So I'm here to hopefully wrap up the day with a little bit of a story as a talk about how one of the companies that we started collect AI, where we decided to use AI, how we decided to kind of get it up to started. We are in an environment where we have this large conglomerate on the one side, which has about 12 million in turnover a year with everything from retail services or retailers to logistics provider, to financial services, providers and auto group digital solutions sits over kind of on the let's say on the edge of that. And we are constantly looking for new businesses to create new business models to test in the last six years. We're now at our 16th business that we're starting. And one I'm talking about today is collect AI, which is a company that services kind of everything from the post sales key invoice, all the way through to that collection that's market.
And when we started this company, we looked at it and said, well, we have the auto group of all of these retailers. And particularly in Germany, we have a lot of the retails retailers do their own all professional. So they basically, they're sitting on massive amount of invoices and credit every month. And on the other side, we had AOS that is basically a larger debt collections company inside Germany. We said that was for us kind of the starting point to say, well, we have a lot of invoices flowing through the system. We have a lot of data that we can potentially use. And we have all this expertise touching on everything from collections, regulatory, all throughout Europe through with let's say even the manual operations call centers, things like that, that go into those processes. And that's where, when we decide to start the business.
So we took a look at, and we operate in the region, not just inside Germany, but we said, well, this, this market and of south 60 billion euros, that's interesting enough for us. So let's go start this business. We were fortunate enough that in the first year with the platform that we built, we were able to handle 40 million euros worth of invoices that went through for both Dunning and debt collection. And with kind of the initial version of the platform, we realized basically a 33% increased collection rate for us. That was how quickly could you get the money back from the consumer or whoever owed you the money. And we decreased cost in comparison to let's say, letter coal, et cetera, by about 41%. And when we started this company, we basically, we were, we were dealing with, and it wasn't just inside the conglomerate. It was also when we went and talked to banks or other retailers or other people that had credit and, and were trying to go through these do processes by and large, they were doing letter letter, letter letter, and they might be going seven months by sending different postal letters where they just increase some fees on it, say, wait, you forgot to pay.
So here's five more euros and fees, but it was basically from our perspective, it wasn't really that exciting from an AI perspective, because all we have were these, they sent a letter, maybe it got paid and our whole value proposition started with, we have all these digital channels. You have emails for the consumers. You might have their mobile number. You might have WhatsApp, you might have line, you might have library. You might have all these different things messenger that you could use to communicate to the consumers. And depending on the demographics, where they live, cetera, what they're doing time of day, there's all these other opportunities to reach out to that consumer at an appropriate time to get the, the payment for the outstanding credit, as opposed to just sending the letter. And then just hoping that the letter shows up and that then the person sits there and on a Sunday morning, then logs into their online banking and pays whatever that bill is.
So that in mind we knew that we didn't really have any kind of, let's say, compelling data set. Even though we had all these invoices inside the group, we had all this expertise in the group. We still wanted to use AI. Cause we said at the end of the day, that's how we optimized this process. And what we did was we took in particular, a type of reinforcement learning subset of AI. We took the math and we broke it down into individual components. And so what we did was we said, no customer that we walk in and talk to is gonna say, I want to throw this into the black box. And I want the black box to do everything for me. We basically said the customer is gonna say, well, I understand if you break this apart and tell me that the AI is gonna pick you for the best possible time to send that communication, we understand that there's then called it agents.
So then the next agent's gonna pick the best possible content for that communication. We have been the agent that says, oh, we're gonna pick the best possible communication channel with this content and timing and whatnot. And also as kind of a, what we've learned from our consumers or from our customers in the German market was a lot of them wanted to optimize fees that they were charging late fees. So some of these, these departments in the back office, they're actually optimizing and operating as profit centers at the end of the day for some of these large corporations. So they can tack on five euros or 10 euros or something onto the bill. And 80% of people pay that that's a big win for them. And so they're incentivized to actually collect these fees, whether whether or not you're agree with that. And so we said, we'll break it down into these individual components.
