Which will give a short introduction. Jack Berger as well, who is the Director of Software Engineering at Otovo. He will give a short live demonstration of the Endurance system. Afterwards, we will end the session with the Q&A, and if you wish to ask a question, please do that in the Q&A function here in Teams. With that, I'll hand over to John and Jack.
Thanks, Albert. Thanks for having us. It's an exciting time with the technology change brought about by artificial intelligence. You hear a lot of things about what artificial intelligence is about and what it can do in the future. Some being a little scary, most being pretty exciting, especially as it relates to businesses and especially service businesses like what Otovo does, which is behind the meter for both commercial and residential, we solve and provide maintenance for power assets, for specifically solar batteries, generators, EV chargers, and load managers. At first, I'm going to go through and talk a little bit more about Otovo and investment thesis, and then I'm going to turn it over to Jack, who's going to take you through a live rendition and walkthrough customer journey, if you will, with Endurance.
This is all live, so if there's anything goes wrong, you know it's live, given technology. Also he's going to walk through including our Pulse system, which is an Endurance subset of HR, maybe some others as well, if we have some time, and love for questions. Please ask what you're thinking about, because if you're thinking about somebody else is most likely as well. Appreciate that. All right. First of all, Otovo at a glance, this is all Q1 information. Obviously, we have earnings next week, so we will not be discussing in any way, shape, form, or fashion, anything financial or change up some of even the major metrics. Largely, we have a very and big market, both Europe and the U.S. Over 37 million installations for both commercial and resi have some sort of power asset with them.
Solar batteries, generators, EV chargers, load managers, et cetera. Roughly, and this number is climbing, about half of them, certainly we've got here on the deck about 13 million for sure, are what we call orphans. That means that there is no service provider, and that's important. I'll get to the investment thesis here and what we're seeing happen in terms of our sales and so forth. When you look at what we can do with Endurance, what's this really all about? It's about increasing margins and not really necessarily the gross margin of our labor technician going out and solving a customer's problem. It is more about the net margin. It's about cutting operating expense. It's about going in and slashing the SaaS cost, slashing the SaaS staffing. What does that mean? Software as a Service. What does that mean? Salesforce, HubSpot.
These are the kind of companies out there that you've heard about in the so-called SaaS apocalypse, and we're here to tell you the SaaS apocalypse is a very real thing. From there, Jack will take you through an agentic AI. What are we doing with agents to displace other costs within the system and to enhance customer service, so to provide a better customer service? What we've done so far, I think you'll find pretty amazing. What we see in the coming weeks and months, notice I didn't say years, weeks and months, I think is really interesting. You can see the lift there in margins on the first page. Consolidator. This is a very fragmented space. If you look at, there is no big global service provider. Frankly, there's no big one U.S. service provider or European.
Now there is with Otovo. There is a lot of fragmentation here, and we'll get to that here in the investment thesis just in a few seconds. You can see the cost savings, the legacy customer base. We have access to near 2 million customer information and so forth. How many acquisitions? We'll close our eighth acquisition. We have more right behind that, as everybody knows. It's in a very exciting space here as far as large fragmented needs technology and frankly needs a better performance and level of service that customers can't get out there. In terms of going through and how does Otovo make money? We make money through memberships, so both residential and commercial. We are monitoring your power asset, think solar system or batteries. We make money on the repairs.
Something's wrong, we roll a truck and fix it. We make money on upgrades. Think you need a battery. Say you're in the Netherlands and net metering's ending at the end of the year. You're going to need a battery. We can do that as well. Retail power and then aggregation of power and selling it back into the wholesale system or what we call virtual power plant or VPP, and that is in the deregulated markets. Currently, right now, we only do that in Texas, but we're going to work to launch that out in both the Northeast and Mid-Atlantic U.S. and all across all of Europe, because all of Europe is actually more deregulated than the U.S., and enjoys a better marketplace and better pricing for consumers as a result.
On the next page, before I turn it over to Jack, just give a quick overview on investment thesis of Otovo. Otovo really sits at the epicenter of three major investment thesis. That's power. We all know about all the AI demand with data centers and driving power demand and so forth. We obviously serve customers, both commercial and residential, for power. We do so with many of these assets at far below the current prevailing utility rate. At a better reliability because that generation is on-site and matches with load. There is an enormous amount of need to keep power assets running and running optimally now, given what's happening in the power markets, given demand pricing moving materially higher and expected to do so for the next several years at least.
We have a nice investment theme here with addressing this secular mega trend with power demand. The second is service itself. Service has been very challenged in behind the meter, especially solar in both Europe and the U.S., for both commercial and residential. What we're able to do, and one of the theses here was to essentially take a very fragmented market, be able to aggregate a bunch of companies that were profitable, then when you start to look at getting economies of scale and cutting out costs and so forth. Being able to serve big customers like major manufacturers of equipment, major asset holders, these are things that have really performed very well.
