Clavister Holding AB (publ.) (CLAV) Earnings Call Transcript & Summary

Hi, everyone. Welcome to our Clavister webinar, giving a bit nuance to our latest news, the acquisition of our AI specialist, Omen Technologies. With me, I have our CEO, John Vestberg. Welcome, John.

Thank you, Jenny.

Yes. And the CEO and Co-Founder of Omen Technologies, Wissam Aoudi. Welcome, Wissam.

Thank you, Jenny.

On the agenda, we will start the presentation with a presentation. And in the end, give you, the audience, the possibility to ask you, Wissam; and, you, John, your own questions in a Q&A. [Operator Instructions] But now, John, should we get started?

Absolutely. Thank you, Jenny. So well, again, welcome, everyone. The intention of this session is to provide some more insights, some more details on the announcement we made yesterday, namely the acquisition of Omen Technologies, a very skilled AI specialist within the field of cybersecurity. So starting off with providing a little bit of an insight for those of you who might not be familiar with the artificial intelligence and machine learning market, specifically within cybersecurity. So basically, in a nutshell, what will artificial intelligence and machine learning add to the domain of cybersecurity? Well, obviously, the trends are very clear. The attacks and the threat actors become more and more sophisticated as time goes and machine learning, AI tools are being increasingly used in crafting new attacks and exploring new possibilities to attack systems to avoid already existing mitigation systems and mitigation technologies. So by using AI, by using machine learning, there is a good way of analyzing patterns, learning from patterns, detecting anomalies and basically being able to identify potential attacks or anomalies that could represent a threat and proactively respond to those. It is clear that AI has a very natural place in any cybersecurity solutions going forward. The actual market for AI in cybersecurity is very big. It's growing dramatically. There are some numbers that point to a CAGR of almost 24% up until 2027 with an anticipated market size of over USD 46 billion by 2027. So already on its own, it's a very interesting market. There are another aspect as well and regulations and new directives that really drive the demand on AI and machine learning. This is one example within the European Union where within the vehicle space, the automotive space, there is a requirement from the European Union that all vehicles produced from 2024 will have to have a certain level of cyberattack detection mechanism, cybersecurity mechanism. So this is obviously a very strong driver and it strongly ties into what Clavister is doing already within the field of CyberArmour where we address a certain type of vehicles. With that, I'd like to go through a few of the use cases where we will see the acquisition of Omen add concrete value to our solutions space. The first one is within our defense sector where we specifically have our CyberArmour solution. What we can do with machine learning and AI in our CyberArmour solution is to add what typically is known as zero-day threat protection or threat prevention. And for these type of systems, for defense systems, for armed vehicles or for any type of defense system, that most common or the most imminent threat, the most dangerous threat is really the attacks and the digitalized weapons that are tailored towards this specific industry. Keep in mind that there is, of course, thousands and thousands and millions of different malware threats out there. But most of them are common. They are known already. If you are an adversary and you would like to attack a nation's defense, you would obviously not use a common known malware because that would be just too easy to catch. So basically, here, we're looking at adversaries with quite powerful resources, be that states, be that terror organization or what have you, that could basically craft zero-day threats and inject them into typically the supply chain or any other means of communicating with the systems. Now in this specific case then, by adding machine learning and artificial intelligence algorithm to the solution, we can then basically advance our position in this space, we can have our solution be completely fresh, look at anomalies, look at traffic patterns. And by doing so, detect these very dangerous so-called zero-day attacks. The other cool thing here is that we can do this not only on Ethernet and IP, basically the classical Internet protocols, but also on a protocol called CAN Bus. And this is the de facto standard protocol that is being used in the vehicle industry, in the IoT industry, in the industrial sector to communicate between various type of industrial devices, be that an engine, brake systems, even weaponry can be controlled with CAN Bus. And this is obviously a very, very, very dangerous attack surface if that would be left exposed to zero-day threats. So with our solution, we can now detect threats not only on IP, but also on CAN Bus. So it makes it a very much more