Okay, welcome all. It's my pleasure to welcome you to this report of Artificial Solutions Q3 2023. Like I said before, the presentation can also be found on our website, under investors.artificialsolutions.com. It's been another great quarter of growth. We also met one of the significant milestone that we have on our trajectory, and we're gonna talk about both of those. We also had some experiences where Fredrik and I have learned a bit more about metrics and reporting. We'll go in a bit more into that as well.
But to start off, I'm coming to you from San Francisco, as we speak, and, I've been here now for, since this weekend and, met a lot of the people in the epicenter of what is basically version two of the big technology revolution. So we had the PC and the Internet, and now generative AI has dragged along with it all the different AI applications that we've been working on for years, that now have become mainstream. It is incredibly, incredibly hyped in, San Francisco right now. The out of the 50 new startups in the U.S. in artificial intelligence, 40 of them are in San Francisco, so 80% of them are in the city of San Francisco. So it's really totally changed the landscape here since just even a year back.
So something that to most people was only known in November of last year, so not even a year ago, has really changed this city dramatically. It's, it's moving in full cylinders. And spent a lot of time with Microsoft, also with McKinsey, and both of those companies are just incredibly straight on into this new AI revolution, and it's very, very interesting to see. And the bit of that discrepancy between the hype and the sort of joy and the forward-looking that we see here in San Francisco versus the valuation of our company in Stockholm is probably also why one of the things that you saw this morning is that the board has initiated a strategic review.
So we'll go in a bit more into all of that, and I'll try to weave some of what we've been seeing here into the discussion during this presentation as well. You probably know us by now. I see most of the people on the call are known to us. For those of you who have not met us, there's a few. So I'm Per Ottosson, I'm the CEO. I've spent a lot of time in the software industry and also recently here for the last... not, that would be like 13, 14 years by this point, been working within this particular space of AI using natural language, and based out of Stockholm. And Fredrik is also based out of Stockholm.
He's our CFO and has worked in all sorts of finance roles, both PE-backed, but also in SaaS companies. We've been doing this now for about three years, going on three years towards the end of this year. We're on a journey, which, when I've been over here in San Francisco, I've been trying to explain where we sort of stand in the journey, and I've been using a slide format that is slightly different than what we've used before. I want to do that here as well, to see if maybe this can convey a bit of also the feeling of why we feel we have a very strong position at this point. So obviously, this is chronological, starting in 2021, going to 2024. It starts with the new management team.
So great technology, it's been around for 20 years, patented, and the new team that came in together with me has changed the technology into being containerized SaaS platform, so software as a service. We've rebuilt that in the beginning of 2021. We also then took some of the really large customers that we needed as references. Also there, we had, we got Microsoft as a partner. Now, at the time, we were hoping that Microsoft would be strong in the contact center, which is where our technology finds most of its use. We saw that with our initial customers as well. However, that's not really what Microsoft did. However, they definitely contribute something else today, and I'll go into a bit more into that as well.
So the first customers came live. I would say we probably had the platform ready realistically in June of 2021. Project started a bit earlier, but the first go lives were in August, September of 2021. We've then been maintaining and nourishing these customers and making sure that they're growing, and we also have now a class-leading Net Retention Rate. Our customers are really growing very, very well. We started to see what these customers were using the product for, and the use case that we prioritized to build out is the same that the customers were most happy about, and that is picking up the phone and asking why you are calling. So it's what we call OpenQuestion. We made all these customers into official references, got those published. That's tail end of 2022, beginning of 2023.
We've been focused on growth in these existing customers during the first parts of 2023. One more quite key thing, which we achieved during this quarter, which is the tech partnering with the two winners in the contact center space. So Microsoft, after being here now in San Francisco the last few days, I can definitely say that my opinion is Microsoft is totally winning the AI race. There's no debate. They're way ahead. What they're gonna be presenting in the next, let's say, up until March next year, is just absolutely tremendous. What they're presenting in three weeks time at Ignite, even that is absolutely tremendous. So they, they are really, really way ahead on the AI side, but not on the contact center side. The two winners on the contact center side, especially in the U.S., are Google and AWS.