And that works very well to take basically somebody who has just set letters, not nothing that, that you can really do anything with in terms of the AI and let the AI start to go from, you know, the way I describe it is, you know, learning to crawl and learning to walk, then learning to run that as it goes and has the flexibility to experiment, to send an SMS today or send an email tomorrow, try different content, try different timings, try different fee structures. And what we did was we take our customers from that initial point where we say, look, the AI's just gonna start with something small. That's easy to understand time of day, the message gets sent. It's gonna go time of day. Plus which communication channel. Now it's gonna go with the content and now it's gonna go where eventually we can get them to a point.
I would admit nobody's at the point where they say, just dump everything, the black box and let it do what do what it's gonna do, but we can then take other components of the decisions, the decision that's being made and let the AI then through those components as well. So we can even offer different payment plans, or if somebody looks at a particular set of payment plans and then decides, well, those aren't necessary to the payment plans that I'm interested in, they break off the session. The AI can take that feedback and then say, okay, I understand you looked at those payment options. Those were acceptable payment options. Here's this plans VC and D is one of these acceptable for you and all the feedback points. Cause those things are digital. We can take all those feedback points even back into the system. And we can see that, well, if you send a particular type of comment or content to a consumer, that the reaction is gonna be negative or that that results in not being paid or it gets questioned, it's not clear those, all those individual digital feedback loops can be put back into the system, which is much more beneficial than let's say the logs that some of the even debt collection company that we have in the portfolio, they, they say, well, we made all these phone calls, but they didn't necessarily record all those phone calls.
They didn't let's say necessarily record all the results or they were in several different places. And, and it was very difficult to get any kind of, let's say, relevant data put together to train anything. And so for us, the, probably the best comment that we got was, I guess it was about a couple weeks ago. And one of the banks that we deal with said, Hey, this, this is working a little too well. And we said, well, what do you mean exactly? They said, well, you know, we're not exactly sure what to do with all the employees that we had sitting around that make all these phone calls to chase down consumers, to try to collect the payments. Cuz the digital solution is actually working much better than anybody ever thought. And they're, they're only at the initial phases where they're just talking about okay, time and channel and content and things like that.
They're not even letting the system make the full spectrum of decisions for them. And that for us was the assign where we said, okay, we took, we took something where a customer had no data. We were able to dump three, six, 10,000 claims into the system that they were working with. And already within a really short period of time were already able to see even on just the digitization that consumers or our customers were, were getting a lot of value outta the system. At the same time, they were starting to see, let's say optimization happen in a very short period of time where then they understood what was going on. They, they could see the results that, okay, the system is saying, the SMS needs to go out at 9:00 AM on Tuesday or the email needs to go out at Wednesday at, at 4:00 PM or something like that.
So they could see those tangible results. You spoke about the have to be able to open the box. And for us that's part of our customer journey is that they may not have any idea where they stand today. They have a percentage that's in a spreadsheet somewhere. They know they send the letters out, they know what that costs. They know what they get on the fees and that's about it. We have to take them along kind of a journey that says, well, we're gonna start with that where you don't have anything. You don't need any data that you have. And the system will then learn to crawl and walk and run and specifically tailor itself to the kinds of demographics or customer groups that you have. And so that was a very quick brain dump of what we, what we did at collect AI a little bit different because we didn't actually have a, a jumbo set of data to work with bootstrap or train algorithms. But they, they, they have done a very good job over the last 18 months of taking customers from zero or nowhere, to a point where they actually can see tangible results of the bottom line.
Okay. Thank you very much. Do anybody have, does anybody have questions? Well, no questions. I got one which is, have you tried this idea in any other geography?
Yeah. So the, the company actually operates globally depending on the regulatory environment. So if the, so if it's a, if it's a creditor that's in Germany, we have couple of these, but then they're the people that they loaned money to are global one 'em was a Bitcoin related kind of lender then legally or regulatory-wise we're able to operate globally for those consumers send SMSs to Argentina or whatnot. We have customers in, let's say the UK, we have pilot in the Philippines, I believe right now that's getting up and running. So the, the, the value proposition of saying that there's a very unsexy back office process of trying to go run down consumers and collect the money is now being digitized and more efficient seems to speak kind of cross boundary, cross border that, that everybody goes, oh, that, that seems like a no brainer. Why are we doing that? So we've been pleased to see that we, we may have started in Germany, but then right from the GetGo, we went at least regional time
International. Yes. I asked that question because there may be in different jurisdictions, different limitations about what you wanna do. Yeah. And that, that, that can also be problem, need to tune it perhaps.
