St. Jan Sørenberg, Ph.D.: Welcome, guys. We are really excited to spend one hour together with you on a topic that is really close to our true passion. It's about turning dead calls into valuable actions, what we also like to call speech to impact. It builds upon speech analytics technology, but it's much more than just installing a tool. It's about translating the insights you get into sustainable impact. It's a central element of what we like to call the digital call center. Before we get started, we have a practical note. Please post. If you have any questions, please post them along the way. We will have a Q&A session at the end, and we will also have a few opportunities to do question and answers along the way. Today's facilitators is going to be me. I'm a management consultant at Implement, and my colleague Thomas, who is a partner at Implement. We are so lucky to have Morten, who is the IT director of Semler Group, with us today. He will talk about his true or his real experiences with speech analytics and their implementation journey. When we leave here today, we have learned three things, hopefully. So why is speech analytics and speech to impact relevant? How do we translate the insights into real impact? And how do we get started? The agenda of today, first of all, we'll have purpose and impact. Then Thomas, he will demonstrate the actual tool. Morten, Morten, he will deep dive into the case of Semler Group. I will tell a little bit more about how do you actually get started with this, and we will have a Q&A session at the end. So let's get started. Basically, the digital call center, that's, well, any service unit that is receiving calls, that is also utilizing a broad range of technologies to ensure quick and high quality resolution of customer inquiries. We have been working actively with helping our clients both install the tools, but also get impact out of the technologies that you see on screen here, voice bots, chat bots, call platforms, speech analytics, and intelligent self-service. We would love to talk about these range of things, and all these different types of technologies. However, with the time restriction of just one hour, we deep dive into speech analytics, which is the core of gaining customer insights. So what is speech analytics? Well, you probably have, or you might know of somebody who has a Google Assistant at home, who have Siri on their phone. Well, this is similar. It's software transcribing recorded calls via machine learning algorithms to categorize the data into meaningful insights. It's about identifying the causes of customer dissatisfaction, customer satisfaction, and reveal opportunities for improving your customer journeys. First of all, you start crunching the conversations, figuring out what is the customer's customer inquiries, actually asking us about how do our agents respond to these customer inquiries. You then establish insights based on defined KPIs. And thirdly, and most importantly, you act on these findings, you upskill your employees, you ensure that you improve your no reason to call rate. There are many, many use case examples that you can use here. One pitfall that we often see with our clients is that many things that 80% of the job is done when you have installed the software. We actually believe that it's only 20% of the job done. The 80% remains in about it's about acting on the insights and harvesting the benefits. So that's what we believe is the difficult part. We know when establishing the insights, how to set up relevant KPIs to ensure that it's the right thing you measure on. We can help creating the right actions and translating the insights into executable actions. We know when we see x, y, and z, it also means that you take x, y, and z actions afterwards. And finally, harvesting the benefit of the remaining 20% of the work. That's really important in terms of creating a structured plan for your leaders, creating the right accountabilities, making sure that people act upon what they're actually supposed to, and to mobilize your teams into acting the right way. So speech analytics, that's also a tool and a methodology to kind of as a point of departure for designing your other channels and designing your other tools for both helping you internally and helping your customers. It can help you create a meaningful knowledge base to help your agents retrieve the right information. directly and help your customers much easier. It can help design and be input for how should your voice bots actually act. What should they be able to answer? The same goes with your chat bots. What questions and what answers should your chat bots be able to convey? Your internal digital assistance helping your employees and your agents, that can be a trigger by some of the insights that you get from speech analytics. And finally, self service portals, how they should be designed? What types of input is there in your self service portals? That's also an insight you get from your speech analytics. So what is the impact that we actually see out there? Well, we see many successful implementations here from dedicated companies that result in in significant impact. And here are a few examples. We've seen 31% increase in sales conversion rates and ups up selling. We've seen improved customer retention due to straight through processing, for example, making the customers more satisfied staying with you for longer time. We've seen significant reduction in customer service costs due to improve no reason to call rates. We've seen more call rates. We've seen fewer or significantly fewer internal inquiries between the agents. And then we've seen an improvement, a dramatic improvement in employee satisfaction since the continuous upskilling creates more happy employees. And I would say most importantly here is you get a direct source to innovation of your products and your services and your customer journeys. A few use case examples. A few use case examples I would like to go through, and there are many use case examples for speech to impact. And the method can help you identify training needs of service agents as an example. It can help being the or knowing what the