, let's say, a tax report when you're saying all your expenses and all your revenues, they will check statistically if you just sat down and made up all these numbers or if it actually happened. And actually, the patterns look very, very different.

People think that they can create random numbers, but they can't really. It's very obvious when somebody just types the numbers. There was some professor that he did this test with his students that he said half of them should throw flip coins and just write what they got.

The other half need to just simulate it and just write it by hand. And he immediately could tell which ones did that and which one did the other. So, it's very, very hard to fake data.

And you can see it everywhere. When somebody, they have to disclose something, it's very easy to know that it's fake, but a lot of people don't know. And what it causes is that today we have a kind of a data economy.

So, organizations are buying data all the time. So, let's say, you're a lead to somebody and they want to sell you something. So, they'll buy and reach the data about you.

They'll pay maybe a couple of dollars or whatever to get more information about you. Now, that company, they might have information to sell and they might not have it. And if they don't have it, sometimes they will create it.

They will create some averages. They'll create predictions and they'll give them something. And so, sometimes it's close in, but some companies are taking this data and they're creating models from it.

And these models will be the models that help detect different things about people. If these models are using bad data, then the models are going to be bad. And on a lot of different things, you end up having AI trained on bad data, making bad decisions and helping people make bad decisions.

Because when you make up data, it's not diverse enough. It's very in the middle. You don't have enough extreme cases.

So, a lot of times the data and the models, they don't know how to deal with extreme cases because they've never seen them. So, this is a big risk of bad data. Especially, you can think of, for example, there's a medical device.

Let's say you buy a medical device and it has models that take your heartbeat or whatever, and they tell you if you're fine or not. Now, they train their models from systems that have medical information. Now, this system does a lot of them, I think more than 200 different vendors in the United States.

And sometimes you will get data for the same patient from two different systems, which is totally different. Because a lot of the data is missing. Nurses, they don't write it down or they write it on the bedsheets and something.

And a lot of the data doesn't get filled in. So, some of the systems, they just copy the old data. Some of them are just predicting what data might be because they don't want holes in the data.

They want complete data. And then they sell this data to the companies who do medical devices. And they're using that data to make the models.

So, now you're hoping that this device is going to be a good model, but you don't know it was created using a lot of fake data or data very skewed to very specific type of patients. Let's say they will buy data, for example, from India or from China. But, you know, humans are not the same.

And some models for healthcare that work in China very well will not work in the United States or not work in South America because people are a bit different, they eat different, the weather is maybe different. So, it's not exactly the same. And another thing, especially you're thinking pharmaceuticals.

Pharmaceuticals, they're always afraid, you know, FDA or a lot of different bodies around them. And it's a group of companies that are not always together because one of them, let's say, factory. The second is the R&D facility.

The third one is like a marketing group. The fourth one is manufacturing something. And another one maybe is a warehouse that has like expiration dates and different, you know.

And they all, they have to all report data which looks good to the FDA. So, even if they have a problem, they cannot say it because if they say they have a big problem, the whole group can go down. So, they all know they're going to keep reporting everything is wonderful, but they don't know if they can trust each other.

Because even if they wanted to tell each other we have problems, we had a really bad manufacturing batch, we have a problem with production, we have a problem with clinical trials, they will not tell others in the group because all the group has to be happy for all the group to succeed. But these kinds of things you can see from the data. If you know how to analyze the data statistically, you can see exactly where the bullshit starts.

So, it's very interesting. No, no, definitely. And with your experience, like with the insurance, because what I'm seeing, this is on the other side for the individuals.

They're, and you know, like they're using AI is what I've heard now, to actually stop paying out people. I've had a few legitimate claims didn't get it. And nearly every single person you talk to, they're having the same thing.

How do we fight them companies to make sure when it's a genuine claim that we actually get paid for what we've invested in protecting ourselves? First of all, insurance companies are not all the same. They're calculating their premium from based on their own pool and their own risk for your specific type. For example, if you want to take a loan and you'll say, like from a bank, and they'll ask you and you'll say you're divorced, for example, you'll pay higher percentage because they think you're higher risk.

If you want to take a loan and you say, no, I'm really good. I want to take the loan and close, you know, all my debt and like be a really good citizen. You have a much lower chance of getting the loan than if you just say, I want to buy a Ferrari.

Because they're thinking, you know how to operate. You just want to spend the money to buy a Ferrari. Good, we'll give you the money.

