Good morning, everyone, and welcome to this OnPoint this Tuesday morning, at least here in Copenhagen Central Time. So, thank you so much for leaning in and participating here in this next 45 minutes. We're excited to share with you a little bit about a generative AI transformation and adoption here in Implement. We have now some insights that we really want to share with you, so hang in there. So, maybe a proper introduction. My name is Kor Ronex. I work in Implement Consulting Group. I've been here for almost seven years. I've been a background within change management, training, education, and recently I really put all of my efforts into AI adoption, which I'm pretty excited to share with you today on the topic here. But together with me is also here to be Tobias. Yeah, thank you and good morning to you all. I'm a consultant here in Implement Consulting Group and I've been working with AI for the past seven years or so. I have been technical. Right now, all my energy is focused into helping organizations navigate how do we do strategy, how do we do adoption, how do we build technical things that can assist us in our work. And, yeah, have a lot of energy in that, I would say. So, super curious, you know, about this topic and looking forward to sharing with all of you. And Implement, we are fortunate to have many talented colleagues who also could have been here today, but today it will be the two of us who will share what we learned and what we've seen out there. But maybe we should look quickly at the agenda here for today. So, we will give you a little bit of a fly-in. That's you to be setting the scene for the AI adoption landscape. What are we seeing? What trends are there? But then after that, we will zoom into our own internal program where we are able to achieve what we say 80% adoption rate. And we will share a little bit of lessons learned to do that, but also a little bit maybe to avoid what we learned down the line. After that, we will share a little bit from a client perspective what we are doing with clients and what we learned from their projects that hopefully you also could benefit from. And lastly, we will consolidate all that down to what we call Implement's approach to driving successful AI adoption program. And if time is allowed for it, there is Q&A. Feel free to put your questions into the chat as we go along and we will try to answer them as we come to it. So, that's a little bit what we have planned for you. And, Sopirius, maybe you will take us through the scene? Absolutely, I will. And thank you for that call. So, let's have a look and let's talk a bit about the AI adoption landscape as we call it. And actually, we're going to start out with a study coming fresh off the press from the US. A study surveying a representative group of workers in the US assessing their use of generative AI. Quite interestingly, I think it's that 30% in this survey actually indicate that they have used generative AI tools. So, it's not just something that is coming. It's actually here and now. And I think also interestingly is a third of this group or approximately 10% of the full population actually indicate that they use generative AI tools on a daily basis. So, that's interesting. And I'm pretty sure this also mimics quite well in the European landscape. But this also tells us that we're still people who are here, but also there's still room for more adoption, isn't it? Exactly. Yeah, exactly. I think there's plenty of room on the wagon, if you will. This is, of course, one study. And in the implement team, we've actually been quite busy. We are following a few of the studies coming out. This is a little bit of a hard slide to present. I think the main part is that a lot of the studies coming out are assessing a productivity dimension. This is the top. There's some smaller lines indicating individual studies. These are across industries, across working functions, etc. But on average, what we see is that when it comes to productivity, the average effect is a net 33% in the product. Positive direction, which is a positive direction, of course. This means people spend less time on their tasks or they complete more in the same time. This differs a bit whether it's, for example, customer service. That's more towards the lower part. It's high volume, but it's roughly around 8%, 9%, 10% effects. In the other direction, it's a lot about IT. It's a lot about coding. There are really some classic 2x use cases, if you will, in that space. And I think around the average are a lot of the creative work and writing tasks, etc. So this is one dimension. The other dimension is quality. Not all studies that we surveyed actually assess the quality dimension. It's super difficult to assess quality because what is a good email and what is a good PowerPoint slide? It's not as easy as math. But some of them do and some of them score each other's responses. And they actually see at the same time as the productivity increase, also roughly an 18% increase in quality. So it's not either or. It's both and. And I think that's super, super