We've done very well on buying service companies at incredibly low multiples, further cutting the cost out of them, then doing a lot of transactions and signing increasing more contracts and getting more business from very large manufacturers, very large asset owners that have nowhere to go. They don't have the scale. The last thing a big manufacturer wants to deal with, is to say they have to deal with a bunch of hundreds of small little contractors out there that frankly are really installers and not the service business. That's two very different businesses. One's a construction business, one's a logistics business measured in hours and days. The other is measured in months. This year there's an enormous amount of demand. There's been a great amount of execution. Can't wait to talk about that on the next week's earnings call.
This has been a huge upside in terms of starting the company, doing these acquisitions. This has gone much, much faster, much better, much cheaper than we thought when we originally started. The last one here, that's what we're all here to do this afternoon, is to talk about AI. This was something that was obviously not well understood. Certainly, I can say that not understood by me that well, but really was able to bring in Jack and his team to be able to come in and say, "Let's lean into what artificial intelligence can do. Let's lean into what the large models can do, and what could that look like?" What you'll see here is the result of that. Again, building our tech stack from the ground up. We own the entire thing.
Displacing that software cut costs, cutting the software support staff, then being able to easily deploy very inexpensive, very powerful agents. As those models exponentially improve, and it is exponential, two weeks is a long time in AI as I think everybody is aware of at this point. That further has cost cuts to it. Down the road, who knows? Embodied AI, there's Elon Musk and a lot of others that firmly believe that that's right around the corner. I have no reason to doubt that. That's something that we can talk about at a later date, should that come to fruition. Exciting to see what we've been able to accomplish as a company. Quite candidly, it's hard to still get your head around how fast AI moves and what you can do with it.
Really what gives us a real advantage here is that we built the company up. We are what they call AI native. We force the AI to be injected in the organization, restructured the organization, which is helped by pulling in a lot of the different acquisitions of the companies. This is not easy. If anybody tells you this is easy to adopt, they don't know what they're really talking about. Quite candidly, this is very difficult. Pulling this off, we have the advantage of being a leader, started ahead of time, so we got a head start, but also being able to build this company up with all these acquisitions from the ground up with AI first and foremost in mind.
With that, I'll turn it over to Jack, and he can talk a little bit about his background and so forth, and get straight into showing you Endurance. Jack?
Hello, everyone. I'm Jack. I'm the designer and implementer of Endurance here at Otovo. I'll get right to it, and show you all exactly what Endurance is all about. This will be a live demo, just a heads up ahead of time. For those of you in the technology space know that this is incredibly rare to do, primarily because of the usual risks involved with doing something like this. We're confident enough in our platform in order to do this. I would say that this is a fairly rare occurrence, to say the least, for most people. Good. This is the main dashboard you'll see as you enter Endurance. If anyone's wondering about the dude on the ship, that's our marketing technician guy, and Endurance is Shackleton, the explorer ship. We have a little bit of fun with that.
I'll be showing you a couple of different systems today in Endurance. Let me see if I can head on over to my account here in Endurance. You can see here we prioritize speed a lot in our systems. We're building everything from the ground up ourselves. We want everything to be very quick, very snappy in our systems, in order to be able to scale appropriately. As you can see, we have all the customer information here, any of our tags. This is my test account, so you will see quite a lot of Kind of junk and test items here as you're looking through this. You'll see all the tags from different campaigns that we're running or anything like that. We at Otovo have a bunch of different systems all built into Endurance.
Often where we're going, we don't necessarily always have the license to be able to operate as Otovo in those regions. For example, in California, we are not licensed. We've purchased a company, EnergyAid, to work in California, and we work under their license, and we are transferring their license over. For the time being, we have to operate as EnergyAid there. How do we keep track of all these customers that we need to say, "Oh, hey, we're EnergyAid," or, "Oh, hey, we're Otovo?" It's really, really difficult, especially on a global scale. We've built a brand ID system in here. You can see Provider Otovo. In this case, we are licensed in Texas, where I live, to be able to operate here.
We bring in all of the customer's information from the phone networks, any of the customer's mood, their preferred channels that they want to operate in. Any of our consent tracking, anything like that is all in here. We have our full marketing engine directly into Endurance here. Any of our sales status, we have a central ledger for everything here, where we have our first touch, last touch, any of our sales, and inbound calls, outbound calls, all throughout the system. Of course, mine has quite a few in there, in order to tune our marketing engine and actually get a real customer journey all the way through the system. Another important part of Endurance here is we need all the data all the time into Endurance in order to be able to actually give a really, really good holistic experience.