comprehensive solution. The other nice use case or part of this use case is actually a positive side effect. And by looking at anomalies and by looking at changes in traffic, we can also detect anomalies that are not the effect of an attack per se, but the consequence of piece of machinery, a piece of the equipment that basically is in need of maintenance because it starts behaving in a different way. It starts to be worn out. It starts to be in need of maintenance. In -- without this kind of solution, you would obviously have to rely on standard life cycle cycles -- or life cycles timing. But with this type of solution, we can basically have an early indication from the system itself to allow the service team, to allow the owner of the equipment to facilitate service in a location under good conditions where it's basically safe, not doing out -- doing it out in the field where you would put crews to unnecessary risk. This is a nice extra value add that comes from this solution. Also, if you take all of this together, looking at a fleet of vehicles, for instance, there is the opportunity, for instance, to gather, to collect operational data and look at this from a holistic perspective. Basically, if you have a fleet of hundreds or even thousands of vehicles, you might want to look at the data holistically, gather it centrally, you apply machine learning to this data. And by doing it, you can really detect anomaly -- different anomalies on individual vehicles forming part of a larger fleet. This can be important as well. In practical terms, we have our RSG-400 product, that's one of the key products within the CyberArmour solution. It has already a lot of strong features to be a viable product for the Defense segment. However, by adding them, artificial intelligence and machine learning capabilities, we add specific features, for instance, zero-day threat protection, proactive maintenance, centralized threat analytics and the multiprotocol support. So basically, we augment the product from already a very strong product to a very competitive and very strong product that can accommodate even more use cases and even more solutions. We have been running a project for a while together with Omen Technologies and together with BAE Systems. This is a project that is funded partly by Vinnova. We announced this a number of months back. The scope of this project was and still is to detect -- specifically, again, detect cyber threats that are arriving over standard Ethernet or IP, could be, for instance, an IP camera. Keep in mind that most of the modern vehicles have a lot of IoT sensors, IoT devices, IP cameras, they could be potentially subject for being replaced with a malware-infected firmware or being modified in a way that makes them the entry point into a system. And within this specific project, the scope is to detect that we can -- or to ensure that we can detect those type of vulnerabilities. The other part of the scope is to detect threats, again, coming over to CAN Bus. So it could be a directed attack, for instance, on the engine control or the weapon system. And again, this might be even more devastating if this happens in the field. Imagine a combat vehicle that suddenly stops in action because of an attack. So far, we're in the mid of the project. So far, it has been very successful. So we have already been able to prove that we can detect these type of attacks and we can do it with live data from a real combat vehicle environment. If we look a little bit more on the horizon, the CyberArmour solution is obviously the first solution where we will implement our new machine learning algorithm from Omen. But the ambition is that this type of technology will find its natural place in all the Clavister solutions. One of the solutions will obviously be our Secure Access Service Edge, which is our cloud-delivered security solution. And within the scope of SASE, what's the purpose of having AI and ML and SASE? Well, it is a perfect location to apply machine learning and AI. Reason being that we have access to vast amount of data that is being stored in the cloud. We can look at user behavior from -- if you have hundreds or even thousands of -- or 10,000s of employees, you start forming patterns and behaviors. And by doing so, looking at the differences, the anomalous change in behavior, this can be the perfect early warning indicator to understand if there is some or several in your staff that starts to behave in a different way and therefore, might be a warning for an attack. This is a bit more long-term project. This is not happening overnight, but it's a natural step on our SASE journey. Specifically, some people in the audience might have seen our SASE technical stack before. But specifically in our stack, the green box on the top really represents what we're talking about here, the anomaly and behavior detection engine that has a comprehensive look at all the data that is being gathered by our SASE solution. With that, I'd like to hand the word over to Wissam to provide a bit of a background and more insight and details.