interesting. If you just pause for a second. I mean, look at these numbers. I cannot recall last time we could see a technology could do that huge impact, right? So this is crazy looking into it. I mean, maybe it's more or less different depending on your roles and sector. But anyhow, this is huge. Yeah, exactly. And I think that could be another webinar diving into the different industries and types of work. But this is the big picture, right? Yeah. I think on the side of this is also interesting that we actually see some places where this is something that also drives employee satisfaction. I think that's pretty interesting and pretty important. But also the fact that I think the younger people coming out of universities, they expect organizations to have a tool or to have some stance on this. For sure. Because they also use it for their schoolwork. So I think this is also something to keep in mind. Yeah. But it's also, it's a very, you know, the tools are super cool. They are super powerful, but they're also quite weird. Let's be honest. Yes. And I think on that part, we sat down and thought a bit about what makes this actually unique for the end users. One of those things is what we call quality dependence. How I use my generative AI tools and how Core uses his, it depends on how well we prompt, how well we think, how creative we are. So it's not our same point of view into an ERP system. It actually depends what we do with it. Yes. So there's a quality dependence and that's super interesting. There's also what we call a dynamic evolution. Starting out two and a half years ago, a lot of this was ChatGPT. Now there's a plethora of tools that you can access. If you write a lot of code, you will likely want to use the model called Claude, Bionthropic. If you're doing creative work, maybe it's something along the lines of a classic ChatGPT. If you're doing a lot of research, it might be Google Gemini. So that landscape becomes a little bit blurred and it evolves dynamically. So both new models, but also different ways of using them. And I cannot recall any technology that had upgraded itself that fast. Just look at the last year. I mean, last three months. I mean, I can never recall. I've seen anything that rapidly changing, right? And keeping up with that. Wow. That's tough. And that's also why we do this, right? Because it's quite new. We also talk about individual use cases. So again, back to the example from before, I use it for different things than Core might do. And similarly, within your organizations, there will be a ton of different things that you can actually use this for. So it's a very individual tool at the same time. And I think this flows nicely into number four. We talk about blurred boundaries. I don't know about you, Core, but I don't use an ERP system at home. But I do use ChatGPT at home and I do use it for work. Gildiest charts. Gildiest charts, exactly. So kind of the lines between the work and the personal spheres here are becoming a little bit blurred. The boundaries may be blurred. Maybe we have guidelines for what we can do and what we can't do. But those definitely don't count at home to the same extent. So this is really an interesting thing kind of seeping into organizations, giving a blurred outline, I would say. And I would challenge the most organizations probably don't have clear guidelines what they can use it for and not use it for. At least being really saying this task is a no-go somehow, right? Yeah, exactly. Or at least maybe. Yeah. And I think that's a great segue to number five, actually, Core, which is about these risks that we associate with it. There's a lot of the ethical use of generative AI tools. There's definitely also some legal risks to this, the EU AI Act coming out. It's really, from our experiences, it can be really hard to actually communicate what do those mean for us. Yeah. And I think this is also something that makes it unique. Sure, we have guidance for how we use ERP systems and Excel and what we can Google and what we can't. But generative AI tools are a different breed, I would say. Definitely. Definitely. So this makes it unique in our point of view. And I think just moving forward a little bit as well, we work with a ton of organizations around this. Some have moved pretty far. Some are still in the starting blocks and some are somewhere in between. How we see these journeys unfold is typically something along the lines of this. From minimal reactions on one hand to super neatly integrated into daily workflows on the other and everything in between, right? So we're seeing a lot of companies having a minimum reaction stance. This is about some sporadic check GPT use, perhaps. People will find their own tools. They will find their own ways into them. Moving towards that initial exploration. Maybe we buy a lot of things. We buy a few licenses. It can be copilot. It can be chat GPT. It can be stuff like that. And we start to explore a little bit. What might we use this for? If I may say transport company, what might I actually use it for? This is a natural