identifying the drivers for dissatisfaction and also customer satisfaction. In a logistics and operations example. In a logistics and operations example. It can help you identify what errors and inefficiencies happen when a product transports from A to B. In a sales setting, it could be improving the cross sales and understanding the customer needs better. And in a more industry specific example here, it could be in a claims handling department where you improve the no reason to call. I would like to deep dive into two of these examples in the explaining you what we actually mean here. Let's start with the claims handling and afterwards go into identification of training needs. So speech analytics when set up correctly, it can be relevant for your customers that you don't, for example, have to read the full policy. Think about your incoming calls, where an example could be 30% 30% of all the contacts that you get are related to product questions. Of the product questions. The total sum here could, as an example, be motorized vehicle insurances. And speech analytics will help you, if set up correctly, break down the reason why you're getting these calls in the first place. So it could be that the customers actually don't understand how their policy works, how their policy works, what types of, or where they can find the policies. It could be that the policies are too detailed and difficult to understand that the conditions are not clear. And there could be many reasons for that. It could also be that the insurance deductibles are not clear, or it's difficult to navigate the self service portal. These insights you would be able to get from speech analytics if set up correctly. In a training situation. It would also help you identify the training needs among employees. So what what normally is done out there is leaders shadowing employees, kind of doing tests on whether they they perform as they should. Here you would get a complete coverage instead of random checks. It could be within the service experience. It could be within the service experience, where the leader figure out that certain employees, they have an inappropriate tonality. Could also be lack of proactivity, continuous silence, lack of empathy, just for example. Within the area of knowledge within service and products, it could be a lack of knowledge within of the product features that you're actually selling. Could also be within return policies or products. Could also be within the service. And then you would be within the service. And then you would be able to identify these pain points with the insights you get from speech analytics. All right. So that was a lot of words from me. Now I would like to welcome my colleague Thomas on stage who would like to demo for you what what is this tool actually about and how does it look? Thank you. Thank you, Nils. As Nils mentioned, speech to impact is 20% about establishing the insights and 80% of the impact. Faceignation data, apple marinating. biking. And now people are pulling out an aperture from There. Usually. Here we have a short description 90% about creating anarlos So a short description of the tool before I go into it. The tool analyzes basically recorded calls, and it does so by splitting up what the customer says and what the agent says. Thereby, the tool is able to recognize when the agent speaks, when the customer speaks. It's also able to say something about how much does the agent speak? How much does the customer speak? When does the agent speak? Is that in the beginning of the call or the end of the call and so forth? And based on that, the tool is actually able to establish insights. And it does so in a variety of languages, and it is able to recognize approximately or plus 90% of what is actually said in the call. Also, as a foundation, it has plus 10 million calls from other clients where we have empirically proven correlations between specific call characteristics and sales customer service performance. This means that you're actually able to set things up so that you from day one can start gaining insights. Let me illustrate how the tool work by showing you a few of the functionalities in the Capture tool. So I'll actually start from the top and demonstrate how in this case, and this is actually live simpler data. And for the same reason, I'll not go into illustrations involving specific agent behavior. I'll just talk you through that. But we've been granted the opportunity to show you live data here from Morten. So in this case, Sembla has chosen five KPIs that are the most important things to focus on, you know, efforting in on the good dialogue. Nils mentioned a little bit about that. The five things are the way we welcome our customers in the call center. And this is an IT call center, I should say. That's the first one. The second one is the extent to which we invite for dialogue. The third one is the degree to which we show empathy in the call. The fourth one is the degree to which we use positive language. And the fifth one is the degree to which we politely end the call. And those five KPIs together constitute the score, the good dialogue, which in this case is 2.7. And we actually have an aim, as you can see, maybe on the screen, of bringing it all the way up to seven. This dashboard is something that is work works as a tool for both the individual employee, but also the team manager and as for management at upper levels. So it's the same dashboard, the same data, live data. And when you go in and change these things that that I'll demonstrate in a little while that you're able to to to to to to do it actually on the fly, recalculates these measures. So it's a very interactive tool. So you could you could ask what lays behind this and and basically it is a set of trackers and I'll just demonstrate how that works. For example, and sorry for this being in Danish for those of you who are not Danish speaking, but in this case, with similar we have set up the tool in Danish. So as you can see, there are trackers that are related both to in this case, the customer, but also to the employee or the agent. Regarding the customer, we are we are we have the opportunity to