So we signed up partnerships with them, meaning that they've now approved our products to be integrated into their solutions. We've also initiated some metric collecting to make sure that we can explain to them why it's better for them, driving more consumption into their clouds, if they use our product rather than building this OpenQuestion themselves in their tech stacks. We did the first deal together with Google during this quarter. That's with HelloFresh. Very, very key for us. So going along with that, we're partnering primarily with Tech Mahindra, CGI, and Valcon.
So out of our partner stack, those are the three that we're working the most with towards this go-to-market motion with the large hyperscalers, and especially Tech Mahindra has a very strong presence here in the call center space, both in Europe and the U.S., and particularly the English-speaking parts of the world. So now we're at the stage of new logo hunting, and we are anticipating to close one or two new logos now during Q4. And that, of course, is part of the plan for us to reach our financial goals, one of them being cash flow positive during 2024. This is the company. It is a very busy slide, but let's just start at the top right. So I said the customers are quite important to us.
We have very, very good customers, and they speak quite a lot about us. There's a webinar next week, for example, with Telefónica, talking about their implementation together with us. We've done several with Swisscom. HRSA just did one in the U.S., where they spoke about their LLM implementation, so how they have Teneo integrated to LLM, to GPT. That was one of the first live ones, of course, also in the public sector in the U.S. And also we've seen that from Telefónica, also talking about cases like that. So very, very strong customer base. We're very happy with them. And then we have these partnerships with the three hyperscalers, which basically gives us a hunting license within their ecosystems. But then also a host of partners that we can work with as we go after these customers.
We are very targeted on 286 customers across the U.S. and Europe, and those 286 were targeted with a very, very targeted market motion. Which is, first we have, we're using something called The Point, which helps us to book up and set up meetings with the right people, and then we have a process that we take these customers through. So building pipeline together with these partners is what we're doing right now. Bottom left, you see the development of our recurring revenues. The revenues, of course, the SEK 8.6, SEK 7.6, is the SaaS revenue and the legacy revenue. But then you saw the...
You also see the ARR development, up there, and there it says quarterly average, and we'll go in a bit more into that, and why we're working with that. 70 live projects, with our technology today. We have ten API calls per second in the platform right now, which is very, very strong and way beyond number two in this market. So we are the de facto choice if you have a large call center and you wanna pick up the phone and say, "Hi, welcome to Microsoft. How can I help you today?" Rather than saying, "Please press one, press two, press three." The market we're in is growing. It's growing quite rapidly, 48%.
We're growing a bit more than the market, of course, but this is a market that's gonna shift now more and more into software and less and less services. So it's gonna go much more into SaaS and less into the integration services. Many headwinds in the market, sorry, tailwinds in the market, one being the OpenAI, GPT, of course, another one being this legislation coming out of Spain, also quite interesting. And also the fact that we have the really large implementations in this market makes it a very interesting market for us to go after. Here's the product we're taking to market. We have this Teneo platform. Teneo.ai, as you see behind me, is what we take to market. On top of that, we built a solution. That solution addresses this problem. Customers are just tired of keypad navigation.
They're trying to figure out what you said, which word matches with what they wanna do. You have your internal, you as a vendor, you have your internal way of expressing, "Press one for support, two for invoice, or three for..." That's not how people think when they call, so they're very tired of this. What happens is that 15%-20% of all calls are misrouted. This is statistics that comes from a company called ContactBabel, that does surveys of the 1,200 largest contact centers in the English-speaking world. They found that 15%-20% of calls are misrouted because people come in, they hear the keypad, they push the wrong key, or they just push keys to get through out of the keypad.
They end up with somebody, and then they have to be routed to somebody else. So it's the smashing of keys, as they call it. And also, the IVR trees grow quite big. So once you've chosen, let's say, I wanna report a damage on my car. Once you've chosen that, underneath that, there's several branches of trees before you end up with a person. So the 15%-20% of misrouted calls means that these calls cost at least two times more than they would have had they been routed to the right person from the beginning. And you now have a person have to understand why you're calling and then sending this off to somebody else.
So this increases the agent load, and just by taking this out, you could reduce the cost in the call center, the call minutes in the call center, which also then translates to the amount of people in the call center by 8%. So the customer experience is not great, but there's also big cost savings to be done in this solution. The change in this is that we can now move to-- Oops, it's actually not updating. Now it's updating. So, you can move to a very open solution, where instead of getting this keypad, you say, "Hi, thank you for calling Telefónica today. How can I help you?" And you state in natural language why you are calling. And here's also where we're now seeing LLM integrations.