Yes. Thank you, John. So hello, everyone. Omen Technologies was founded by 2 researchers from the Computer Science Department at Chalmers University, myself and Associate Professor, Magnus Almgren. And shortly afterwards, Christian Löfvendahl joined as late Co-Founder and Business Developer. The start-up is a spin-off from Chalmers University. And the focus of our research at Chalmers was on developing AI-driven cybersecurity monitoring solutions that can detect advanced cyber threats in real time and purely by analyzing data streams. Back in 2018, our research work got accepted at one of the world top venues on cybersecurity. But our innovative solution was not only recognized by the research community, but also by the industry as we came to realize from the research collaborations we had with major industrial actors in Sweden. And that is why and when we decided to start the company and commercialize our solution. So Omen was founded in 2019 in the city of Gothenburg. We are currently 3 full-time employees and we are backed by Chalmers Ventures. So nowadays, systems are becoming increasingly connected to open networks and many of these systems are mission-critical. This increasing connectivity is adding more layers of complexity and is rendering these systems ever more exposed to complex cyberattacks that can target safety-critical components. And to cope with this rapidly changing cyber threat landscape, it has become clear now that classical cybersecurity techniques need to be augmented with AI and machine learning capabilities. And this is precisely what we do at Omen. So although our AI solution is a stand-alone technology, it is meant to be complementary to classical existing security mechanisms in the sense of adding an extra layer of sophistication and intelligence, which is necessary these days to stay ahead of the game. For instance, as John mentioned earlier, a targeted attack is unlikely to be detected by a classical antimalware solution that lacks an AI capability. And in the last couple of years, Omen has attracted considerable attention from the Swedish industry for having an interesting technology that is capable of solving difficult emerging challenges and for having the domain expertise, both technical and commercial, to bring the technology to its full potential. We have been working with Clavister and BAE Systems within the Vinnova-funded R&D project that John talked about, which is going well. And the main reason why we decided to join forces with Clavister is the realization of Omen shareholders that our goals can be achieved faster and at a much larger scale. So now I will give a brief overview of our technology and what we bring to the table. So in a nutshell, Omen's technology is an AI -- an innovative AI-driven methodology that is basically, specifically designed for real-time anomaly detection in a security context. The methodology is rooted in state-of-the-art research on mathematical analysis of temporal signals and data streams. And it possesses key features that make it practical and suitable for modern connected systems, including the following: it is extremely lightweight and resource-efficient. And this means that it can be deployed on edge devices, if needed, so closer to the data source, which is obviously favorable from a security perspective. It is specification-agnostic, which means that it requires very little knowledge about the system or the process generating the data. And this makes it applicable to a wide array of systems. It is purely data-driven, which means that it can adapt to the changing behavior of the underlying system over its life cycle. And finally, it is based on the unsupervised machine learning approach, meaning that it doesn't require labeled data sets. This is particularly important because it enables the detection of zero-day attacks that are otherwise undetectable. So the following video presents a visualization of one of the monitoring and detection use cases within the Vinnova project with Clavister and BAE Systems. So the plot on the left displays real-time output of our AI algorithm while monitoring CAN traffic, that's contain-and-attack scenario somewhere in the middle, and the plot on the right showcases the internal working of the algorithm. Our technique works by first learning the normal behavior on -- normal traffic behavior on the monitor CAN Bus from CAN message payloads. And this is the training phase of the algorithm where a so-called baseline behavior is identified, which is represented by the blue cluster that you see on the right-hand side of the screen. The training phase is not shown in this demo. What is shown is the operational phase of the algorithm where it continuously analyzes the payloads of consecutive CAN messages transmitted on the Bus and detects any significant deviation or departure from the baseline. One of the key advantages of our unsupervised learning approach is that we don't need to know anything about the attack to be able to detect it. So it can be a completely new attack on CAN Bus technologies in this case, and the algorithm will still manage to discover it because it will induce a change in the overall traffic behavior. As mentioned previously, in this project, we are monitoring IP traffic as well, and the results are like the ones displayed here. And this ability to provide multiple monitoring protection is due to the specification, agnostic feature, that I pointed out earlier, which allows the technology to be applied across multiple data sources at the same time regardless of the nature of the processes generating the data, as long as a baseline behavior can be identified. Our technology is quite mature, and we have the skills and the know-how to evolve it according to the market needs. And now that we have access to Clavister's enormous resources, we are looking forward to enabling AI in most or all of Clavister's solutions. Thank you. Back to you, John.