But if you're saying, I can't make ends meet and I need the money just to survive, they're thinking he's not going to pay it back. He's going to keep falling down. So like that with insurance, they're classifying you.

So they have a model that classifies you. But this is a different, I get that. And that makes sense, what you've just said.

But this is more like, say, if you were like car robbed, like that's something that happened to me. They broke into the car, took everything and they wouldn't cover the cost of that. A brand new computer.

And there's other people that are paying insurance, house insurance. And when a leak happens or something boiler breaks that they're supposed to cover legitimate cases. Like I've seen this all over the world.

Nearly everybody I talk, them kind of claims that you've paid for it. Like I had a holiday health insurance one as well. I had medical receipts proving that I had an accident and they had, they wouldn't pay.

And that's not just me. That's nearly everywhere. Them kind of claims, like how is there AI that we have to start using that can kind of put together the claim? Because usually people just be, they're just speaking from the heart.

They're putting it as it is, which it should be. Hey, this is what happened to me. Here's my proof.

And you just expect to check. And yeah, 90% of the time they're not paying out. I think first of all, the reason that they have a contract that are so complicated is first of all, you wouldn't understand.

And second of all, to give all the small terms that basically put the odds in their favor. Because you know, they have to always win. So they will say, for example, we will protect you.

We will pay back if you have water damage. And then they will have in small print, but not water coming from within the building or water coming from the outside, like weather. So it doesn't really leave any water, any other water unless they just, you know, come into existence on their own.

So I think one of the things is to be able to get this complicated agreements and have maybe AI trying to triangulate what are the cases that they are willing to pay for. And then you understand exactly what is the risk and what, when they will pay and when they will not pay. And also when something happens, you know how to describe your story in the terms that they have to pay.

And sometimes, you know, a very small question that they will ask you will just want to clarify, you know, exactly what happened, exactly what you did. And if you just use the wrong term, they don't have to give you anything. And they can even sue you sometimes, because you operated against what you said you're going to have.

You know, some people, you know, they have to have safe in the house and the safe has to be, you know, bolted to the wall with specific, you know, bolts or something. There's a lot of very specific things. And some of them is to confuse you.

Some of them is that they know you're never going to do it. And some of them, you know, it's just, it makes the claim almost impossible, like with the water. The water has to come from somebody, from somewhere.

But if it's not external, not internal, and not pipes, then where is it going to come from? And also, you know, insurance companies, it's not one, it's like a tree. They're, you know, like Lloyds or like other, like they're big insurance. And then there's other insurers, some of them are creating like very creative insurance product, which they will buy some insurance product and they will add their own thing.

And sometimes make it cheaper, but then they don't cover as much. And, you know, some of them makes it, I saw insurance that says that it's like an additional insurance. And they say that if you're covered by any other insurance, then they will wait for the other insurance to pay everything.

And only after they pay everything, then they will see if they need to pay more or not. And they know that the regular insurance is going to take you like 10 years to chase them. And I've seen as well that sometimes people try to make, they might have a claim, say with our credit card or whatever payment processing, and they make a claim and then they have insurance and they go bold, but they don't realize they're talking to each other.

And then you can have a case against you for fraud. It's like a casino. The house always wins.

Almost always wins. I know that, you know, there was a saying that in every bet there are two people, one is cheating and the other one is stupid. So it's something like that.

They're selling you de-risking. So they're selling you, if you pay us, we'll make you feel like you're less in a risk. But basically you have the same risk.

But not only you'll be sad because you lost the money, you'll also be upset at them. Fighting them with AI means it's like having an AI lawyer next to you, analyzing that agreement and sending them, before you sign it, sending them specific notes about things that you don't agree with and trying to get the reps or somebody to agree to some of the things you're saying without them fully understanding what they're agreeing to, because they want to sell you. So, you know, maybe there is a way of fighting back.

But a lot of big organizations, you can read it carefully and just sign on the bottom because nobody will change anything for you. They have no problem losing you as a customer. I just want to tell you something about insurance, that a lot of organizations, they have to take insurance as part of a lot of different projects that they do.

And sometimes if you need to make insurance, you can go under the umbrella of somebody else and you can save a lot of money for that. So this is something I did in other companies before. For example, if you provide services to a big company and you must have insurance, then you might be able to go under their insurance and save doing your own insurance.

Because their insurance covers everything. And just on that, because it's kind of connected with what you're doing, because sometimes insurance companies give a kickback for the person that gets the client, which in turn, sometimes it's the person in the company getting the kickback personally without letting them know. Are you able to pick up on it? Because that's a hard one because, you know, they're not disclosing it.