Number one, to be able to vary the response that you have, the welcome response, but also to understand better what the person is saying, and then also to provide an answer immediately to that person who's calling. And initially, that gets checked by humans, so it's still, we're still using reinforcement learning, as it's called. But this is where LLM comes into play as well in this initial, trying to automate what that customer is calling about. You see here some of the customers and what they're saying about this. Lots of videos on YouTube now and lots of videos on our website, lots of case studies. I encourage you as shareholders to go in and have a look. These are some very impressive results and, and great customers that we have, on this, platform, Teneo, and using it for OpenQuestion.
So the existing customers are our growth engine. It's been our focus. We initially were hoping to sell this in a slightly different way than we ended up doing. We were, of course, building a pipeline together with Microsoft, but that ended up not going so well in the contact center. So we then focused on taking some large customers and really building them and helping them with supporting them. We're not building for them, we're supporting them in their building with our customer success team, and they're doing a great job. So we have a net retention rate of 156%. That's the first time we'll report this metric. It's something I heard a lot about here in San Francisco. It seems to be the big thing now that we have interest rates again.
People are very interested in this because, of course, existing customers have a lower customer acquisition cost or a lower cost of revenue. So they're quite interested in this. What I also cut in here is just a range. I got this when I was over at McKinsey. You could essentially say that a really good retention rate would be in the 125% range. What you see here is that for our size of company, in the SEK 3.8 million ARR range there, 109% would in 2023. This is from August 2023. 109% would be 108%. No, 109% would be a really good NRR number.
So Stellar NRR, I would say that this is the best in the best in class in software as a service companies in, in the Swedish Stock Exchange, but also looking out internationally, Snowflake, which is best, best in, in class, is, just slightly above this number at 172%. So very, very happy with the performance here, happy with the performance of our sales team, account management, but also our customer success team, of course, and then not to say the least, our, our product teams. So, one last, slide that I want to talk a bit about before I also let, Fredrik talk a bit more about the numbers. Here are some of the operational highlights from Q3. So first of all, we launched what's called a conversation IVR plugin for Amazon Connect.
We got that approved, and it's now on the Amazon Marketplace. This is very important. Amazon Connect is by far the leader in this space, and if you look at the pipeline that we're building together with Tech Mahindra, that is within the Amazon Connect space. So it's customers that are replacing their old legacy technology. That could be, for example, Genesys, and they're replacing that with Amazon Connect, which is a cloud-based integrated solution for customer care. That's where we are addressing with OpenQuestion. We're also launched our connector to GCP into Google Dialogflow. Google was in the beginning of this year, they were working with a company called UJET which they were also looking to acquire, which was their contact center platform.
Now they come out with their own contact center platform built on Google Dialogflow, and we now are connected to that, and we also did our first deal together, which is quite key for that part of the ecosystem as well. We also saw Telefónica launch an integrated Teneo-powered LLM pilot, and this is something that, there's been several presentations about. We have one on our website and on my LinkedIn. You can also find an invite to a webinar about this. It's quite interesting because also here in San Francisco and Microsoft, people raise their eyebrows when they hear that this is what Telefónica has done. So, quite cutting edge from them, and of course, helped by the fact that they can do it from Teneo directly and use our development environment to build a solution.
We also, and this is a bit more of a difficult one to understand, and also to see the value of, but confidential computing is something that's not very prevalent. Not a lot of people use this today, but we have launched and implemented this now for two customers, where essentially the data is encrypted throughout the whole process, also as is processed by the model. And that means that you never leak any private identifiable data or anything like that, as you work with the Teneo solution. So it has overhead, it's slightly more costly for us and for the customer, but for some highly regulated industries, this is key. Haven't heard that anybody else has implemented anything like this in the contact center space.
I haven't even heard of anybody implementing a real confidential compute implementation in the enterprise at all. So very, very much the first implementations of this type of technology in the world. So this is something that initially was invented by Google, but it's now supported by Google and Microsoft, and which will also be supported on the Amazon cloud. We also released our second version of OpenQuestion, again leaning in on the customers that are using this, putting some of those features in, and also cutting the implementation time down to 60 days. So within 60 days, you can have this system routing your phone calls and getting rid of your keypad navigation, which is brilliant for large customers. We signed an agreement with isolutions, which is a Microsoft one-stop shop in Switzerland. Switzerland remains an interesting area for us.