look at at five things here. We have chosen those things as the most important ones. You could add a lot of other stuff here too. So I'll just briefly go through them. A couple of them I mentioned before they are among the five key KPIs. So from the customer's point of view, to which extent does the customer? Does the customer? Is the customer satisfied at the end of the call? To which extent does the customer show something that we could that we could interpret as frustration in the call? To which extent are there confusion in the call? To which extent can we find something in the call that indicates that the customer has called us before on this topic? And again, we also have a measure on waiting time and frustration in the beginning of the call. And basically only imagination is the limitation to what you could look at. The way you set this up is that you if I go in, for example, on this one, you know, to its extent to its extent the customer is happy at the end of the call. As I mentioned, we can look at individuals, employees here. I won't do that today, but we can also see the things that the tool captures. So in this case, these are the phrases that the tool captures. In this case, thank you and the same and thanks a lot and a thousand thank yous. The way it is set up is that the tool is that the tool captures. is that you set this up as I mentioned that you set this up in a way where the system looks for something that the customer says at the end of the call and actually actually in the last 15 seconds of the call. And all the things that that you look for are mentioned down here. So that's what what you do. And as I said in the beginning of the call, when the solution is implemented, it comes out of the box with all these things based on what has worked on the other 10 million calls from other customers. So it's very easy to get this up and running. I'll just go back here and illustrate something from the employee point of view. Here we have to its extent does the employee or the agent end the call in a polite way? To which extent do we have a lack of knowledge? And that is kind of turned around. You don't want that one to be to be so high. And again, empathy and to which extent do you do you welcome your your clients? And we can look at another one here invite for dialogue. Same thing you can edit the tracker here. So in this case, you can see is the employee who in the beginning, the first 30 seconds of the call should mention things that are mentioned down here that will trigger this tracker. So that I hope gave a short but meaningful introduction to how the tool works. I don't know. Maybe we could maybe we have time for a couple of questions. Let me just see. Yeah, so so there's a question here. We've now demonstrated how it can work in a service desk environment. But what about sales? Can you give some examples? What which concrete results can be derived in a sales or am I saying? And what about sales? And we have, for example, from from a Danish newspaper, where we are able to get 20 percent higher retention rates from customer calling in actually to cancel their subscription. And also from a bank where where we have plus 10. customer satisfaction and actually seven FTEs reduction in handling calls. So this is also a cost issue. Let me see, do we have other questions here? Yeah, so and that is actually maybe a good bridge to where Morten comes in, a concrete example of the insights that are the actions that you create based on these insights. One example could actually be what we did with Simpler was that we, based on the insights we established, we gathered the team and we talked about from a team point of view, what does this tell us without focusing on individual performance, but actually say, okay, as a team point of view, we didn't really illustrate this to the individual on the individual basis, but we said as a team, you perform like this. And that way it made it a little bit easier for the employees to take this in and say, okay, we see that we actually do not welcome our customers the way we should. And then afterwards, when we worked with this for a while, there were still a couple of employees who didn't really get it. And then we provided individual sparring sessions with those. So that's one example of how we worked with it. So I think with those words, I will hand over to Morten. Morten is the IT director with Simpler. And we have worked together on this. Thanks for letting me in here. As Thomas said, I'm the IT director from Simpler group. And I'm here today to tell you about our journey and how we have gained a lot of insights using this tool. And first of all, let me just talk half a minute about Simpler group. Who are we? Simpler group is the largest automotive company in Denmark, founded more than 100 years ago. We have a wide range of companies, mainly importing, importing, selling cars from Volkswagen group, but also other companies who are in the mobility solution area. We are 2,700 employees in Denmark and Sweden. And in Simpler IT, we are servicing not only our own colleagues in Simpler group, but also private dealers who are private dealers who are partners with Volkswagen. So totally about 5,500 users. The past three, four years, we have been through a very big digital transformation. We have replaced a lot of legacy systems and moved into new services, new platforms, platforms in the cloud. And on that note, we also have seen that the systems that are required today in a modern company is not only, it's much more complicated than it was just five years ago. So today, when you talk about a mechanic, a mechanic in a workshop or a mechanic in a workshop or auto technician, it's not only two tools, it's a lot of different tools, it's a lot of diagnostic tools from the OEMs, from the car manufacturers, because the cars today is actually a lot more digitalized than it was earlier. And that also actually requires a lot of new skills in our service desk handling these customer requests. And when I say customer requests here, it's not the end users buying a Audi or a Porsche. It's all the employees and the partners who are servicing these end users, these customers. With