But is it possible to pick up on them once? It happens a lot when you have a rep that is talking to customers. Now, either he's talking to a lot of small customers or he's working with a few very large customers. And when they see money coming in large quantities, they will try to get some of it out to their own.

So they might say, you know what, we'll give you a really big discount, but then you can meet me under the bridge, you know, and give me $2,000 and I'll save you money. But then, you know, if you do it once, then maybe it's, you know, it just happened. It's this type of customer.

But if you look specifically on that rep and how it starts to change with different customers, and they're all starting to get very high discounts. After this rep has been selling, regular prices for a long time, then if you isolate this, it starts to look really bad. Why? Everybody's starting to pay very, very low.

And obviously him buying a Porsche is also, but you know, and that's after that. But yeah, when people start small and then they feel success, they'll start to do it more and more and more. And the problem is by the time it actually explodes, it's too late.

It's like a year and a half, and they've done so much damage and so much reputation damage. But as a business owner, you can catch that, you know, in a month or two, after like two or three transactions, you'll immediately catch something is wrong. They threw six too many times.

You can see them throwing six, you know, 10 times in a row. You don't have to wait for a thousand times in a row to know something's bad. So yeah, so you can do that as a manager looking at your organization data.

You can do it as a citizen looking at your municipality data. A lot of municipalities, they're releasing like a full disclosure of their older transactions and contracts. There's a lot of things there.

A lot of things that are, you know, very, very strange. We, you know, to train POVA, we have to scan a lot of different things. And we cannot just scan synthetic data that is made up because the system will just spit it out and say it's bullshit.

It's just made up data. You can't give me the made up data. Just everything will be red.

It's not. So we're looking for all the open public data, all the different cities that, you know, New York, Chicago, a lot of them that have open data. And all the, uh, there's a big medical, uh, paybacks to doctors, uh, databases.

There's a, there's like real estate biddings. There's a lot of data that is very interesting and they count on it, that it's too big for anybody to analyze it meaningfully. So you can take it, you can make a few graphs and they say, oh, we released so much data.

Look at the graphs. It goes nicely up and down. Yeah.

But, you know, it's like they say the average temperature in a hospital is 36.6 Celsius. So everybody is fine, but no, some of them are dead and some of them are 41. So, uh, the average, the average doesn't say anything, but yeah, when you start to analyze very, very carefully, you say you see something very strange.

Like, like why would somebody make this kind of agreement? Maybe it needs to look at them. You know, if a city, for example, commissions, uh, I don't know, 10 times, 10,000 times to go and do, uh, land appraisals and it doesn't matter where it is, they always pay the same for like three years. That's weird.

I mean, wouldn't you negotiate and get, you know, uh, terms that are changing by like complexity, time, distance, things like this, or, or something would change or something else, but, but it's exactly the same transaction, like tens of thousands of times. So, you know, that's, uh, does a city that use a lot of, let's say water bottles, would they make like a hundred thousand transactions for water bottles and not to make some like a better agreement than that? Uh, it's a bit strange. So some things are strange.

Some things are, uh, just holes. There's some random names that they put for different things of explanations, and you don't really know what's behind it. And it can be a lot of money and some things are just missing.

Um, so at least somebody can go in there and ask the right questions. Like, uh, what the hell is personal services? Why is it 3% of the budget? Personal services, 3%. That's many millions of dollars, for example, in one of the municipalities.

Uh, so it's a lot of fun taking these databases and trying to, to hunt for, for different things inside and don't even get me started on the, you know, on the, all the reports. And it was very, very obvious that, you know, a lot, some countries, even all the reports for many months were just made up. They didn't know, I don't think it's fraud, but I think they just didn't know the numbers and just had to make up something and report something.

But you can see that a lot of the reports are totally synthetic and they're just made up random numbers. It's not, uh, they don't follow any, any pattern that, that the real numbers would follow if they actually happened in time. And I think if you want a good analogy for that, it's like, if you're looking outside when it's raining and you look at the floor, you can see the pattern of the drops.

And if something is wrong with the pattern in some area, you'll immediately see it. You don't understand, you know, the mathematical reasoning, but you'll see something is in the wetness and the distribution is different. And that's what you see there immediately in the data.