first step going into this. Then it becomes more structured. There's something about structured adoption. And this is usually where we see there's a broader implementation of a tool. It might be one that you build yourself, an internal GPT, like we call them. But might also be something that we acquire on a license basis. Moving even further, we talk about advanced implementations. It's where we actually still have the same backbone more or less. But this is where we might put in our employee handbook. We can take a little bit of load off our people and organization teams by actually embedding this and having not just the normal chat GPT, but also actually being able to ask, you know, can I bring my dog to work? That's kind of the next step. This can look, I think, differently across organizations, of course, what brings the most value. But I do think the end state is more or less crispier, right? It's about across our teams, across our departments, what might we actually use it for? If I'm a P&O team in the pharmaceutical industry, what are my use cases? There's likely a lot of onboarding. There's likely a lot of hiring. There's a lot of contract work, etc. And that's where we really go deep. And I think on this picture, we're actually also curious a little bit on you attending today. Where do you actually see yourselves? Where are you sitting right now? You may not be in charge of AI adoption in your organization, but maybe just what's your point of view? And I think we will have a small poll popping up in just a second. But we're curious, you know, from the minimum reaction all the way up to what we call workflow-based use cases, where do you see yourselves being right now? Yeah, this will be interesting to see. And as you mentioned, we see organizations be all over the plate here. So just good to see where we are. Yeah. And I think, so we need to lean in just a little bit, but we are seeing some of the first results. We have a little bit of answers coming in already. Maybe people are sitting in the car or something, maybe hindered to click in here, but could be. Right now we have a pretty neat spread, you could say. I think there are a few people on the initial expiration. So licenses coming out, we're testing it a little bit. We're seeing what's happening. We're seeing where the value might be. And also we have a few on the structured adoption and the advanced implementation and also actually one with the full shebang, if you will, the full integration into our business workflows. What can you relate to that with what you see out in industry? Is that also the picture here? I would say so, yeah, actually. We don't see a lot in the minimum reaction. I'm sure there are still some, absolutely. But I do also see that it can be hard for organizations to take a true lean back stance on this. We know there are still a few organizations that I work with that have guidelines saying maybe we don't want to use this too much. And I think that can be a difficult stance. People will find these tools. They will become more and more available. Maybe not necessarily just as tools, but it can also just be a button. If you use a minty.com sometimes, for example, you can actually group your responses. That uses generative AI as well. So this is definitely coming out there even if you want it or not. Yeah. Okay. Very interesting. But I think this is a pretty good picture. We are definitely seeing some moving, also moving super fast. Yeah. But also some that are in the initial exploration phase. So super interesting and thank you so much for responding here. And I think it's time to move on because I think also given this picture, it makes sense to talk a little bit about some of the benefits of the data that we're doing. Yeah. And then I think we'll do another poll after that. Yeah. And then we'll talk a bit about our own experiences. So moving ahead, talking about typical barriers to user adoption of AI. This is, I would say, a hard picture because this is not true for all. But we do hope that some of you might see yourselves in some of these. There's definitely one part about the lack of access. We talked a bit about it already. If we do not have a tool, people might find their ways around them. They will find different tools. But also we might have a tool. We might have a certain subset of licenses that we need to distribute. We have that as well. Copilot. Copilot, for example. So there's definitely something about actually giving people a structured, secure, safe access to these tools. This is super important. If that's not in place, it can be a barrier because then it becomes a mixed landscape. Some will use Claude. Some will use Mistral. Others will use ChatGPT. And it becomes a dispersed, adoption, right? Yeah. And you don't need to have done that many times here. Don't have access. Then you don't go see and do have access now, right? Exactly. Yeah. So it needs to be easy, right? We'll get back to that as well, I think. There's something about lack of time. We hear this a lot. And we totally understand it. And we do our best