Thank you very much, Wissam. So yes, to conclude, very impressive results that we have seen so far and the ways of working and the similar culture that we have between Clavister and Omen, together with obviously the very interesting market opportunities ahead were the compelling reasons why we, together with Wissam and his colleagues, decided to join forces. So with that, Jenny, I think we open up for questions, if any questions.

Yes, we will open up for questions. [Operator Instructions] Wissam, how do you feel about moving into -- becoming a Clavister?

It feels good. Actually, as a person with a technical background, it's always good to know that I have a lot of resources at my disposal. So basically, what I really -- what is most appealing to me when joining Clavister, first, the excellent harmony that we had when working with their team for the last several months. And second, the idea that I will have the technical support needed to bring the technology to the market sooner and faster. So of course, that feels really good.

We have a question coming in. First of all, congratulations to the acquisition. It looks like a very good fit. How will Omen as an organization be integrated in Clavister? Will it be a stand-alone company within Clavister or integrated in any other way?

I can answer that. Omen is, of course, a small entity. And we decided, together with Wissam and the team, that it would, of course, really not make sense to have yet another entity as a legal entity within Sweden. So there is a full integration happening where Wissam will form the role of our Chief Data Scientist and had a -- to start with a small tech team or scrum team within Clavister's larger technology department; and the commercial colleague of Wissam, Christian Löfvendahl, will join our sales organization to drive sales on a global account basis towards certain selected accounts where this technology is of extra interest.

When do you think you will have a commercial Clavister product for the market?

The idea is basically that the first step now is that we conclude the project that we have already started with BAE Systems, with Clavister now and the funded by Vinnova. The outcome of this project is a -- close to a commercial product. It's not really a commercial product. There has to be some additional steps afterwards. But there would be an integration of technology and proof points enough to form a commercial product shortly after that. We hope and our assessment is that the CyberArmour product will be the first product where this technology will see its commercial daylight. In exact timing, not decided yet. But given that the current project has run quite far and will end sometime during the spring next year, thereafter, we will be looking at the commercial phase of that one. Yes. So that's where we stand.

Omen is obviously very dependent on the key founders' competence. How will the Omen founders be incentivized to continue with the Clavister long term?

If I start and Wissam can complement, this is a very relevant question. And obviously, one of the discussions we had early on with the Omen team where we are not only buying a piece of technology, we are, in essence, buying domain expertise and very deep competence in this field. And it is absolutely crucial for the entire structure that this competence is maintained and sustained going forward. With that said, we were very happy to learn early in the process that the founders of Omen instead of getting paid in cash, getting paid in Clavister shares because that's, in essence, the best proof point I see of really believing in the strong future together. Wissam, maybe you would like to complement something from your side as well.

Yes. So I'll start where -- from where you ended. Basically, having a significant stake at the company, that's the way we get incentivized. And that is achieved by being shareholders. And also being influencers, especially when it comes to our own technology, and that is also granted. So that's always a good incentive for us.

What about the competition in AI cybersecurity then, in detections?

Yes. Again, I can start from my perspective and Wissam can complement again. So it's very clear what the trend is. The most cybersecurity vendors have some kind of an AI ambition. Now it's very different how they have realized this. Some have -- some went about to do the same as we did now with acquisitions. Some have implemented some -- well, very limited type of AI capabilities, but they still call it a quite comprehensive AI, but it's really not. Then you have the more specific cybersecurity AI companies, for instance, Darktrace and others, who have their core business rooted in AI. So there is certain competition, absolutely. But I think the important message is that the future within cybersecurity is mandated by AI. There is no cybersecurity future without AI. And the reason is very simple. The attackers are using AI, then the defenders need to use AI. If you can't use AI to defend, you're basically out of the market over time. It's not happening overnight. But over time, in a number of years from now, that's the landscape we're seeing. Wissam, what type of competition have you seen from Omen so far?