I think part of the thing with AI in general, and then also, of course, in Endurance, is making sure that you have ownership of every single point of your data. That's been very important for us, is that we have every single thing in Endurance. There's nothing outside of Endurance. Any quotes to be paid here, I'll show this in a little bit. We work with Stripe and Ramp as our really only outside partners for primarily financial reasons and regulatory reasons. Stripe will handle our PCI compliance, and Ramp will handle our actually issuing our credit cards to our technicians and things like that. Both are very, very tightly integrated into Endurance.
All of our information on view in Google Maps here, and any of our answers to our questions about if we're okay to proceed, if not home, or anything like that is all stored here in Endurance. Any open customer cases, and any of our activity in here. On the sidebar here, you're going to see our central communications system with the customer. Now, we have this sidebar available pretty much everywhere in Endurance, and the idea is that this is a full holistic view of everything that the customer has experienced. All of the outbound calls, inbound calls they've had, any of the transcripts, any of the call summaries, the dispositions, the recordings, all of it is immediately available right here. You'll see in here, someone named Iris. Iris is not a real person. This is your first AI agent that we put into Endurance.
When we set out to build Endurance, of course, we set out to build Endurance with AI. That wasn't the primary system of putting AI into Endurance. We set out to build a really good CRM first and then get to the point in which we put AI into it. We, for the past couple months or so, have gotten to that point. That's where we see Iris being able to schedule appointments, as you can see here, or notify appointments, confirm appointments, set up a checkout for a subscription, for our membership, which if anyone on the call is in one of our service areas, check us out for membership. Give us a holistic view of exactly everything that's going through the system. Any files needed for permitting or anything like that, all available at the click of a button.
I'll show you one of my services. We've got EV charger service here. This is the start of the service. This is, you create a quote, we need this customer to accept it. In this case, it's me. I did change a little bit of this before the call, all of our inspection reports from prior systems, if we need to create a change order on the quote or anything like that, any context that we need for the system, anything like that, our AI will inspect the inspection report and get an actual overview for the quoting team right here. This is the second part, is we're actually starting to pull in This one's a bit older than our newest version of the system.
All quotes that are inbound for our system now, all B2B, all B2C new quotes are made by AI for us now. We have about a 90%-95% accuracy, given the correct information submitted by the technician on site for the request on this. That is a reflection towards our actual estimating team. We've been using this. We saw those numbers, we were very happy with those numbers, we went ahead and changed our entire system over to that. All of our stuff is first intaked by our AI estimating and quoting team. Any of our commission that we need to pay out for the sales team is also brought directly into here. Also, our sales tax, if we need external financing, if it's covered under warranty, linked to external accounting systems, all brought directly into the system here.
Let me show here. We actually had a very unique problem here at Otovo that I think is a little bit interesting. A few months ago, one of our salespeople was complaining that we had a customer complaining that all of our quotes seemed like AI. At the time, we actually didn't have any AI in the quotes when we were presenting to the customer. Right now, you're looking at the customer view. This is where an example of what we're using AI to resolve a problem that you just wouldn't really think of. It was because their quotes were too technical. The customers couldn't understand it. They didn't know what a 40-amp breaker meant. They didn't know why a type QP breaker was important versus a regular breaker. They just didn't understand it, and they didn't really care to understand it.
They just wanted their stuff fixed. That's where we have actually brought AI in here, and we've not gotten a single complaint afterwards, since we've built this system of our quotes sounding like AI, which is a little bit ironic, but very interesting to me at least. We have a little breakdown here, then we also bring our eSign system directly in here as well. This eliminates the need for DocuSign. This eliminates the need for any eSigning system. We put it all directly into Endurance, also you eliminate the back and forth where like, "Oh, hey, let me send you the quote on Square or one of these other platforms." Then, "Oh, and I also need to send you a second email. It's going to come from DocuSign.
You need to sign that and pay the quote." Here in Endurance, everything's the same. Now, when you've signed your quote, you'll see this here, where we'll be able to add a card or bank account here directly, select a payment, then complete the charge directly on here. All of that is directly into Endurance here as well. Additionally, we have quite a few systems in Endurance. When you're looking at a Salesforce product or a HubSpot product or any of the other different systems that we've ripped out, a lot of them end here. This is your base CRM system. We've got our base dispatch system. We've got all of that. This next part is really where we take Endurance further, is we bring everything into Endurance. It's not just some things, we take everything.
Anything in terms of our call systems here, of course, there's no calls today in the U.S. at 7:00 A.M. here, but we're bringing everything into Endurance. You can call directly out of Endurance. Usually with Salesforce, it requires a separate talk desk, separate air call, then you have to connect those into each other. Oh, okay, they don't sync. We know all these issues because they're the exact issues that everybody that we've acquired has had problems with. We have a dialer directly in Endurance, being able to dial on our outbound leads and anything like that. Again, another separate SaaS. When we acquired EnergyAid, we cut over 30- 45-ish different SaaS subscriptions precisely for this reason, because they had a Salesforce system that was fairly good.