And with what you've said there, I know, cause this is very interesting regarding the cities, because there are, anything I've kind of investigated on this, they're all as corrupt as you get, even when I'm like, I'm Irish, but I'm living in Poland for like nearly 20 years. And there was, um, the guy that was representing the city for development. He was trying to sell stuff himself, privately developments, but he was so dumb.

He was using the city's email. So it would be very, and I just refused to deal with that kind of thing, but which like, can a service or can people come together and provide, like get, hire you to go in because there's stuff that's available. And as you say, like, it's easy to kind of find a few things, but there's also the freedom of information and can do like the AI that you figure out, okay, we need to create this and get the freedom of information.

Cause you know, they're supposed to give you the information. I know they go around it. And sometimes you can't give a list of 50 things.

You have to do it individually. But I'm sure a lot of people would actually benefit from that because we know like every city does so much potholes for so much everything. And they're just taking more and more property tax.

And so we know this fraud does. And like, I see people that are working for the, the, the city, they're all on junkets. Every time you see them, they're at an event, they're all getting freebies.

They're going on flights. It's like, okay. I go along and it's like, I'm having an epileptic fit as I'm driving on the road and they're, they're just spending the money.

So like, how do people get you with the company to do that? Cause obviously I think a lot of people would invest in something like that because they go, all right, let's get rid of the scum, the cheaters, the fraud and ruin a city properly. There's actually two methods to do that. The first method is what's called like OSINT, like open source intelligence, which means you can see that two vendors are basically using the same phone number or whatever, you know, emails are easy to make.

Phone numbers take a bit more effort to have a lot of phone numbers. So, you know, somebody might create like a hundred business names and then tell small sums from each of the business names to the same department run by his uncle, you know, and he'll collect like a million dollars a month, but it's from a lot of accounts, like small siphoning, so nobody will see. But this pattern will be strange in the data.

Why are we, why always paying these a hundred vendors one after the time, one after the other within a week, every month. And it's like, these sums are keep growing and this cluster looks strange next to other cities. So like identifying a DNA of a fish, which is great, but not if it's inside a cat.

So another way, yeah, so one way is to do the OSINT and to see the relations of the thing. And the second one thing is the data, you know, there's a lot of, when somebody is trying, let's say he works in a municipality, when he wants to get my coat, they're trying to think of different things, but what always happens is that they start to scale it more and more. So for example, they will get like some relative that he is a contractor to fix the potholes, okay.

But then, you know, they'll think, what if he makes also the, there is these bumps that supposed to make you slow down. So they pay them to make them and then they inspect them and they realize they're not up to par with exactly what they need and they need them to be a bit taller. So they break them apart, they build them a bit taller and they pay for it again.

And then they get complaints because people, cars are getting damaged. So they say, okay, we're with you citizens, they destroy it and they make again in the proper size. They actually paid that guy three times to do it.

So you will start to see that these kinds of projects, you know, maybe they look okay on the project plan, but if you compare them to other departments, other cities, other, it looks strange. That's not how it usually happens and not in this kind of situation. Yeah, shit happens and, you know, projects get redone and, you know, I saw a bridge built on the wrong side of the road, you know, actually between two walls by mistake, you know, there's a lot of different stupid things that are done.

But when too many stupid things are done in the same place, in the same terms, it's a question to be asked. No, the answer, you know, there is a saying, you know, everything happens for a reason. And sometimes the reason is that you're an idiot, but, you know, but sometimes somebody is cheating.

So, yeah, when you get, you need access to data, but not the data that's like summaries or aggregates or averages. You need the raw data, the data as it happens, like the raindrops. You don't need to get a bucket.

This is how much rain we collect. It's not going to help you. You need the actual to see the raindrops.

And then you cannot hide from the data. And, you know, actually I had, when I was younger, my first company was a CRM company. And I actually tried to manufacture a lot of different data because I wanted to show the system.

I wanted to show, for example, a system for sales management. And for that, you have to make a dashboard and to show, let's say, by a salesman, by something. And because it was like an analytical dashboard, I would try to think, but if somebody will try to see by product type, by day, by color, it will break.

Because when you just make up numbers, it will not look good from every angle. It's like Hollywood, you know, it's like cardboard. You cannot go behind the scenes.

It's very thin. You can never make the data well enough that if you look from any possible direction, it's going to look good. It's always going to break.

And that's what POVA does. It looks from every possible direction until it breaks. Until the Hollywood sign falls down and you see it's thin.

It's a nice facade, but there's nothing behind it. So, this is, it's a lot of fun because it's not like hacking that is like real time. You're trying to escalate your privileges.