So, so, so one of the tactics that we use for that is we basically kind of, I call it, define a box. And so if a regulatory environment, for example, would say, well, you can't send an SMS to consumer at midnight, then there's a rule put in place that says the AI can still try to make that decision, say, wanna send the SMS at 1:00 AM, but then it gets feedback right away saying, no, you're not allowed to do that. And so based off of then where the particular consumer is, ends the, let's say the, the origination of the credit, we have to put these boundaries in place. For that specific reason, we had interesting enough as, as an American, the, the only place where we had anything, where there was even a question come up, was the us where the, a consumer said, I don't understand what this is and no German can come talk to me about paying back money ha high. And on the other side of the Atlantic. And then we said, okay, but, but we have a collections company there in the us, we handed have any information and then problem solved. But it is an interesting environment cause of
That. So is at what point do you come over to?
So it's, it is a little different depending on which country it is and what the, let's say, the regulatory environments, there, there comes a certain particular point. And in Germany, I would say, as soon as you start to go towards, you are having to take it to court. Then there becomes a point there's cutoff point where the, the software can't go argue to a judge, or it can't, it can't do the bureaucratic processes. There are certain things in Germany, which are automated, at least with the filing of those things, but you still then have to have a, a person there. So on, on, there's a little bit of a handover phase where we do have a couple of operations people where the technology will say, okay, we're gonna hand this off to the court system now, cuz it's, it's, it's just, it's at that stage and there's no, no way around it. And those people will support that process once it gets submitted. And then at that stage, it usually gets handed off to somebody, but then they have 60 years of experience and the exact lawyers and whoever that, that need to do that. And most of Europe we're fortunate enough that the auto group has a collections company in most of those countries. And so it's a very smooth handle off for
Us. So would you, would you say that you're going to reduce through this, the number of cases get to that point? Cause then it becomes
Yeah. Expensive. One of the things that we did at the beginning was we said, we're actually providing a customer service as opposed to collections, cuz merchants are potentially throwing away good consumers just because they throw 'em into that collection right away. And we learned that people in the accounting departments and whatnot didn't really that they didn't care about that. So we, we dropped that. But what we noticed was because we get about a 33% increased, let's say speed or overall collections rate on the things that are in the Dunning process. That means a third, less things are ending up in collections. And what we've actually seen was the, the German, let's say sister company, that is the collections company. We have several of, let's say their customers in the group that are using our technology on the Dunning side. And they already, after about four months of the technology being in place started to kind of complain that they were getting less low hanging fruit where they would, they would be able to go over to a consumer and say, well, now it's at a collections company, you have to pay us an extra 30 bucks.
That was all disappearing because our technology was telling the consumer while they're sitting on the subway or at home or whatnot here, click on this link, resolve this payment. So that way it doesn't go to collections. That's that's part of the goal or our whole philosophy is, is why I treat the customer kind of like crowd and send them to a collections agency when it would be very easy to use the digital channel to reach them in advance.
Okay. And having gone to the point where they agreed plan, whatever, all those plans is there evidence that they're,
So yes, most of the plans are relatively shorter term or we're not necessarily talking about like a 10,000 Euro loan or something like that, right. At this stage, most, most of the merchants and or the banks that we're dealing with, it can be anywhere from 50 euros up to maybe a couple thousand. It is working. The part that I would say is it's, it's, it's, it's a little too early for me to give any kind of statistics to at least at this stage where I would feel comfortable doing it. But what we've seen is is that once we realized that the payment wasn't made the system then fires off the next, let's say digital channel communication to the consumer to say, well, for whatever reason, if it was direct debit, that was, let's say already set up and it failed. We can reach out to the consumer and say, Hey, before this becomes a problem, could you pay this installment? And, and, and that seems to then catch people before then they fall back into kinda default on these things. It doesn't catch everybody. Right. There are gonna be snares where it doesn't, but, but it's working
Very good. Well, thank you very.