implement, we did a lot of work last year, actually started last year with a big customer survey, not only for IT, but for the whole group function. So it's finance, it's marketing, and so on. So it's finance, it's marketing, and so on. So when someone actually contacts a service desk, that's the last option. Our end users actually likes to fix it by themselves, and they are asking for more self-service opportunities. Another thing was one entry to our group functions. So it should be easier to make a request to make a request to IT or finance or HR, whatever. And then it also required more empowerment. And empowerment is actually not only for the end users, for the customers, but also I'm also mean in IT. And it could be. And it could be that the help desk technician gets more administrative privileges to actually finish the request instead of handing it over to, let's say, someone from infrastructure or system admin. And then be more data-driven. And when we talk about insights, you could say, you could say, you could say that we have been using IT service management tools for many years, like I guess all the other companies. And we have a lot of data from all the tickets that are sent to IT. But when it comes to all of the calls that we receive, we didn't have any data. So that's, of course, that's why we look into speech for impact. Learn more about how we can improve our customer experience based on the insights we get from data. And just a short intro to the service desk. So our service desk is, we have 14 agents. We receive about 5,700 tickets per month. Half of them is actually calls to the service desk. We have an average wage time of about three minutes. And the call length is four minutes. And that's not the. We are pretty happy about that. Of course, you could say that maybe we can get the wait time down. That's at least what the customer expects us to do. And as I think that Thomas also mentioned it earlier, when we get the customer, when we have them, the agent receive the call, we can see with the insights here that the the mood, the mood, the mood of the of the customer calling in that about 35% is actually in a bad mood, because they are a little bit frustrated. They have something that doesn't work. And this is, of course, interesting for us to see that how we can how if there's anything that we can help doing better, this way, this way. So so we get the we get the customer to be happier, even when they call in with a problem. One thing could be that we have a better self service tools, better FAQs, things like that. And as I said before, we have about 5500 users. And there's a lot the diversity among these users are huge, because because we have technicians in a mechanic workshop, because we have marketers working with marketing sales force, they are calling to ask IT service desk for help. And that really requires a lot of, of course, of knowledge in the first line of service desk, but also some tools about how to speak with these type of people. One thing that we also see from from from the new tool is that 90% of the problems are actually handled in the first call. So that's very positive. When we started this project with implement, we we we set up some hypothesis, we have a couple of hypotheses, we have a couple of hypotheses, as you can see here, five different hypotheses, as also as Thomas mentioned before. These are the these are the trackers that we are measuring with the Captui. So how can we have this nice welcoming message that will actually improve the customer experience in the call? This is one of the metrics. of course. And as you can see, we have like our our our our goal is that 60% of the calls is started with a warm, a nice welcoming message. And that's still we are still on it on our way. There's still need for improvement on that note. And as the same time, if you look at the goal number two, or a better an invite to a better dialogue. And I think that I was actually quite amazed about what you can do if we looked into if we look back like two, three years, the speech analysis and the NLP, the machine learning and all that was not something that we saw in the Danish language, at least, that were, you know, the, you know, we have to go systems out there, who could actually analyze calls spoken in Danish, but it really has improved the last couple of years. And I think as Thomas also said, based on the millions of calls, the AI and the, the, the engine really improves each day. It gets more and more smart. And of course, we have also try to train it with our with our corporate language, we have a lot of key phrases, we have a lot of specific words that we are listening for. I think that one of you asked earlier today, in Thomas sessions about this, if you are into if you are working in pharma, is that possible? Is that possible? Yes. But you need to train the the machine learning, which behind the tool. And this is also what we did. Out of the out of the box, you get the you get the you could say, the tool that can listen for the mood from the from the caller. A tool that can listen for the chat. And so on. And so on. But if you want to be more specific, you need to train it. And that's, yeah, as I said, that's something we have done as well. I think that as Thomas said, looking into the hypothesis here, getting the data, that's really a nice base for a further understanding and a further conversation with our agents, whether it's leader to employee or you, you do it in a team session, because now we have the insights. Now we have it's not it's not it's not just a gut feeling. It's actually based on data. So I think that's really exciting for us. I won't go through all these KPIs. But I have a dashboard. And as as Thomas just showed us before. These are very, very nice insights, both for the team leader in our service desk, but also for the for the for the for the for the single employee working with this. And as Thomas also mentioned, we have these trackers, trackers that are listening for a nice welcome, a polite end of the call. All of this is is calculating into this KPI, the good dialogue, where the goal is seven. And right now, at least yesterday, I