In fact, we got a couple of offers to try to maybe think about even buying their setup of the system, because it was so good at a setup for Salesforce. It just didn't end up making the cut because you have so many different SaaS subscriptions that then get put into the system directly. We have our technician management. We have our clock in and clock out here. Some of this is going to be sensitive information that we won't want to reveal on the call here. All of this is directly into Endurance. We also have an automatic dispatching engine, where we automatically route our technicians based off of our skills and the nearest technician, then load balance it according to what is the best profit margin, especially on the drive time.
Drive time is the worst thing in the service industry because it's purely lost margin. You really want to be able to eliminate that. We have a deterministic system that we've built into Endurance directly. We originally tried to use some of Google Maps and a little bit of their API there, but it ended up being extraordinarily expensive. Again, as you'll see is a common theme here, we built it in-house. I'll move through this quickly here. Legal, any of our issues in Endurance here. We've also built AI directly into Endurance with the context of all of our systems. If you've ever seen a Bolt or a Lovable or whatever else this is, we've actually built that directly into Endurance.
We have each of our team members able to build out their own dashboards directly with AI, with our AI chat system. Our Canvas system, which allows them to pull directly from our database and get any of the data that they need for whatever they need to do. We've seen some pretty creative stuff, or at least I've seen some pretty creative stuff over the last few months. People, that is actually probably our biggest success on new features in Endurance. Another thing I want to show is Endurance is more than just this one system. We bring a bunch of different systems together, especially systems that need individual permissions or whatever else this is, all with the same login. Companies struggle with a bunch of different issues with HR.
A lot of it is managing, like how do we manage a global organization? Because we've got about half and half in Europe and the U.S. That's going to change even more as we're working with Green Panel. This is our solution to that system. Clock in and clock out. There's specific laws in different countries that we have to manage, all of that directly into Pulse. This is another one of our systems under the Endurance umbrella, and it syncs directly with Endurance. The Endurance mobile app for the technicians and for the customers syncs directly into here. If you clock in in Otovo Endurance, it clocks into Pulse here as well. All of our policy documents, any documents that have signed, again, obviously very sensitive information. This is actually live.
I did sign our Otovo whistleblowing policy. Now we've got one of those. Any members of my team, any time-off requests, attendance, anything like that will be directly in here. We will be bringing some payroll systems directly in here as well. That is not there quite yet. We also have service access and system access directly in here. All of our systems are directly accessed in Pulse. We have a API connector to Google Workspace, Slack, NetSuite, Ramp, where with one press of one button on Pulse, we can terminate access to tons of systems all throughout the organization with just one click of a button. There's no, "Oh, man, did we turn off Fred's access yet? Shoot, I forgot. Now he's had access for three months." There's been a couple of cybersecurity issues with that with companies in the news recently.
Here at Otovo, we're striving to not make that an issue. Technician leaves, decides it's his last day, or we do, his Ramp card is turned off automatically with one click of a button through here. No hanger on here. No, "Please remember to shred your card and return it." Additionally, we're building another system here at Otovo with Themis. I'm going to need to log off here. We put additional security measures into Themis. We start building a legal assistant in Endurance here as well. Again, under the Endurance umbrella. This is allowing anyone in the company to email, or with the allow list in the company, sorry, not anyone. The allow list in the company to email the central repository of all of our legal documents, and get responses, usually in the form of just making documents and things like that.
The issue that we've had with AI is a lot of people have trouble with how long it takes, how much time it takes to really get these responses or building out new NDAs. At the very least, doing a preliminary red line before the legal team takes action or takes a look. That's what we've strived to also resolve with Themis here as well. This is our platform for being able to handle this. Themis is special, actually. Themis has access directly, as you'll see in the credentials system here, which again, a bit sensitive on that, but you can see the systems directly. Themis actually has access to its own browser, and it has its own credit card as well.
It's able to submit trademark applications, anything like that, and pay the ceremonial $11, $12 fees to the U.S. government or whatever government you need to pay for the licensing fees. Things like that. Themis is able to operate its own browser here as well. As you can see, I was working on some home improvement contracts, and working with Themis to be able to handle quite a few of these at the same time. This is really bringing down our costs on a global scale, because this is a lot of work to be able to handle all these home improvement contracts. As well, the last one that you saw here is Nessie. Nessie is our AI bot that we use. This is our recursively improving software system directly into Endurance.
