Teneo AI AB (publ)

Good morning. It's top of the hour. It's nine o'clock. So I suggest we get started. Just as a reminder, this presentation is being recorded and it will be on our website later today. And you can also find the presentation on the investors site on tineo.ai. So with that said, we'll go ahead and start. Today presenting is Per and Fredrik. It is of course the Q3 report. We have been around now for quite some time, Per and Fredrik, although last quarter was somewhat different than maybe we had hoped for when we started the year, with the whole sense Q1, Q2. But we do have a background from this area, both of us, so quite experienced in the business of software. So I'm going to start off with an update from me, and then Fredrik will go into the numbers. It's not too much to talk about in the numbers. It's pretty much the same as we did forecast in our Q2 report, which of course was in connection also to the rights issue. So for those of you who were there, it's not that much news, but I'll thank you all for joining again. We have three phrases sort of to describe where we stand today. So it's debt-free, recapitalized and rebuilt to act. So it was a very demanding year and quarter. It all, of course, has to do with our former US partner that broke our contract. And there's just no way to sort of sugarcoat this. It is a big drop in revenue. and that of course resulted in a severe cash squeeze but as they say you never waste a good crisis so in the third quarter we worked hard to make sure that we could emerge stronger faster more agile than before so first of all debt-free uh our balance sheet has been a major issue for shareholders, for M&A, and also for our large customers, and it's been so since Fredrik and I joined. What happened in the quarter is we completed a full debt-to-equity refinancing with Capital Four. We no longer have a loan, meaning that Teneo is now debt-free, and Capital Four is likely to be our largest shareholder with 29.9% of the shares. We're also recapitalized, so thank you shareholders for supporting us in this tough period. You contributed through a rights issue of 74 million. We ended the quarter with 25.8 million. That has a lot to do with the fact that we, of course, had during Q2 several months where we couldn't pay the bills at all. So there was also quite a lot of catch up that we had to do, apart from also the transaction and some repayment of loans as a part of the transaction here, the rights issue. So from October, we run the company at about five million kronor a month, which is close to half of where we were a year ago. So the cost is about half of what we had a year ago. So we also say we're rebuilt to act. What does that mean? Well, the new organization is slim, it's small, it's agile. And the first proof of this is the delivery of Tenio 10 and its reception that we've seen in the market. So we've rebuilt Tenio. But instead of taking sort of the old code and starting to recoding it, we still took 20 years of code and design principles, the concepts that we've been building on, the concepts that customers love, and we've redesigned the platform from the ground up. Why did we do this? So there's two reasons. Number one, we want to enable an AI driven software development lifecycle. What does that mean? Instead of developing where all code is written by humans, both the code and the coordination is done by the AI. meaning that we have a model in the middle and then our developers are working towards that model. This enables a far faster development at a much lower cost than before. Also, since Tenio 10 builds on everything the customers required from a contact center solution, it's the stability, it's all the confidential compute, it's the auditability, it's security and all this in one and the same platform. But now you don't really need to log on to Tenio 10 to change a flow or to build a solution into Neo, you can simply prompt Claude or ChatGPT. This is what's called headless. It is, of course, a big craze in the SaaS industry because if you charge for the people that log into your platform, charge for seats or users, this is an issue. For us, it's not because, of course, what we charge for is the API calls. Now we transform that into charging for minutes, but it's the same thing. We charge for volume. So that's a sort of very quick update of where we stand. And give you a summary here on what I've been talking about. There's one thing I didn't mention here, but the Spain exit is completed as of this Monday. And we are holding our gross margin. We are having a much lower delivery cost with 10.010 as well. So going forward, we will definitely be able to climb up again on the gross margin. And we also stabilized the ARR, of course, in this quarter. A lot of senior debt has been eliminated. And of course, that's transformed into a 29.9% stake, implicitly actually valuing the company a lot higher than what it's valued at on the stock exchange today. But that is, of course, a different matter. So let me talk about Tenio 10. I think that's the most important thing. So this is what really is gonna take the market now. and what we're focusing on. We're focusing first on a few large engagements we're engaged in to roll this out, but we're also at the same time presenting this to new customers. Was in Turkey last week, presented this. I was in London the week before, we presented this to prospective customers and they're loving it. It is a very, very big upgrade to the usability of the platform, not to the stability, not to the confidentiality, not