Yes. So I can add that -- so obviously, the employing AI -- obviously, the industry is overwhelmed today by so many challenges. So there are so many areas where AI can be deployed. So I think that this market will really fit many companies in the near future. But in terms of competitive edge or competitive features that we had is, I can say that we have the -- our technology, as I discussed during the talk, has certain features that make it quite competitive in certain areas. And add to that, that it is patented or patent-pending methodology, and it is based on research. And the researchers behind it are the ones that are going to keep developing it and integrating it into other products. So this also makes a difference. But I can mention, for instance, that one of the key competitive advantages is that it is extremely lightweight. So usually, when you talk about machine learning and artificial intelligence, you always assume that you have a lot of resources available because they are quite demanding, these types of algorithms. But our algorithm does perform a lot of work in this in a very resource efficient way. So it's extremely lightweight despite the big job that it does. And this makes it deployable or makes the component -- the AI component deployable on really small devices, and we have tested that many times now. So I would say that this is one of the key features, then there are other features as well, but -- yes.

Question coming in. So will this open up the market for Clavister to a larger TMM since Clavister is such a small company in the space? Can we expect to see much higher growth figures?

So this is really in sort of 2 parallel tracks here. One thing is augmenting our products and our solutions using AI. That will -- it will, per definition, not open up a larger total addressable market for those products, that will sort of remain the same, but it will absolutely increase our competitiveness. So our probability or possibilities to take a bigger part of that market gets higher. I can mention a very few or a few very important examples on that. When -- especially now after the pandemic has been released at least a little bit and people can travel and meet again, we have been very active and participating in a number of defense fairs across Europe over the past months. And when we meet with the -- both the peers of BAE and the likes but also smaller defense-related contractors and we talk about cybersecurity for defense and we talk about our CyberArmour solution, it is from day 1 an interest. That's clear. It's confirmed that cybersecurity is needed in this space. But when we add the concept of AI to the discussion, then it's a complete icebreaker. Then they welcome us to the discussion, then they invite us to proof of concepts and invite us to continue discussions on a completely different level. So our takeaway so far after many, many, many discussions in that industry is that, well, this is really a value add that makes us so much more competitive. So that's one track. The second track is that, obviously, this is a high viable technology. And as Wissam mentioned, even though it's sort of -- it's meant to be integrated into a larger picture, it still stands on its own as a very capable technology. So we see potential of licensing this technology to vendors who need this capability as part of their industrial set of some kind. Now that's a bit more, let's say, open-ended type of business case because it's super early, it's very hard to talk about the specifics of it yet, but that has the potential to expand the addressable market and provide additional growth vectors. So 2 interesting growth potential is coming from that. Jenny, I think there was a question also regarding Darktrace. I'm not sure if we had that one already or missed that one. The very first...

I missed -- I haven't seen that question that's why I can't really ask it.

All right. So I can just take the question then.

Yes.

So the question was if the technology is comparable to that of Darktrace in the U.K., which is trading at 18x enterprise value over sales, that's a very nice trading number, obviously. So I -- well, if this is the trigger that gets the Clavister trading to that, I'm happy. But Wissam, I mean you have insight into Darktrace and you've been comparing it a bit to Darktrace. So maybe you're best off to answer this one.