It's built off of some lower-level agent loop systems like OpenClaw and Codex, but is uniquely us. This is built directly into all of our systems. This takes customer requests from Iris. If Iris realizes someone has a technical issue, it actually passes it off to Nessie. Nessie fixes it, and passes it back to Iris. As you can see, we're, at least I hope, we're really running away with it on this system with multiple different AI systems with different permissions, different security settings, working directly with each other to be able to resolve issues. I have an approval system here. I have my runs. As you can see, Nessie's running right now with approvals, and I also rate how Nessie is working throughout the system.
Nessie learns over time through my ratings and through my team's ratings in the system to be able to improve the response behavior. We've seen quite a lot of improvement recently just from that system alone. Any approvals that it needs to make new features or do runs here as well. Some things don't need approval, and most actions can happen just autonomously, but we have about 70%-80% resolution on an autonomous basis. Most things that Nessie asks for approval end up getting denied, because there are other, usually customers or outside features that we don't always want to build. This is our way of actually sorting through the system properly. We have about an 85-ish resolution rate with Nessie on all technical issues throughout the company. That includes new features.
That's not just bugs, that's new features as well. I'll open it up for Q&A. Trying to think if, there'll probably be some more items that I'll show you in Endurance throughout Q&A, but would love to hear some questions.
Perfect. Thank you very much, Jack. We have gotten some questions already, but if you have any more questions, you can use the Q&A function here in Teams. Maybe we can start with scalability. Could you give some more color on the scalability of Endurance versus how it was with traditional SaaS providers?
Yeah, actually we've found with Endurance, we still have to, and as we go into these new companies that we've acquired, we get to work with their systems for a few days to a week or whatever else it is. Our median time to transition from a SaaS system to Endurance is about three weeks. We do a pretty good job of actually, we've done quite a few of those transitions at this point, about five of them now. We get to play around with those systems. We find them slow, we find them very, very difficult to scale appropriately. With Endurance, we've been able to modify the source code to directly impact our scalability. All of our hardware in the United States is a one-time cost of about $6,000-$ 7,000, roughly.
That is quite low compared to what you see for an AWS-like system. I think if I did the math, the number on an AWS on a per month basis was about NOK 40,000-NOK 45,000 to buy that equivalent hardware in the AWS cloud directly. Quite a bit of expense otherwise not used. Endurance in general is very scalable. We can edit the code to be faster. We actually did have an issue at one point a few months ago where we were really starting to feel the strain. We were adding all these companies. Endurance was slowing down. It was an issue. In just a few days, we were able to completely rewrite most parts of the stack underneath to optimize them, and we've been running ever since just fine. As you can see in Endurance today, it's a very quick system.
It's very scalable accordingly. We have a lot more fat to trim on that. I bet we can get a lot faster. Endurance is the fastest CRM in the industry.
Yeah. Also, you recently acquired EnergyAid. Did you experience any hiccups with the integration, or did it work without any issues?
There's always going to be some issues. Nothing ever goes according to plan. I would say a lot of times too, what we do with Endurance in general is, okay, what is this company doing well? Because when you're acquiring a company, they're not doing everything bad, right? They figured some things out for sure. Every single time we take a look at what their processes are doing, take a look at our processes, and then merge them together into the best possible process. Every acquisition, we get significantly better in Endurance. The great part about controlling your own software is your company's processes are built directly into the software. They're not like, "Oh, hey, you're supposed to do this in Salesforce," or "This is the way that you're supposed to do this." It's like, you can't do anything else except for the process.
We got to learn a lot from the operations and sales team there and build that directly into Endurance and improved our numbers dramatically. Each acquisition is a learning opportunity for us, and that's how Endurance evolves over time, is how do we get better with new people entering, new suggestions, and things like that constantly.
I want to add in there something, Jack, if you don't mind. One thing to look at is when you look at traditional SaaS stack, and you've got somewhere between 25 and 100 different SaaS providers in your tech stack, pretty much all of them have a per-seat cost to it. As you're building up a company like we've done in a very short period of time in a large company, your costs explode. This happened to me in a previous company that I built, and it turns into tens of millions of NOK a year. Trust me, and those who are running companies and know about it, you'd see that. You don't have that here. All-in costs, all-in total, even with tokens, we can talk about that, is about NOK 1 million. That is very scalable.
In fact, that is inclusive of bringing more agentic AI, so agents within the company to further cut the costs out. When you look at, for instance, 50% of customer solutions or service today is done by Endurance 50%. I mean, think about that. Think about the cost structure. What is that going to be in a few weeks? Certainly 90 days or so. Materially higher than that, especially as you have the new models that are coming. The cost reductions here are tremendous. We've listed out what we said publicly, it's far greater than that. As you move forward in the next few weeks and months of the back end of this year, you're going to see a tremendous more millions of NOK in savings against both the SaaS that's been cut out.