You have to go through. Data is there and it can't hide. You need to find, you know, the X on the map and then, you know, you can take it to court.

You can do whatever you want. It's there. And even if somebody deletes the data, the deletion, the missing part is there.

So, it's if you're going to the field, trying to dry some spots of the rain, the dry patches are as loud as the rainy patches. Because why is it dry there? So, you can't really hide. You can't really hide.

And I don't know anybody that can really create data synthetically as believable as the real thing from every possible angle. That's a very, very long process. I'll give you an example.

I was taking a database that had 40 million transactions, okay? 40 million transactions is about, let's say, 10 gigabytes of data. So, it's 10,000 megabytes of data. Now, when I was running POVA analysis, you know how much data POVA created just with analysis? About 200 times that much.

200 times than the data. It's like replicating the data 200 times just to find that needle in the haystack. It's like, it's very hard to hide.

I used to have another company that was using security cameras for catching, you know, people do unsafe behavior in construction sites. And, you know, or helping them. You know, if somebody falls on the floor, you get help in 30 seconds.

Because the camera sees, analyzes, sees somebody's on the floor and gets help. Or if somebody is, you know, doing some work in a risky place, so, he knows that the camera is watching him. And even if the guy watching the camera went to the toilet or something, the camera itself can, because it will identify him.

So, it actually made some people feel very, very at ease. And one of the distributors told me, what if your camera will miss me, will miss the fact that I'm on the ground? And I told him, you know what, in one blurry image, you're right, there's a good chance that it will miss you. But, you know, in 30 seconds, and you have 30 frames per second, that's 900 pictures.

You know, it's a chance that the camera is not going to see you lying down in 900 pictures in a row. That's very small. It's very, very small.

I told him, you know what, if you run across the room and the camera doesn't see you, I'll give you all the products for free. He didn't try. He understand that it's a low probability.

So, yeah, it's all about probabilities, putting them on our side. Yeah. So, curious, because, like, I get how you're able to help all these kind of big companies and they're doing everything, spending big budgets to make sure nobody breaks in.

But it must be difficult for them, because they then have to trust you, because they're basically giving you the key to the house and letting you go inside. The risk factor, do you have that? This is, I'll tell you, I didn't know, I didn't know how paranoid some customers are. First of all, I had customers that made me take polygraph tests.

And also, POVA can create encrypted reports. So, POVA can take the data, run for two weeks, and create an encrypted report that I cannot see. And I just send it to them or give it to them.

And I hope there is a report there, because it might be, you know, error on line one or something. Couldn't find data or, you know, something like this. They wasted two weeks when computers are sitting and waiting for the data, you know.

Press enter to continue, you know, something. No, but in many cases, in the beginning, you know, you get the stories of, you know, everything is great, everything is wonderful. Yeah, let's do something, maybe we'll find optimization.

But as you spend time with them and they trust you more, you start to hear more of the really bigger issues that are. But also, you know, you had some cases that, that, you know, in some countries, we only go in with the legal of the company. So, backed with the legal of the company, because we had the case that one of the executives wanted to run POVA, and we had a very loaded report to give him.

And we were afraid that he's going to go to the other executive that had, like, very serious questions about him, and to blackmail him. So, we wanted to make sure that we are covered by the legal team of that company, that it's being done on behalf of the board and of the management, and not a tool for one of the executives to bankrupt, to really go and try to blackmail the other executive. So, I don't want to be in the blackmail operation.

But yeah, you learn, you know, you learn. And, you know, sometimes you go to businesses and, you know, who are the biggest problems? A lot of times, it's not the secretary. A lot of times, it can be, you know, the CTO, CFO, CEO, who's stealing the big money.

So, if you're going to, you know, if somebody, like, on the board asked me to go to the CEO and tell them what he does, and I see them start to get sweaty, I think maybe they're not going to buy it. Maybe they don't want to know. Because I was going to actually ask about that.

Is there anyone that's like the CEO? Especially, you know, if it's like the tech guy in the company, maybe is a director and decides, this will be good for us, we'll figure out everything. And if your man decides not to... That's exactly what they said. That's really bad for us, because we're going to figure out everything.

A lot of people get worried. A lot of people get worried. But, you know, sometimes you have like a private equity, they want to buy a business.

Sometimes, you know, $200 million. And they want to get all the skeletons out of the closet really fast. They want to know what they're dealing with.

They want to know what they're buying. They might still buy the company, you know, even if there are a lot of problems, you know. That's why they will usually buy, you know, good deals.