Yes. So obviously, Darktrace is very big. But the main distinction between what we do and what Darktrace does is that we monitor what we call process data rather than only classical IP traffic. So our technology is -- was tailored from the beginning to modern cyber physical systems, meaning that the kind of protocols that it -- or the kind of data sources that it monitors is a bit different from what Darktrace monitors. That's one. The second one is that, again, going back to what I called as a key feature for our algorithm is that Darktrace solution is a cloud-based solution, whereas our solution is meant really to run and train and operate on edge devices, which is -- or edge devices that are -- which is becoming more and more attractive with the rollout of IoT devices in pretty much everywhere. So these 2 distinction points, I believe, make us a bit different than what Darktrace does. Yes.

But -- and with that being said, Wissam, with the baseline of the current technology that you have developed and augmenting that over time, training it with new models, it -- my impression is it has the -- at least the baseline capabilities to do quite a lot of what Darktrace is doing. They do it on the cloud side, mainly. We can do it now on edge devices, which is especially in the industries that Clavister are targeting with, for instance, defense where you absolutely do not have a cloud. You do not have a combat vehicle that communicates with the public cloud or even a private cloud, that would be completely disastrous on the field. So that's why this technology fits very well in those segments. And of course, I mean that spills over into IoT devices and industrial IT as well where cloud connectivity might be limited or restricted for various reasons.

Right.

More questions. You're mentioning industrial partners, which industry is most interested -- interesting besides the defense industry? Can you share some details on partners on defense?

Now it's -- if we -- if you recall, I mean Clavister now has 3 primary customer group targets, the public administration or public sector, the service providers and defense. Within public sector, you have, of course, certain subsectors. You have the typical public agencies that everyone is familiar with, like social security and so forth and municipalities. But within that scope, it's sometimes hard to forget or sometimes easy to forget that you have also a lot of mission-critical industries. For instance, state-owned power grids, locally or government or regional-owned power companies, water treatment companies that basically, if you call it mission-critical industrial IoT applications where cybersecurity is super important because it -- the lack of it might threaten the daily lives of citizens. So that's an area where this technology would fit very well as well. So coming back to the initial statement that we see that the Omen technology gets implemented or integrated in the full scope of the Clavister product portfolio because it will actually appeal and apply to many use cases in -- across those sectors. And those mission critical ones are the ones who absolutely have the strongest incentive of using this type of technology.

You started the presentation mentioning the car industry. Are you talking to OEMs already?

No. We're not doing that at this stage. So I -- we started talking about the vehicle industry on more sort of broader terms. The vehicle industry or automotive or transportation is, of course, a big scope. It's a big sector. We start now again with the CyberArmour solution because it's targeting the customer group that we have decided to focus on. Within that space, we have, for instance, the business cases and the business we're doing with vehicles within defense. So it's a subset of the vehicle industry, very much specific to defense at this stage. If we grow, if we are very successful and we get openings into other customers, of course, we're a small company, we have to be open for creating new businesses as well. But right now, our focus is on the targets that we have already set out.

Moving -- how will you -- you continue to build more competence and capability within the AI fields, do you look at more acquisitions growing or growing organically?

Good question. Very good question. This is obviously the first start. We have had AI/ML on our road map, on our wish list for quite some time because we know that this is where the future is heading. So getting the Omen team onboard is the first and most important step. You need the seed to get into this. And Wissam and colleagues is the perfect seed for that. As the company grow as basically any other capability and type of knowledge in our company, we need to grow that as well. We need to follow our general growth path. The good thing with -- and this is where -- Wissam, please complement as well. The good thing here is that this also gets a very natural -- this opens a very natural path into the academic world with Chalmers specifically, of course, but also outside of Chalmers. We have academic connections in other parts of Sweden as well. So this is obviously a field, a domain which is highly research-intensive. And as such, it's really nothing that you build from scratch or organically. It's something that you build in close collaboration with the academic. So Wissam, maybe you'd like to complement.