It's reflected in the last payments that we make under the contracts when we acquire these companies in the first part of this year. That, by definition, will not show up again because we've terminated all those contracts. On top of that, now you're starting to get displacement as far as administrative folks in the office and so forth. We're seeing a very significant amount of cost reduction, dispatch over 80%. We will actually be moving that to almost completion, will be done by Endurance procurement, I mean, marketing and financial analysis. There's so much here that's already being, in terms of cost structure, wiped out, and you are seeing a significant impact on the bottom line and the widening out of the net margins as we went through and have said in our Q1 call.
Frankly, it's unbelievable. That's the purpose of this call is to say, "Here it is, this is real, this is working. Here are the costs that we are avoiding. Here's where we think we can go. Here's our cost to run the system." It's not millions of NOK of tokens and some things that you've heard out there. We'll give you the straight answers as always about here it is. This is what's actually happening. This is real.
Yeah. I'll add onto that. I was just reviewing the numbers actually right before this call. Our handoff rate, which means our rate that AI cannot handle the system, is 17.4% right now. Over 82% is handled by AI on all customer communications. Sorry, on text and email, text-based. We have a pretty high rate there. I will say come Monday, we will be turning off the ability in Endurance for our dispatching team to set the start time and the duration time of all of our appointments globally for service. We will be doing everything with our automated system directly into Endurance. That'll be a big paradigm shift even for us, we're really, really pushing the frontier with that one.
I would say that's going to be definitely a big deal for us, it's going to rapidly give us the ability to actually scale the business.
Just another one here. What are your estimated running cost of the system, and what are your estimated cost savings also related to specific use cases like CRM?
Our running cost for the system, U.S. server costs was the one-time cost of about $7,000, $6,000- $7,000 as I said earlier, and then that's pretty much it. You need the salary. I have one person on my team in California, two in Houston with me, and two in Poland. The running cost of their salaries, and that's pretty much it. We do do some renting of physical metal servers in Europe just to comply with GDPR regulations. It's not like AWS where you're having a VM. The actual metal is what we work on directly there. Those costs come out pretty materially lower from the AWS side.
Great. Then I saw Sanity on the tombstone on the slides. Is that Sanity.io?
Yes.
How do they fit in or rather don't fit in anymore?
Sanity is still hosting our webpages for the last few days actually here. We've been ceremoniously taking them out here. We've built our own CMS system directly into Endurance. We host everything there. We're trying to find the last few websites in Europe that are still hosted with Sanity at the moment, we've already replicated those over into Endurance and are in the process of finding and turning them off here in the next few days. Sanity is going to be ripped out. It was a holdover from the old Otovo European team. We've built our own CMS directly into Endurance as well. Again, the philosophy at Otovo is we build everything ourselves. If it needs to be built, it needs to be built in-house.
Also on your graveyard slide, you mentioned Twilio. How have you managed to bypass a voice or SMS provider, or is it primarily add-ons that you have cut down?
Yeah.
How much have you reduced the expenses?
Yeah. We still use some of their services at a very, very low level, and a low, low cost. Much, much lower than the overall regular cost. Twilio and others, as you know, have SDKs that allow you to bring everything directly into the system, handle your calls that way. We don't use any of that. We use the raw SIP signaling infrastructure, which I get is a little bit technical at the moment, but is about as low level as you can get. It's the actual connection from the phone lines into the internet, so literally the switches. That's built at a much lower rate, and then we've actually built our phone infrastructure to connect that to the human, connect that to any of our customers, ourselves. We're looking into building our own infrastructure there as well.
That one's a little bit harder in terms of a regulatory burden for us. All of their software on the top end, any of the add-ons, anything like that, none of it is used. None of it is not built in-house. Any of our actual attribution, any of our AI phone systems, all of that built directly in-house. We started with some open source repos that we've now significantly forked, and some MIT licensed open source repos, and then we've significantly forked and built on that infrastructure as we found certain things work and certain things don't work for us.
I think, however, on the cost side, we need to stick with what we previously publicly disclosed, and that is about NOK 4 million in SaaS with the companies. There's obviously more that goes with that, with SST and so forth and the additional acquisitions. We've talked about at least NOK 1 million already in savings on essentially back office administrative costs. That is going to explode higher. We'll talk about that in the earnings call. When you look at every time you add a company, there is, and when this goes into talking about scalability, there's no movement up in cost. I mean, a little bit in tokens, few hundred NOK an acquisition or something literally in that kind of number. Jack can correct me if I'm wrong. This is really going to be flat line.