We've had our attempts, together with Swisscom, to do something in this market. That hasn't quite worked out. I still believe in the market, so we are now gonna work with isolutions, which is a big Microsoft partner, and of course, also has a lot of AI projects as they are the biggest Microsoft partner locally there. So, having some hope for that, too. Switzerland is not a big market, but because of the great reference we have and the fact that we support things like Swiss German, which is quite uncommon for solutions to do, we believe that we still have a bit of potential there. And then the board of directors initiated this strategic review, which, in my opinion, again, this is from the board of directors, and they did publish a press release this morning.
From my opinion, this has to do with the fact that being here in San Francisco, we're a bit of kings on the street here. People really like what we're doing, they really understand how we are driving the adoption of these AI solutions, but we're not really seeing that we get traction for this in our stock price, and the board is probably looking at how they can somehow get some of that value unlocked. But again, that is something for the board and not for Fredrik and I really to comment. Recurring revenues up 50% year-over-year. We'll dive in a bit more into this, and the SaaS ARR also growing a lot, so we're very, very happy with our growth.
The EBITDA is down to SEK -13, so we're still losing money, but as you can see, a lot less, and we're definitely tracking on our plans here for our financial goals. With that, I'm gonna give it over to Fredrik and rest my voice a bit.
Thank you, Per. Thank you. Good. So let's dive into the financial highlights in the Q3. This is a very busy slide, so I will a bit kinda jump around in the slide. But as Per mentioned, I mean, I think a key metric here that we introduce now is Net Retention Rate. Obviously, I think this is an important metric in the macro environment we are in presently. Also looking around, also as Per mentioned, relating to how are other companies doing on you know, growing upsell on existing customers, and we are doing quite well, as Per also mentioned, in that regard.
We are growing recurring revenues, so we continue to show record high quarterly numbers, SEK 16.2 million in the quarter, so up 50%. Very pleased with that. Also, SaaS ARR continued to grow, so +32%, and total ARR, +39% year-over-year. And the same goes obviously with net sales, more or less in line with the above metrics as well. Then on gross margin, we experienced 63% gross margin in the quarter, so quite much up from last year in Q3. But this quarter, we also experienced quite heavy commission, which is part of our, you know, compensation plan to actually then drive SaaS ARR growth to our sales team.
And that is also what we experienced, where we had very strong July and August and a bit softer September. But overall, that turned into quite a lot of commission, and we will come back a bit more on this topic later on. But that also means that, I mean, in order for sales to get commission, there has to be an average of 10% growth every quarter. And that is also then looked up on a monthly basis. So in order to then get commission in Q4, they will also need to recover what potentially has been, you know, not recorded in September. So that also has some implications on...
Could potentially have some implications on the October commissions as well, which will be in the company favor and perhaps not in the sales reps' favor. As Per also mentioned, EBITDA is continuing to improve. And also we are growing top line, and also then, as you can also see, our OpEx was SEK 9.1 million in the quarter. We will also come back a bit on run rate numbers and how we look at it going forward as well. Cash position, SEK 37 million, so very much according to plan, I would say. And then also the metric we have on the right-hand side is, I mean, 53% SaaS revenues.
And then also, which is incredibly important in the macro environment we are in at the moment, is that we have 99% recurring revenues. And coming back again on net retention, revenue retention rate, it's good to have loyal and growing existing customers, basically. Next slide, Per, please. This is a familiar slide for most of you that have been on these calls in the past. A key metric for us is SaaS API call volumes. And API call volumes on SaaS, I mean, that is a key indicator of how our business grows. And this KPI is an indicator on how our customer application and the usage of them are growing.
And the more application, solutions, covered regions, or how we call it, is more use cases, perhaps, that we are internally using that word, the higher API call volumes. So, that's a bit what we are driving our business towards. But also, in our operation, we also tend to see variations month-over-month. Just to look back into Q2 this year, we experienced a bit the same that we now experienced in September, in Q2 as well. So we had a very strong March month, where we also had perhaps January and February, which were not as strong as March. And then we had a bit of decline in April, and a bit soft May. And then we had a recovery in June again.