to try to navigate around it. This is true for ourselves as much as it is for the clients we work with. Back to those productivity effects and the quality effects. Of course, there's likely some time coming back. Hopefully more than we we can invest in the training and communication. But this can be a key barrier. Because if we don't have time set aside to actually start playing and start using and start getting value out of it, we won't get it quite simply. I have this picture in my head where people are pushing this old wagon with square wheels, right? And one is offering the round wheel and say, oh, we don't have time for that, right? Yeah. Exactly. And I think this is the same. And I think the interesting part here, of course, is these tools are roughly two years old. They will develop a lot over the next year. And I think the faster you get through this time barrier, the more benefits you're going to have in the long run. Mm-hm. There's something about awareness. I think this taps into more or less the same, really. But of course, do we have an awareness about these tools coming out? Yeah. What is our structure stance? Maybe we even talk strategy about Gen.AI. How do we do this? And how do we actually bring this out to, you know, not just within one office, but across offices, across geographies, across different types of work, what does a finance person need to know versus what a P&O person needs to know? Mm-hm. I think this is quite critical. And that awareness can be difficult. Mm-hm. There can be a lack of trust. Mm-hm. These tools are in part new. We sometimes talk about what we call GBT hesitancy, right? Yes. And there can be an individual part of that on appearing more or less skilled. Yeah. But there can also be an organizational part of that. Mm-hm. Do we trust these tools? Do we believe that they bring value? Do we trust that the tool we have gotten is the right one? Yeah. Do we trust that, that, you know, we can use this freely, et cetera? Yeah. Can I talk with business sensitive information there? Yeah. Yeah. Right? Exactly. So, so trust is super important. Mm-hm. I think trust first and then, you know, maybe time afterwards, right? But, but this needs to be in, needs to be in place. For sure. Yeah. There can be something on the, we call it poor quality of outcomes. Again, it goes back to the way that Cor and I use these tools, just like you might do yourselves. It differs from how we use them. It, it, it depends what we use the tools for. It depends how, how good our tools are. And sometimes we can get a bad experience. And I think from the adoption point of view, how do we tackle bad experiences? Mm-hm. It takes a lot of good experiences to build new habits, but it, it only takes one bad experience to, to start breaking those, right? Yes. Yes. Um, so, so that's super important. And I think this ties super nicely into the next one, which is about the expectations. Uh, we need to manage expectations. We need to understand that sometimes these tools may have a hard time solving certain tasks. And sometimes it doesn't make a lot of sense. It can be really good at making a photo of an elephant, but it can't read our contract for some reason. Yeah. Right? Yeah. Um, so we need to set expectations on that. And I think the good part is that they are only moving in the right direction, these tools. They are getting better and better. Mm-hm. But it's something we need to, to cope with, uh, as well. Mm-hm. I think, again, we're gonna do a, a short poll. We are curious about what do you see in your organizations? How does this map that map? I think you can only choose one. So we're curious, you know, which one of these actually fit you the most of, or, uh, describe best your organizations and your current status. Um, really curious about that. And maybe at the, uh, the same time we can ease up a bit on the next section call. What do you see for us going on here? Yeah. So, um, definitely there's something about, uh, getting the last, uh, 20%, getting to prioritize the lack of time. Uh, we're definitely seeing that the more senior you are, uh, the more the calendar is booked, the more the harder it is to take it down. And, uh, we start now to do and dedicate. We're hacking actually, uh, these, uh, managers, uh, uh, uh, calendar, uh, uh, going into where would they already have scheduled meeting? And then we can see, okay, now they can see the value in it, but we somehow need to hack it to give them that time. Yeah. Into it. That definitely, that's a big part of it. Uh, but I can recognize all of them in our initial, uh, journey, right? Yeah. There's also some, there's, there's a, there's a, there's a neat part about, um, we don't need to convince a lot of people about the ERP system. No. Uh, but we do need to convince people about the general survey are tools, right? And that, that needs to be value based. And I think that's the new thing for sure. All right. So you can see here, uh, 33% say lack of time. So they definitely also recognize that. Uh, but then, uh, 70% say the limited awareness, uh, also into it. And, uh, the