to the auditability, but to the usability. So make it easier and faster, even easier and faster to build. So the typical voice AI today, which we've seen hundreds of companies pop up to do, is that they have the language model to both generate facts and actions. So they use open AI, they use cloud or some other language model, and the language model goes into your backend systems looks a bit, it has read-only access because nobody allows the LLMs to write into a backend system. So into SAP, into Dynamics, into Salesforce, into ServiceNow, because if you would, you'd probably get dumb data into your system of record, which you don't want. So the language model in a typical implementation, Sierra, Perloa and the others, it generates the facts and the actions. This, This means that the owners of the systems, so say ServiceNow, SAP or some other backend system, the security people and the compliance people cannot prove what the LLM agent will do. Therefore, they're saying, no, stop, we're not going to allow you to interact with our backend systems. This is very much what we saw at Neutronic, where other LLM systems could not interact back to ServiceNow and then also to the patients or the users of their devices. So what happens instead is in a traditional voice AI implementation, it answers the question, but then a person still has to work. So you still have a person to sort of check and make sure, or go into this backend system and actually enter what the LLM has sort of decided to do. So instead, what you see on the right is the Tenio approach here. We have all facts and actions running through a controlled, tested, auditable layer. So what does that mean? Well, first of all, it's not going to put out anything that is not correct. It's not going to put out anything that you can't afterwards see why it put out that answer. Where did that come from and who did it give the answer to? It's very, very important for enterprises, even more important in the EU today and even more important in regulated industry. The language model, which we still use, we still use LLMs, but it will only take the words and answers that were already verified. So if there's a placeholder, which is a fact, is never going to change that. It's just going to be able to change the sort of dialogue around this to making it a more natural sounding dialogue. So all the three roles, the security person, the owner of the system, and the compliance people that typically block these projects can say yes. And that's of course why we have read write access into customers' backend systems, because we can prove that Teneo is not going to change things which are incorrect. What we also have now is a replacement model. During Q1, Q2, we primarily used the American LLMs. What we found is that customers are no longer, and that is both in the US and Europe, customers no longer trust these models. And of course, a lot of that has to do with what's happened in Anthropic. When Anthropic says they're retaining your information, they retain the right to also train on it, meaning that they can take your processes, your way of working, and train on that. But also in Europe, a lot of customers are asking for either a European, like in Mistral, or a local LLM. So having no U.S. ownership. And most of our competitors today have U.S. ownership. So all the big ones that you might hear about that have lots of money, they're all U.S. owned. So this is quite a unique selling point for us that we're not U.S. owned. We're obviously owned by you guys on this call. We can also deploy anywhere. This is quite important in, for example, Turkey. It's also becoming more and more important in Europe and Germany next week, where this is becoming quite important to its security concerns and its concerns about the US in general. So we can deploy now to Neo 10 in the cloud as usual. So we sell the software as a service. We can provide a private cloud solution. So if you have Azure private cloud, Microsoft private cloud, we could provide there. which is in-country in your data center. So that gives customers a lot of flexibility to meet the demands that are coming, the EU AI Act, it's the GDPR, it's KVKK, it's new ISO standards. And one of the most important factors is that we have a fixed price per minute. This is the API call translated into minutes, but it's a fixed price. So you know the cost of your solution before you even start writing a line of code. That is not the case for any other voice AI solution based on LLMs because the LLMs will always charge you per token. So you don't know upfront how much cost your solution is gonna cost, which also means you don't know the ROI of the solution upfront. Whereas with us, you get a fixed price because we don't need to use that LLM in the actual production, meaning that we can price according to we know our own compute instead of somebody else token cost. And you probably saw the numbers this morning in the Wall Street Journal, that Anthropic lost 74.6 billion dollars in the first nine months of this year, which of course means that they're going to need to raise the prices at some point. So this token costing is a big issue. We've seen customers paying one and a half dollar for simple FAQ calls, so simple answer question calls about facts on your website, essentially. to these voice AI providers. So Tenio 10 has a hybrid AI margin lever that's for the customer and for us, meaning we provide a fixed cost, but we also know our cost in the background, so we don't have to use the LLM. The hybrid AI is good for