And they know it's like a fixer-upper. They know it's not perfect. Because if it was perfect, it would be too expensive.

But they want to know how deep the holes are. And yeah, that's scant. You know, we can give them, at least to give them in like one or two weeks, like a piece of mind to know if there's like somebody, something really, really big, you know, brewing there.

Or just, you know, small policy things and like a lot of things that maybe should be better managed or... And also, another side to POVA is not just looking for fraud, but is really explaining how operations work. So, for example, if you see that a certain facility, once they have more than 30 people come in every day, all the operations break. All the ROI becomes more expensive.

They start to get external consultants. They're starting to get, you know, to pay for meals and they work late. Like all the ROI goes bye-bye.

Then, you know, you have a breaking point. Or let's say there's a new manager and he starts replacing all the vendors with his friends. So, you see regime changes or when there's subscription creeps, that you start to get more and more subscriptions.

And then the pricing changes, you know, when you sign something because it's $9.99. And then you check it a year later and it's like $15.99. And the year after that, it's like $29.99, you know. And yeah, they sent you an SMS that the promotional period in the beginning is over. And if you don't cancel, they'll have to charge you the regular price or whatever.

We've been upgraded for free, but now you have to pay for it. You know, something. So, it catches also all these things.

And even if you have thousands of them. So, it has a lot of value. And I created it because I used to do a lot of consulting and I was very expensive.

So, people expect me when I come in to get results very, very fast, right? To go in, do my magic, whatever it was, and meet with the CEO or the board or somebody, you know, within one, two weeks and tell them what's going on in that company. And people would not cooperate. They look at you like they want to kill you.

You're the one who's coming there trying to tell them like they don't know how to work and trying to like be smarter than them and whatever. So, all I had was the data. So, I had to figure out ways from science, you know, from everywhere.

What can I know about the data without talking to anybody? And, you know, even like really, really cool things, geoprofiling. You know this theory of geoprofiling? Let's say you have a criminal. Then he wants to do the crime far enough from his home that people will not catch him, but close enough to his home that he feels, you know, feels familiar.

So, if he does more than one crime, you can see the patterns of the points in the map and you triangulate what places are equally distanced from his home and the crime and you can pinpoint where he lives, a building block. So, that's really cool and use it actually for crime. And POVA, one of the algorithms it has is also geoprofiling.

So, if there's like 10 branches close by and somebody's trying to do something every time in different branch, like in half an hour, hour apart, we'll catch it immediately because we're looking for that stuff. And the branch managers or something will not see it. We did like for supermarkets and like, you know, drugstores and a lot of times they have so much data that it's very hard for them to cross-dissect, like inter-branch.

It's like too much data. But if you look for very, very specific, you can see that pattern. So, it's a good opportunity.

And then like, obviously with a mergers and acquisitions and like buyouts and things like that, like going in, because usually it's trying to get the price. But I'm just wondering for like the smaller buyer, are somebody that's looking to buy a company, like does it deal with that or do you only stick to the big kind? Like for example, you go in, you get all the information. I mean, I've bought buildings, I bought loads of things.

The buyer always knows more than, you know, the seller always knows more than the buyer. And they'll always tell you, you know, inflated everything. Like, is it possible to actually use it for something like that, that it goes and analyzes everything and then says, no, it's not worth like a million, it's worth 800,000 or whatever it is.

Yeah. In two different ways. One is if you're given some numbers, let's say an investment company wants to sell you something.

So, they give you like prediction of numbers or to give you some past numbers that they did. So, you can check them statistically and see bullshit numbers. Did somebody just write them down or did they actually occur? And there's a different, like a statistical test to do that.

In Excel, if you just paste the data, and you can see that. One of the simplest and most famous formula, it's called Benford's law, about how numbers are stacking naturally on top of the other. And you can get online an Excel sheet and you paste your data and it tells you like what's the chance of that actually occurring in nature.

The second thing is when you look at bidding data, for example, you look at an area and there's like bidding and pricing of different like apartments or buildings or things like this. Then you can see where there are market forces that are artificially inflating, like what kind of trends are like pushing the market. So, sometimes there'll be like attacks on specific areas to get the prices down, to get the prices up.

They started to do a lot of advertising with like manipulative pricing to try to push prices up or down or do something. So, you'll be able to spot these attempts that are not natural, that something is happening there that is very different than other neighborhoods, for example. And it will not tell you why, but you know that this area that they're trying to promote now is heavily manipulated.