think, it was three. So so this is a nice way of of of of of measuring our calls and and the way we behave. So the key takeaways. And our next steps is actually you could say that data is. You could say that data is nice. Data gives us some insight, but but it hope. Oh, sorry. It really it really helps us give us this common understanding of of how we do a great conversation. So it's a nice way to to to to have this common understanding and and also helping the help desk technicians with and and and and and tools to to to you could say helping them remember to to start the calls with a nice welcome or all these the five goals that we have. It's not a we don't use it for performance of and and of course, it's a it's it's it's it's it's based on data. This is as is as it always is with IT tools. This cannot stand alone. It's not just a tool that you can implement in a week and then you have all this. It requires a lot of a lot of training and a lot of training and a lot of common understanding between the technicians, the help desk technicians. It's it's but it's a very, very nice way to at least start the journey towards a better customer experience because now we have the data and and yeah, I think that the it's it's it's actually what I have and I think that if if some of you have some questions, I'll be happy to answer that. I see that some of them are one is asking about how many resources have we spent spent working on this speech analytics in the initial implementation as I said and as I think as Thomas said. It's something it's it's a cloud based it's software as a service. Our phone system is capable of recording the calls and we are sort of just sending the calls to this platform. And then of course, we it requires some initial set up out of the box. Out of the box. Out of the box. Out of the box. It can it can measure the moods things like that. But but as I said, if you want to use it for more advanced analytics, you need to train it. It requires some preparation because all the topics you need to to to to put in all the topics that you have to do. And and and a catalog of services and catalog of your IT systems. It's like the same as you do if you have an IT service management platform where the help desk technician can tag the ticket tag it with a application service or business service because you use that for analytics afterwards. You will do the same thing with a platform like this. So you need to to give it all these kind of information. I think that we have we will have a short break and because we need to rearrange a little bit and then we will be back in in 30 seconds. All right. So how to get started. Let's just touch upon that really briefly. It's actually quite easy to get started with the pilot project as we did at similar. Our our thinking here is that you need to nail it before you scale it, of course, and that's why we have a pilot project. What is important here is that you need to install the software as Thomas also showed you what it looks like at least and setting up the KPIs for what do you actually want to track in the similar case. It was really focused on upscaling the employee but it can be used much broader in a much broader context, figuring out what types of calls that you're getting. That all depends on what KPIs you set up. So think as think about this as a quite broad range of KPIs that you can set up. Then you need to mobilize an improvement team because as I mentioned in the beginning, it's only 20% of the task at hand is is is the software part. The much harder part is how do we actually act on the insights. And then you start your pilot project. That will take between one to two months. And an example of a project team could be that you have a project leader, you have one to two experts and then one from IT and a process SME. And that all depends on what types of KPIs that you set up and what improvement areas that you're most interested in in your pilot project. Then you need to scale afterwards. So after the pilot project has become a success, you need to scale it, of course, and do informed scaling. So apply the learnings that you had at the pilot project, refine the ways of working in your improvement team, build up the next improvement pipeline and figure out what cases are really interesting for your company. And then of course, ensure the benefits realization that's extremely important. This can take normally two to three months depending on how big your scope is. So that varies quite a lot. All right. I think we are ready for a Q&A session where you post all your questions and all the interesting considerations that you had during this seminar. So let's just use 30 seconds more and we will get Thomas and Morten on the stage. As well as me. So please 30 seconds. Thank you. Hi. So maybe I should start. I just noted a few questions down from the chat that I'd like to address. So we have some questions regarding the degree to which the tool can recognize what is being said. And basically the answer here is plus 90%. We have actually a funny case with Simpler where one of the employees, and he's really a nice guy, but he talks so fast. Yeah. And you know exactly who I'm going to ask. And actually I was listening to some of the calls. I didn't demonstrate that, but you can actually go in and listen to the calls. And some of his calls, the customer in the other end said, what? What? And he has to say it again really, really fast. And so we worked a bit with this guy. And I think maybe the reason for me mentioning this is I think the recognition degree for him was like 60 or something. Yeah. And 96 for the other. So of course it depends on how you talk. Someone was also asking about the dialect. And I think in Denmark we have different kinds of dialects as well. And it's really not a problem. No. But of course that really thick dialect is problematic. Languages, no solution on the market comes with all languages. languages. And no solution comes with all the main languages either. So, but this solution comes out of the box in the Scandinavian languages and in the UK and in German. And then we also have examples of adding language models, for example, the