And then we reported, of course, the June number, but still, in order to a bit represent a bit on how our business actually is doing and also to avoid this fluctuation, we have decided then to instead of using the last in the quarter month as the basis for API call volumes and also for ARR, the ARR metric, we will instead, going forward, use a quarterly average, which is more kind of representing our operations. And maybe, Per, you can elaborate a bit more on the drop we experienced in September.
Yeah. So we had a quite significant drop between August and September, and we were a bit surprised by this, given that we have several customers that have taken new solutions into production. But the thing that happened is that we had robocallers from a certain country on the eastern side of the world that came through WhatsApp. And they couldn't quite. The customers couldn't quite figure out a way to get around this, and this was eating resources, but they were also trying to figure out things which were behind sort of the IVR, so automated things that you could do, that they were trying to do through these WhatsApp robocalls.
So what ended up happening is that our U.S. customers shut off the WhatsApp channel, meaning that a quite significant chunk then disappeared because people use WhatsApp to make phone calls as well on the American continent. And that we believe that's gonna come back next month, so that is in the November numbers. It's not gonna come back in October, but it's solutions that not we are working on necessarily. It is the operators who are working on that, because these customers buy the WhatsApp traffic through operators. So it's primarily AT&T and to a certain extent Sprint, but they are working on fixing that problem. So we had a significant drop. It has to do with switching off one channel entirely.
This channel is typically 14%-20% of the total traffic, so that had quite a lot of impact. So, September, we weren't expecting that, but we had a drop between August and September, and we believe that will be recovered by November numbers.
Yeah. I think we can move to the next slide, Per. Looking then on applying average quarterly API call volumes, we see a smoother and less volatile growth, obviously. And we think this better represent, you know, the way our operations work, because we, as... I mean, what we have experienced this quarter, but also previous quarters, is that there can be some fluctuations in between months. And we think this is a better way to measure the operations, basically. And all in all, I think we are experienced quite a lot of growth. So year-over-year growth of 72% in our SaaS API call volumes. And as Per said, I mean, we expect to return to same levels that we have experienced here earlier in the quarter as well.
So, that is our understanding going forward. So very positive outlook, even though, of course, we would like September to have been stronger, obviously. Next slide, Per. And what we showcase here is basically ARR growth, applying average numbers in the quarters. And as for API code volumes, in this way of measuring ARR, we smoothen things and avoid monthly fluctuations. And the SaaS ARR grew with 53% year-over-year, Q3 versus Q3, 2022. So a very strong growth rate there. And also total ARR growing 50% year-over-year. So overall, a very strong growth, and also a big coming back to net revenue retention. It's showcasing that we have, you know, customers that continue to grow with us.
It's not only the SaaS customers that continue to grow with us, but also a very solid growth from our non-SaaS customers. And that is also something which is very pleasing, I would say. Next slide, Per. A lot of perhaps the same message, but I think we're very pleased with, you know, growing recurring revenues of 50% year-over-year in a kind of macro environment where we are in at the moment. And we are growing the recurring revenues in a solid way, both on SaaS and non-SaaS, and also continue to show growth on LTM numbers on recurring revenues, as well as ARR. So very pleased with how we are performing operational. Next slide, Per. Then coming back a bit kind of on the OpEx situation.
As we have guided for in previous reports, we are simplifying operations, and we are also trimming the organization. As was also evident in our report for the third quarter, we have also continued to reduce headcount, to some extent, also in Q3. We have decreased the OpEx run rate from SEK 116 million in Q1 - SEK 109 million in Q3. The monthly OpEx we have now in Q3 is approximately SEK 9 million. We are continuing to trim and, you know, work a lot on performance management measures in order to drive our operations in the most efficient way.
And we will probably see some reductions in some OpEx going forward, but there may also be some increases in other places. So I think the level at where we are at the moment is probably a fair assumption going forward in what we will see in Q4 and the coming months as well. So SEK 9 million run rate is a fair assessment going forward in the coming quarters as well. I leave over to you, Per.