What we've done at Otovo, just forget about the power service business for a while. Just overall, there's always been a relationship between the number of guys or trucks or however you want to phrase it, in the field doing work, whatever the work is, fixing an air conditioner, fixing a solar system in our case, or whatever it is. You have the back office, and that's inclusive of your SaaS, your software spend and everything else to try to make that back office as efficient as possible over the last 30, 40 years. We've broken that relationship where every time you add somebody in the field, you got to add somebody else that does dispatch or customer care or whatever it is. We've broken that relationship completely.
What that is entailed, again, we'll talk about it more next week, is essentially what you've done is you've created a highly scalable. As you add more revenue to it, your actual cost structure stays flat to actually declines, and your revenue is moving materially higher, like orders of multiples. As you buy companies, as you do organic growth, your cost structure doesn't grow. I mean, that's never been done before. We see that that is where you're going to see a lot of competitive advantage, particularly over the next few years. We have a head start that we will literally have a cost structure that nobody else can even touch, and that's where we sit today, and that's what we talked about in the Q1 call. Again, we'll update some pretty exciting updates next week on that.
Perfect. Then over to security. How is the security on this system? What is the risk of hackers invading your system? What if the system crashes from attack or own bugs?
We partner for a fairly low cost on DDoS attacks with certain other companies like Cloudflare to be able to shield us from the worst of that, of course. That is very low cost for us. I think that is about NOK 15 a month. DDoS attacks there. Your biggest risk as a company, in general, with hacks, with anything like that, as you will see, if you go take a look at all the past few hacks, LLM hacks notwithstanding, it is always a human. A lot of IT people will tell you it is always the human in the loop. Someone gave access. Someone who still had access after they left the company somehow gave access to the system. With Pulse and Endurance in general, we are able to end access and control access directly at the level there.
There is no possibility of, with some of the other companies that we found once we go and audit as we go to purchase them, "That guy has not been here for a year. Why does he have access still to critical infrastructure?" That is not great. That is where we really excel there with Pulse. Additionally, we have built all of our infrastructure in-house. Your next real issue is another SaaS or another system getting compromised that then compromises credentials and then compromises your entire system. With building everything in-house, we have control over the security of everything. We have built our own network infrastructure. We have built everything in-house. We know exactly how locked down it is. We can audit every single process, and we do audit every single process.
Every single day, my team takes a look at any security issues, any potential avenues for attack, and we fan out multiple agents across the system to be able to go find them in every nook and cranny. This is not your IT audit of old. This is extremely thorough. Throughout the entire system, we are very constantly aware of this problem. We also have our always-on engineer watching for scans, Nessie on the Nessie control that will monitor and is able to triage any ongoing issues with our system. Even if you stage an attack while I am sleeping, you are not safe.
How many people are involved in developing and maintaining this software, and how dependent are you on those people?
Yeah. Honestly, it was just me until May. In May, we decided we needed to at least de-risk from me building the system in general, and we brought on a couple of really great engineers onto the system. As I said before, we have one in California, two in Houston, and two in Poland, for a grand total of six on the team, including me. Yeah, pretty sparse team.
Let's say you decide to quit Otovo, Jack. How would the system work then?
The team knows the system very well and has done it in my absence if I ever need to do it as well. They're really great. When they first started, maybe not the same answer, but now we have a team that knows the system deeply and knows it very well.
How does the model control for LLM hallucinations?
We have a variety of checks in our system. When you see Cursor, Lovable, or any of these other systems, they only have to build one agent loop. Here at Otovo, we've actually designed over 11- 13 different agent loops for specific tasks. If you're wondering, "Why are you doing that? It seems like such overkill." Because each agent loop is really good at one thing. A big mistake that we made, we actually made this mistake initially, was we want to build one agent loop and it handles everything. Handles all of our controls, handles all of our hallucination controls, whatever else this is. We realized very quickly that that was not going to work. We've now designed the system with individual loops for this. "Oh, this is customer facing.
This needs to block against hallucinations on this kind of data." We have reviewing agents before agent responses go out the door to protect against these hallucinations directly. There's some things where you don't need that hallucination control. You don't need those additional costs. Especially something like quoting, where we have our estimators and our FSMs back those up directly. Okay, we don't need as many controls. Also, that agent loop needs to be able to see our pricing, because it's not customer facing, but we don't want to necessarily show off our exact price book to every single customer that tries to ask, right? Different agent loops for different things and different hallucination controls for different tasks as well.
We have touched upon it, but you have decided to quit AWS and run the model on your own hardware. Could you just give some more color on your take on the cloud providers? Obviously, we see quite dramatic price increases each quarter recently. Why do not other companies take the same decision, and what's your take on this?