Yes. Of course. So as I said in the beginning, I mean the start-up was founded by researchers. So -- and as John just said, the -- this type of technologies are pretty research-intensive. And it's only natural that -- given that the -- we have research background and we have great connections with academia and specifically with Chalmers University and other universities in Sweden, it's only natural that we will maintain this relationship during the development of our solution at Clavister. And it is also natural that our -- when our team grows, we will have talents that are relevant from the university basically. So of course, the academic effect will remain with us, researchers, on the team.

I want to say thank you to all the participants because you're really pushing in questions now, and I'll try to put them in some order. First, what about the cost side on R&D, et cetera, and getting to cash flow positive? Maybe, John, please?

Yes. No. Absolutely. Good questions. So again, I'll just start by clarifying that when we, as Clavister, outlined our strategy, some of you know this already, we identified that M&A has to be one of our growth pillars because it would be impossible for Clavister to organically find all the necessary talent and there will be technology gaps and there will be market access that will be best served by acquisitions. When we outlined that part of the strategy, we also defined a profile, a mapping of attributes that would be relevant to really address whether or not a certain acquisition is feasible. And as most of you know, the acquisitions can go horribly wrong, if you have, for instance, the wrong cultural fit or, of course, the wrong technology fit and all of that. So we took a long and hard look at what is the cultural fit, what is the geographical fit, what is the technology fit and of course, what is the financial profile fit. Knowing that Clavister needs to grow into profitability, grow into cash flow positive, et cetera, obviously, we cannot take on companies with massively negative performance financially, that would be against the growth principles. Obviously, we cannot onboard much, much larger acquisition targets with a very healthy P&L, that would be, of course, very good for Clavister, but it would simply be too expensive, just referring to the 18x EBIT sales multiple that was indicated earlier. So in essence, Omen, as now the first acquisition that we have been doing since we have defined the strategy, matches all of those criteria. So cultural fit, both companies are Swedish, both companies take a pride in building carrier-grade type of technology, which serves mission-critical applications. We have been working together now for several months, and it's a good rapport between the companies and the team. The financial profile is such that it will not add -- of course, it will add OpEx to the group, but it will add a revenue stream which matches that OpEx as well already with the current engagements of proof of concepts with customer -- paid customer proof of concepts and other revenue sources that really makes this a slightly but still net positive contribution to EBITDA already next year. Then we have not taken into account any potential upside or any commercial additional license income to Clavister following this. That's part of the upside, obviously. So as an acquisition, it really, really fits all of the parameters.

And on the topic of ongoing projects, the BAE contract, what does it mean for the BAE contract? Will the size of the contract increase for Clavister? Or what is the share of Omen in the -- in that contract?

Good question. So when we signed our framework contract with BAE back in June last year, this was obviously not on the map at all. So it's -- it was not part of the original framework contract. Then just as a reminder, every end-customer engagement through BAE Systems has its own project contract, which is a subpart of the larger framework contract. And each project contract forms the detailed business rationale and business commercial specifics on that project. What we see here is twofold, either we can augment the price of our product because we add new capabilities, which is requested/required by end users or -- and/or we treat this as a general upgrade of the product to make sure that we win more deals, right? So those 2 are the options. And they do not exclude each other. It could be project specific, it could be something we do more general. But sort of the conclusion is, anyway, that, of course, this will increase the value of the contracts we have one way or the other. The exact way, to be seen and to be communicated when it happens, of course. But clearly, it's something that adds value to the contract.

In the future then, do you expect stand-alone Omen products to apply in similar term-based recurring revenue model?

I can say, I mean, in general, our ambition, and you might have seen the consequences already, I mean positive consequences, I would say, in our preliminary Q3 numbers that we are moving as much as possible, preferably the entire business in Clavister is moving towards term-based recurring revenue because that's in essence where valuations exist today. And the Omen technology, whether it's integrated in other commercial products or if it's sold as a stand-alone sort of product, if you wish, it should ideally follow that same principle. So that's the ambition.

And there's no more questions. So I will actually conclude this session. Thank you, Wissam. Thank you, John. And thank you all the audience for participating. Thank you.

Thank you very much for all the questions.