Dutch and French language model. And I think Morden mentioned this about the technical language being added. If you remember me demonstrating, there were a series of words that the system recognizes. And of course, what we did also with Simpler was we added some of these technical words. It's actually my experience. It is not really necessary to do that because to a last extent, at least what we need to know, at least about how we answer our customers, the way we do, is to do that. So, we engage in a dialogue with our customers. The model doesn't really have to know a lot about what a Porsche is or what an Audi is or what a specific server is. But to some extent, we need to do it. But there was one here at the chat asking for, okay, so if we get some more information about what are people calling about, could we then use it for improving our chatbot? Exactly. You have a chatbot because then we can actually turn these calls into a self service. Yes. And I guess Morden, it goes even broader than that, right? Because chatbot is one channel, but it could also improve the knowledge base. Yeah, exactly. Understanding what your customers are actually talking about or asking about could improve the website design, your self service portals, etc. Right? And that is actually part of the solution. I didn't demonstrate that. But there's actually a portal with all the words that the customers are talking about right now. And the question that you refer to Morden was actually, to what extent can we use this afterwards too? And of course, it can be used. I think as I said, I think as I said, also, you should see it as because I know people who are working with IT service management solutions, you are always in a help desk, you're always tagging each request with some kind of tag, whether it's an application service, business service, whatever. And it's the same thing that the system is able to do here. If you have put in the right tags into the machine and trained it because I think there's a lot of, you can gain a lot of useful insights, whether it's for more, you could say, FAQs or to use it for a chatbot or something. There's a lot of data. I think we have a really interesting question here. Maybe Morden, you can answer that. But recording calls that need some kind of approval? So from customers and from employees as well. How did you do that at Sembla? It was actually, yeah, because we didn't do that before. So we are just announcing it when we are receiving the calls. I think many of you know it from, if you call some other service that this call will be recorded for training purposes and stuff like that. So that's how we handle it. And of course, you can opt out if you don't want to be a part of that. And do you recall how many did not agree to this? I think it's less than 1%. So it's almost no one. And I think maybe it's because it's business to business. It's within the company. I think it would be different if you were dealing with B2C. and you know, normal customers. Yes. Yeah. Great. Good. So one other thing maybe to mention, we talked about that we can use data afterwards. But one thing that Capturi, the tool provider here is working on is actually also to, on the inside, to take in emails and chats as long, you know, the same way as we take in calls so that we use the same engine to analyze emails and chats. And that is something that we'll work on for the coming release. I guess you will get 100% coverage then, right? Because it's all your channels. It's both email, chat, phone, and then you get 100% coverage of what concerns your customers, what is important to them, and why do they actually call you in the first place, right? Exactly. And I think just on a technical note here that this tool, Capturi, is a software as a service, doesn't need any kind of installation. And it's really the API is very simple to set up. It's really easy for us. It has been easier than I expected to integrate our phone system and also our ITSM solution to this. So that's always something that you are a little bit afraid of when someone is buying a new tool or you are implementing a new tool. Okay, this is going to be one hell of a job integrating, but it's really easy. And actually just on that, there's a question from Nina here regarding the resources needed. And that is actually what you're referring to here. And maybe we can even be more precise. So we've had a couple of your resources in your department working, but I would guess 20% of that time for a couple of months with this. Yeah. Yeah. I think that's it. Yeah. And maybe a question for you then, Morten. So what surprised you in this journey that you had with Implement and also Capturi? What was the biggest surprise in this journey? I think that actually just the sentiment analysis that we spoke about today with the warm welcome. I was actually surprised that only 10% of the calls have a nice welcome, at least what we define as a nice welcome. And I think it's something that when you hire new help desk technicians or just someone supporting in a call center, the proper training and the proper training and the speech notes and all that is really, really important. And I think it's really important that you need to notch them to remember this because otherwise it will just be day to day and they will live in their own habits. So I think it's really interesting. Just these simple analysis of the sentiment. Also, also more than being able to say to a new employee, go in and listen to these three ways of welcoming in a good way. Listen to these three ways of showing empathy in a call. It's really easy. Yeah, it is. So I guess there's a big management task as well, right? Making sure that the good habits are actually sustained. Yeah, yeah, exactly. Okay. Great. Great. I guess this was it. Was that it? Yeah, I think so. I think we're out of time. Okay. But thank you so much guys for tipping in with the great questions and I hope that you enjoyed it and have a great Thursday. So maybe just on a final note would be let the questions in the chat that we weren't able to answer here will just reply to you individually. Yeah. Okay. For sure. All right. Okay. Have a great day.