that, but also, of course, for making sure that you provide the right answer. We have the observability, which is something that, for example, Medtronic could never be without. There's no way to provide information about a pacemaker without afterwards being able to say what information you provided if something goes wrong. We have a voice gateway today that we built in, meaning that we have both voices and STT engines that we price in our minute price, which means that there's less third party spend for somebody else, but also we can retain larger portion of the margin. So we can provide a larger portion of the total value to the customer. The authoring UI, our own UI, is very fast, very easy to connect MCP connections into. It also has, of course, a more, like customers now are expecting, like a chat, like an open AI chat, JPT-type interface. But it also has the graphic interface that a lot of the more trained developers are going to need. We have a testing suite built in that's quite important because it's going to be testing the solution and providing recommendations as you go, meaning you can improve the solution without really taking human talent to use that either. And it's still the endpoint that the engine is still the endpoint and engine that had 120 million interactions last year at Microsoft with an accuracy and scale that really hasn't been proven anywhere else. So you get all the good stuff of the new way of developing all the good stuff of security and compliance, and then all the good stuff of a proven model. I want to talk briefly also about the team. So it's a lean leadership team today built to first take the company to cashflow positive and then grow from there. So we slimmed down the organization enormously and we're now 22, 23 people, but retaining that 5 million cost baseline, which is, which is the most important thing for us, including compute, of course, that we're providing to customers. So we have Johan, our chairman. Per is, of course, me. Marie now runs, she's a chief revenue and marketing officer. Marie has a long background in both sales and marketing from Sweden, also from Qlik in the US during their stock introduction, and also in our space in the US in a private company as well. Patrick has worked both in sales, pre-sales, and chief operations officer roles before. Also manages our existing customer. Fredrik, of course, on the call, our chief financial officer. And then Andreas, and this is where our biggest organization sits under Andreas. He's our CTO. We have a head of cloud and ops there. And then we have product marketing and head of product sitting under him. And head of product, of course, there's AI and there's people underneath there. and, of course, also in cloud and operations. We are operating a 24-7 operation for our customers, of course, in the SAS product. So also important just to mention, we're keeping M&A and the legal position with the board, with some external parties working on that. So the M&A project is ongoing. It's managed by the board. it's not something that we as management team spend time on except when we have to present something but it's not it's not our focus same thing for the legal case in the US we of course do not have the money to finance that legal case so we're working still to find a financing partner for the case and I present in those but I'm not involved in in the whole process at all from that perspective so it is really very much a focus on sales so we focus on growing the revenue and reaching that positive cash flow. And the way we do that is primarily to focus on our existing engagement. So it is Spectronic. It is FGS in Turkey. FGS is our partner in Turkey, which has several customers, one of them being the Telco, which is our initial implementation here. We had to, because of the data center implementation, it's not in our data center, this is on-prem, so this is in a Turkish data center owned by this partner. We have to compress the onboarding now because we've been a bit delayed in the local data center. And that has to do with the fact that the data center is moving technologies, going from VMware to OpenShift, which is something a lot of data centers do today. But in the middle of that, we can't really deploy. So instead of going live in September, as we were projecting, we're going live in November. But what we're doing is we're training, for example, the B2B sellers now. We're also going to be selling the solution to the telcos customers. So right today, actually tomorrow, we're training the business to business sellers at the telco. And we're also working on starting to build the first flows in workshops, although we don't have the live data center yet, but that's to speed things up and to work during while we wait for the deployment. We also have a Nordic Telco that is projecting to go live in December. These go lives are not fully controlled by us, but we're very focused on them because of this is what's going to take us to cash positive. So AI Cold Guardian going into production at Nordic Telco. So what is AI Cold Guardian? I didn't talk a lot about that before here today. AI Cold Guardian is essentially a receptionist in the network. So the network, the Telco, in the network has a receptionist, which picks up a phone call that goes to you, Axel. Axel picks up the phone for you, says, why are you calling me today? Or whatever greeting you would like to have. You talk to it in some language that you choose. And then the person calling, the person's phone gets an SMS and says, would you like to let