Sorry, I was muted. I will summarize with this slide. So again, a very strong quarter, another strong performance by the team. We are growing more than the market. We are, at this point, a very well-known entity in this market with the strong references that we produced. So very happy with our position and our growth. We also haven't spoken about this in this report, but our gross margins on the fully onboarded customer, they approach 95%, which is, of course, what we want to get from a software as a service company. So we definitely have the right model. We just need to get customers to grow to a certain size to be able to realize those type of margins.
It's a technology that is built for large blue chip customers, and these are in industries where people right now do wanna save money, and it both saves money and provides better customer experience, so it fits in the macro environment as well. It is difficult to switch from us to something else, primarily because of our markup language, which is one of our patented technologies, the way you store the data, essentially, of the conversations that you've had. And it would require quite large projects, but there also isn't a solution today that can support the type of volumes we're doing with the right performance. And it's all about once you say, "Hey, welcome," and the customer starts speaking, it cannot take a few seconds before you speak back, because then the customer will hang up.
So that's the performance metric that is difficult for other solutions. So far, there aren't other solutions that can build that. Now, also, what we have is references, and we're part of all the three hyperscaler partnerships, and that's what we're gonna be using for sales expansion. And that's my focus now, during Q4 and Q1, to expand the sales, to get new logos in, that can then also start growing. So with that, we're gonna open up for Q&A, and what you need to do is raise your hand, and then I unmute you, and after I unmute it, you can unmute yourself. That's how the system works. I see a hand now. Let me just find Victor in the list.
There you are, and I have now allowed your microphone, so now you can unmute.
Great. Can you hear me?
Yes.
Yeah. Super. So, my first question is regarding the WhatsApp disruption here. So, I mean, what kind of implementations can you do to mitigate such effects in the future hereafter?
That's a good question. I don't really see that we can do anything in, in our system. It's not really something that has to do with us. It has to do with somebody figuring out how to use WhatsApp to sort of fool the system. So I, I'm not really sure that we can do anything in that. It's the implementation of our software, and that's something, of course, that, in, in this case, it's primarily one of our partners that's working with this, and what they're working on as a workaround is a simple identification script, essentially, before, before you hit the actual IVR to see if it is one of these. They come from one and the same... The IP addresses come from one and the same country.
But of course, the numbers that they use as their WhatsApp identifiers are numbers that they bought in all sorts of countries. There is ways to find out who's behind the call. And we will find a solution, but I'm not really sure if we can sort of will we find the future such problems and then fix them preemptively? I don't think so. I think that this is an ongoing struggle in the world to find these. The hackers keep moving the ball around. If that makes sense.
Yeah, I see. And then I have a question regarding the ARR. You mentioned that you expect the last error to be back in November, but should we interpret that the November level be back at August numbers that you press released back in, yeah, for a month ago, or how should we view that?
Yes. So that's, that's what we're expecting. As soon as we can turn on that WhatsApp channel again, we also have expanded volume, so definitely, yes.
Okay, great. The last one for me, you still have around 11 customers on your SaaS platform. I mean, what does the pipeline look like of new customers, and what would you say is the biggest challenge to convert leads into new sales?
The biggest challenge right now is attention. It has to do with the fact that a lot of CXOs are pushing the organizations to do LLM-type things, which takes a lot of resource and attention in the organizations. Now that we have two cases that are documented and they're also gonna be talking to our customers, I think it's the week after next that our next webinar is. Now that we've done that, that means the contact center guy can now say to the CXO layer, "Hey, I have this great GPT implementation, which, by the way, also saves us money." So, the pipeline has suffered a bit from attention deficit during Q2 and Q3, primarily.
As far as what we have in the pipeline, we anticipate at least one customer now in Q4, probably two. So we have four that are in the end stages of negotiation.
Great. Thank you. That was all for me.
Thank you, Victor. If there's any other questions, just raise your hand, and I'll unmute you, and then you get to unmute yourself. Well, there seems to be no further questions right now, so maybe I'll just summarize again. Yet another strong quarter of growth, and also, AI is definitely, it's definitely gonna be the second revolution. The first one was PC Internet. Now it's AI, and we're very happy to be in the middle of that tornado that's taking over the world. So being in the U.S. and seeing what we're doing and where our position is in this, it's incredibly fun. So I'm really looking forward to the next few months here. It's gonna be good, good fun for us. So I'm gonna go ahead and close the call, and thank you all for listening in.