So, it's better to wait or it's better to understand why. And the same thing with investments. Sometimes everything anybody knows is already in the price.

But if you can see that some of the traffic of selling and buying is manipulated, then you know maybe somebody has inside information, or maybe somebody is trying to do like a currency attack or like some other attack. So, there is a body that is doing something. It's not natural.

And you can decide, do I want to ride on that? And whatever they're trying to do, they want to make money. So, I'll just buy what they're buying when they're selling. And hopefully, I'll make money because they know what they're doing.

Or just be afraid of that and say, okay, something unhealthy is happening there. I don't know what. So, like the choice between the cheater and the idiot, I will not be either of them and I'll just stay out.

But at least the power you're going to asking all the time about the small person, what the small person can do and not just mergers and acquisitions and things like that. And the small person is getting access to a lot of public data. He can understand exactly what the municipalities, the government, what everybody is doing around him, if he just had the tools to analyze it.

Now, POVA does a lot of projects for open source data. We do it even without anybody asking us, because that's how we train our system, to find different things. And after we find things, we also need to know, how do we explain them? How do we verbalize them? So, it's a great exercise for us.

So, we even, we have this thing that if somebody submits to us like an open data request, we will analyze it for free. Maybe let's say an investigator or somebody. So, I'll even tell you that.

If in some of the things that you're doing and you'll find that there is data that might be very, very interesting to your listeners, I can analyze it and tell you what I find. And then you can share it or, you know, if it's interesting for you. This is one note.

And I know that a lot of people mightn't believe it's true or not, but the chemtrails, because I've known, because a lot of people share stuff with me in different things, but there's apps that tell you when the planes are going across. And usually when it's the chemtrails, they're not seen. But this one does.

And it actually shows you the plane, all the details, who owns it. I was like, oh, this is good. And I could see basically one was in Riga, leaving for a few hours and going back.

And another one was in Prague. And straight away, have you ever seen, say, a plane leave Dublin, not land anywhere and go back to Dublin? So, that's kind of, I assume that that could be analyzed as well because that would put to bed a lot of things, because I know the States and America, they've banned it, but yet it's happening all over Europe. I made a friend living in Romania.

They had the fog there for like over a year. So, you know, they're actually attacking people's health and something like that. So, I'm assuming that could be analyzed as well, given that now we've got apps that actually go, hey, this flight is here.

And it's simple because like if a plane is taking off and going back, we're not talking about someone taking a flight lesson. We're talking about that it's going in a good distance and then heading back. It's exactly the same thing.

So, if you only had one flight, you immediately see that something is wrong, but because you have millions of flights, nobody will find that specific flight unless they're looking for it. So, this is exactly what POVA does for ships, for flights, for cargo flights, for whatever. You can go and ask them.

So, there is something I really like about conspiracy theories. They call it conspiracy theories because they're very, very good. And a lot of them are very plausible.

But the problem is they require a lot of people to keep secrets and people are not good in keeping secrets. So, but yeah, a lot of these things, you can check them in the data and you can see all these, let's say abnormal flights that people, they report. A lot of people, you know, they see patterns and our brains is very good in seeing patterns.

Like we have an internal POVA to see patterns, but also make up patterns. We can hallucinate the patterns. So, this is also something about our brains.

So, sometimes we can find out something that is very hard to calculate, but sometimes we can just find something that is not really there. I don't know, but the only way to know is to test, scientifically test. And the good thing about POVA, it's not just an AI system that gives you, here's a list of 50,000 anomalies with no explanation.

But each one, each thing that POVA says is problematic to give you exactly which algorithm says and why and in what terms. You can replicate that. You can say, you can go to Excel or any other tool, look at that specific flight and see the parameters and see the, you know, see the path or see the, you know, you will see it, you'll understand why it's problematic.

The only reason we have so many algorithms is just to let the AI understand without you if it's problematic or not. But once you see it, you will understand because your brain does it automatically. Excellent.

This is being fascinating. It's a very powerful tool for people, yeah, for people to use it. And I'm really, you know, also thinking about our call right now is I need to make a way for ordinary people to be able to use these kinds of tools, you know, without having an enterprise agreement and, you know, paying a lot of money and things like this.

Because, you know, scientific tools are just algorithms, they're just mathematical, they can run it on their own computers. You just have to give them some kind of a, like a package that they can run or run for them on public data if they request it. But yeah, this can be done and, you know, it's a lot of fun.