this person through? And you can also, of course, filter and say, do not let these types of calls through, et cetera. so you don't even have to get the SMS. This is a very powerful solution today, but it's even more powerful when you consider the fact that, for example, I use Grogbot today when I need to book a hotel or something, and I have Grogbot call hotels, which means that now people are getting calls from bots. So the amount of scam calls or sales calls or calls you got before are just going to multiply as soon as somebody has your number in some database. So this is, I think, a quite important development AI call garden as well. And then we're projecting that with these, we're gonna be in Q1, around Q1, we're gonna be able to have that cashflow revenue flow, which is for a breakeven target. So very much a focus on existing. And then of course, a few of us are also spending time building up a pipeline on the side for the going forward. And that has a lot to do with with enabling partners. And it's a different set of partners than we had before. We learned quite a lot about how to work with partners from this big debacle in the US. So let me start with the last point here. What we're looking at here is smaller mid-sized partners that sell to smaller mid-sized customers. So for example, contact center providers and not the large ones, the ones that are gonna have the money to be able to try to build something themselves on top of an LLM. That's the voice AI, the left side of what I was showing, which a lot of these large ones are building like Genesis, for example. Remember, we did have a great partnership with Genesis at one point until they thought that they would build their own voice AI on top of LLMs. So smaller players in the market that where we can be, where we know that they're not going to be able to develop this themselves. That's our focus right now. So bringing projects into revenue from the top here, AI CoGuardian, of course, very important, FTS important, Medtronic, very important. Medtronic is also important because it's on top of ServiceNow. We're going to go hard after the ServiceNow market once this is done, not by going necessarily through ServiceNow, but through all their smaller partners. Strict cost control, very much on Fredrik's plate to make sure that we maintain this rate. We can't have cost overruns that we need to find things to tune, which is much easier in the new setup than it was in our old setup, of course. And then the Tenio 10 that we've been around showing, of course, we want to convert that into signed contracts. What we see is that regulated European companies, very, very important to them, what we provide in Tenio 10, the fact that we can provide a purely European solution. So this is where we're aiming also as we speak. And at this point, that's the main focus. So with that, I'm going to hand it over to the financials. Again, it's not much more than what I said, but so we had the supporting slides, but I think Fredrik will go through this and give us a sort of top line view of it. Fredrik, the floor is yours.

Thank you, Per. Yeah. So in line with what you said, Per, I mean, What we see in terms of revenues and costs is essentially where we operate from as a new organization. And also already mentioned, I mean, the pure and sole focus for the management team now is very much to increase the revenue line. And the increased revenue is going to take us to cash flow positive and with the existing organization that we currently have. So, significantly smaller organization, lower costs, obviously. And key here now is to drive revenue expansion. That's what we are all working every day and every hour on reaching, essentially. And for that reason, I will not go into too much details on the financial slide, because I think it's rather straightforward. And as you all are aware of on this call, the main deviation that we see in the revenue line in sales and SaaS ARR in the third quarter is very much linked to the decline in sales from a former reselling partner, as we have mentioned in a number of calls and also in the reports in great detail. So I will not cover too much about that. As also Per presented, the company's pipeline of ongoing engagement is continuing to develop favorably during the third quarter and fourth quarter, and we expect that to also generate revenue growth going forward. In the quarter, net sales in constant currency amounted to 5.5 million, so a significant decline from the previous quarter last year of 20.4. SAS ARR in constant currency amounted to 12.6 million versus 66.3 million in the same quarter last year. Total ARR in constant currency amounted to 21.4 million versus 82.2 million in the same quarter last year. So a significant drop in top line growth in all those metrics. Gross margin wise, we deliver a 76% gross margin despite the significant volume decline. And as Per also mentioned, we expect to improve the gross margin going forward with higher volumes combined with also a significantly lower delivery cost. That is what we have worked a lot with Teneo 10. In the second and third quarter, we have reduced all costs possible in order to adjust the cost base to the lower sales volume. And that has also resulted in that we are reporting an adjusted EBITDA in Q3 2026 of 11.3 million SEK. And this number is obviously adjusted for one of cost and restructuring costs. At the end of the month, we also report a adjusted cash position of 27.7 million as of September 30th, 2026.

With that, I think we can go to the next slide, Per.