When we inherited Otovo Cloud to be clear, Endurance has never, ever used cloud providers. We went from the outset to never use it. We inherited Otovo Cloud in Europe, that did use AWS. When we first got it in October, I think the monthly costs were NOK 7,000 a month. When we retired it in March, it had grown to NOK 14,000. Not exactly an acceptable amount of cost increases. As we were shutting down and archiving the data into April and May, our archive, not our live running of the system, our archive of the system had reached about NOK 8,000. Quite dramatic. The price increases are real, and they will not stop. They will continue to squeeze people because they have a captive audience. I will say it's a difficult thing to do.
My background is in computer engineering, which is a mix of computer science and electrical engineering. I know how to do these systems very, very well. I know how to do some chip design directly even. It's in my wheelhouse, in being able to do the hardware and the software at once. I will say it's difficult. I do think it's doable. All you need to do is look at Amazon's earnings report. What are their profit margins? It's ridiculous. It's like over 90%. At some point you've got to say to yourself, "There's some fat to trim there." We found significant fat to trim.
It is a worry when you do this changeover to your own system is like, "Okay, hey, is this just going to end up being more expensive than buying Salesforce because AWS is charging me too much?" That's why you've got to do the double approach of building your own infrastructure and doing that. We've done that here at Otovo in terms of building out everything ourselves. Is it super easy? No, not necessarily, I would say. It is very doable with the AI agents, and it's gotten a lot easier, actually, for us over time as these agents have gotten better at it.
Perfect. We are starting to run out of time, we can end with a question regarding the migration process from the SaaS providers to Endurance. Could you just give a bit more color on how that process was and what were the key bottlenecks in that process, and how long did it take from using the SaaS providers to your own system?
Yeah. We do about a standard three-week procedure. Week one and a half, we actually don't do anything. It's just like, "Hey, meet the people, talk to them, say hi," give them their Otovo emails, give them their Slack, say hi to them on Slack and stuff like that. Work really begins week two, where we map over their existing processes. As I said before, all the learnings that we can take from them, what works operationally, what doesn't, what can we improve, what do they do better? We map over what their data and what their statuses are. We rip it out of the existing SaaS systems. I will say I've been surprised at how easy it's been to dump existing data from their systems, exporting CSVs and all of our files out of these systems relatively easily.
I don't think that it will be that way forever, as you're seeing, if more companies end up doing what we're doing. It has been relatively easy to do that. We rip it out, we map it all over, build a mapping index, do a couple dry runs, on Friday you log out of Salesforce. On Monday, you log into Endurance and all your data is there. We work over the weekend, move all the data over, shut down the old system. We usually leave people's logins available for the next week to make sure that we map the data over correctly, there's no status mismatch, we don't miss any emails or whatever else it is to Endurance, shut it off after that.
Perfect. We're on time, I will hand over to John for some final remarks.
All right. Well, thank you. Thanks for having us. Thank you, Jack, for walking through Endurance. It's obviously the heart of Otovo, the new Otovo, it's pretty exciting what we've been able to accomplish in such a very short period of time. Think about just the next few weeks and months, as I said in my opening comments, about what's going to be accomplished here. It really is changing the business materially. It's going to change business generally, particularly in the service industries, significantly over the next few months and years. When you look at the financial side of things in terms of what we talked about, we've talked about dramatic cost reductions. We've talked specifically about SaaS software cuts and the staffing associated with that. Not insignificant. Actually improving performance on a software.
For any of you who have ever run a business within a big SaaS tech stack, you know what I mean. Every time somebody comes out with a new iteration, it interferes with somebody else in your tech stack, and that's really what caused instability and security problems and so forth, as Jack went through there. We talked a little bit about this, but service improvement. How do you get customers better served by using AI? We're doing that with having constant around-the-clock, we can answer your questions and so forth. Increasingly, we're going to be able to do eventually all of service in terms of the back office, any sort of questions and so forth outside any rolling of trucks and field repair. How do you run the businesses more and more efficiently so that you can supply faster and faster service to customers?
That's where we're really starting to dig in and be able to figure out how to do that. Productivity, dramatic costing decreases. Significant, obviously. We can talk about it on a forward financial basis next week about the improvement there, but I think it's fairly straightforward. When you look at on a future basis is dramatic customer service improvement and speed and both quality is something that I think will leave us unmatched, particularly as we grow rather quickly, continue our acquisition trek. Looking forward to having more discussions about this. Glad we had the chance to say, "Hey, this is real." Talk about real dollars and costs. We've talked about previously. Talked about answered questions directly about there's no dramatic token usage. I've seen that out there and everything else. Hopefully, you've got a sense of what is happening out there.
We are in the cutting edge of AI, but we, as well as others across other service industries, are going to have a very significant change in terms of what the business is like. It's going to become better businesses as we move forward in time, both for shareholders but also for our customers as well. Thank you for joining us. Look forward to continuing the conversation.
Thank you very much. Have a great day.