Not everything is business and, you know, it's really hard to come to explore data. Yeah, and I think, you know, if something's like that provided, then you've a lot of people that's kind of, everyone's got their own issue in their own city. But like when it's done properly, because like even, you know, there was one of the things before, like on the roads, you had the sound barriers and a company that didn't, Concrete Prefabs, it meant that their ones would last 25 years and they weren't even double the price of the kind of plastic ones they use.

But the ones that they're using every five years, they need to be replaced. And, you know, it's like major way. So everybody is seeing this kind of thing.

And I don't, I can't, I didn't get help because people feel, you know, there's no way of stopping this. And like what you're saying, what you're able to provide, it kind of gives hope to humanity. I'll tell you something.

It's very hard for humans to hide from data. And this is sometimes can be scary also because, you know, we analyzed once the log of a parking lot in a big industrial area. And within a week we found, you know, all the cheating couples.

Everybody that leaves together before lunch and comes back together after lunch in specific days, specific times. You know, when it happens once, you cannot tell anything because people go in and out all the time. But when there's a pattern and people behave with patterns, and that's a problem, you know, with fraud, that's a problem with, you know, bad operations.

It's people create patterns and they don't even notice it. You know, one of the things, if you want to tell somebody how to stay safe, as they protect themselves physically, you tell them never go out in the same time, never ride, don't ride often in the same car, don't ride in the same route, you know, always make variations, random variations. But people don't know how to make random variations.

They will keep doing the same patterns. Also, you know, I was doing martial arts for many years and, you know, people sometimes they do paint patterns like katas or whatever you call them, depends on the martial art. And when they actually get into the fight, they are so stressed, they start doing these kata moves because that's what they know in their DNA how to do.

And that's very predictive. So once they start doing that, they're dead because you know exactly what's going to follow because it's like a rhythm. And they do that when they're stressed.

So when you have to train a fighter, you have to untrain them from doing all these like ceremonial katas. It actually happened. I did kung fu for about eight years and we used to do it that if I was training with you, I'm trying to hit you.

And if you don't block, you get hit. We had a guy, he was a black belt in karate. And when he joined us, he said he got into a fight, screamed and stopped right in front of your man's face because that's what they were doing all the time.

And your man beat the shit out of him. So that's why he joined. So exactly what you've just said.

Yeah. Yeah. So people are training themselves to work in patterns and that's like the easiest for them to like be efficient, but also the worst thing for them to be or undetected or, you know, we're not good in that.

We're social beings where we fall into patterns all the time, which means that we can learn a lot from data. Yeah. Well, listen, this has been fascinating.

Is there anything else about the company that we didn't cover that you'd like to mention? No, I just wanted, you know, people to be more aware of kind of work that can be done with data. And I can tell you something, you know, that might, might be a bit shocking, but you know, most people's phones is selling their location like a hundred times a day from the browser, from different apps, anything that has location services is selling your location to whoever. But the only point is that the only way that they can use, they can do that is because it's anonymized data because they're selling somebody's location.

They're not saying it's you, but different ways to know it's you. So once you, you're in specific spot, you know, you're the only one there or you identify yourself by some way, then you're not anonymized and they can trace you back years, everywhere you were at any point of time. Who was next to you? Who did you meet physically? Who did he meet physically? So all this information is out.

Nobody can get it. It's not a CIA, NSA, whatever only you have there. Anybody could buy this anonymously and analyze it and know anything about anybody.

So there's a lot of different things. A lot of our data is there all the time. I think the only advice I can give people, if we start talking about cyber is that don't be interesting.

That's the only thing. Don't be interesting. So I hope I'm not doing a big mistake here by being a little bit interesting, but... No, no, no.

Listen, thoroughly enjoy this and you might let the listeners know where they can find you. Well, they can go to pova.ai to have a look or just find me on LinkedIn. And, you know, anybody that wants to start a conversation about one of these topics, either from the small person side that wants to, not small, but, you know, less empowered person that wants to make sense of some data around them or organized patients that want to be more in control of what's going on and better understand the patterns around them.

Let's talk. Let's speak. Excellent.

I'll make sure I post both the website and the LinkedIn on the audio and the video. Thank you very much, Roy. Thank you.

No problem. So that's all for the awakening, but also the speaking, because I think this is relevant. There's a lot of value in this episode.

So it's going to go to the two of them. You'll find everything about me, scan the QR code or go to RoyCoughlan.com. If you're looking for virtual assistance, go to va.world. Until next week, take care.