lack of trust also, uh, 33% speaks into and then, uh, unclear expectations. So I think that matched really well also what we're seeing out there. Yeah. I, I agree. So, so I think, uh, from here, this is the landscape we see, uh, we see a bit about this is a unique tool. It brings pretty big value benefits. Uh, there are some typical barriers to this that are a bit different from other tools. Um, and I think this is a great segue to talk a bit about our own experiences. Yes. Uh, so maybe you'll take us through that. Yeah, for sure. So the next part of the agenda here is from our internal, uh, program. As I mentioned, we managed to reach 80% adoption rate today. So we will talk a little bit through how we got there. And I think we are a little bit over in a year into that journey. Uh, so what took us there and what did we do? So the three things that really started out really saying that was our three non regret moves was saying we need to have a dedicated task force working with this. We call it the three T's. So the task force tool and training. So a team that was focusing in on this, who have the knowledge, but also have the organization knowledge and also having the right mandate and leadership support into it. I think that triad really was a key in that. So the next. So the next step is that we're going to implement, uh, uh, uh, success so far. Next is having the right AI tool. Of course, it goes without saying that you need to make that available. So we develop our internal GPT, but also acquired the copilot licenses. We're starting building other AI tools inside and that really gave us some insights. What people are using for and not and it had developed a sense in. And then of course last we need to learn people how to use these tools effectively and responsibly. Uh, and we did that with training and just to say that was not a good idea. So it's not a one off training. We're doing many things and maybe that's a speak to in a second, but with you are not doing these three things in your organization. That would be our recommendation. We call it the AI starter kit. Uh, so start doing these three things would really help you out the way. We actually wrote an article about it. Uh, you might want to dig deep into. So if we look into the upscaling part. We are doing many things because we recognize this is really an adoption challenge based on these barriers just sought to speak to. So these are the steps that we're doing. So these are the six things we're doing at the moment. We're doing a virtual training sessions. I think right now we have three different courses, uh, in our internal, uh, academy. Uh, and then we learn people the foundational, but also a little bit more advanced how to leverage and use it. And we actually go with a virtual class, which works really well at format. Um, and the next thing we did, we did, uh, intensive weekly newsletters for the whole organization for quite a long period where we went pretty practical on what are we, how can you use it in your organization? We're, uh, daily context with identified use cases and then we, uh, communicate that really, really simple. We call it almost micro learning, right? Uh, and we can see that have a directly linked with the, uh, adoption rate. Next is what we call the Gen. AI case challenges. What we did there, we invited ourselves into a scheduled department meetings and then we spend, uh, three hours together with the organization or the department and actually working with a real case how they could use that solving with their Genitive AI and being really practical working with your partner. statements and uh, as we as its global 했어요, uh, we you know, Хот, um, this was all of the time that was no particular part of the cloud ecosystem in marketing practically. prioritize. So, we find out who should be writing of the community, how often should we do that, some big learnings into that. Next is the leadership AI training. So, we can see that to make sure that we get everyone aboard, we need to do dedicated training for leaders in this space here. That's really something they value, but also we need to have to mention a lack of time to hack them. Yeah. Again, they're right. Yeah. Lastly, which is quite odd, we have something called AI DevLab in here. But what we did, we actually do a lot of prototypes, small projects, pilot projects to see how could we build AI tools and like why is that upskilling? Actually, because we saw this was pretty important for our clients to see could we come out with product services that are Gen AI based, but also developing these was a lot of learning as a team to dig into the research, dig into the tool, what can it do, what can't it? So, actually we see that was actually 50% upskilling initiative. Yeah. And I think on that note, right, it's also where we move from maybe being at the center or a bit center plus on the maturity journey to the full step. This is where we go deep, you know, we sit down with finance and say, what do you do? Just hearing that I think brings a lot of value for the technical people. And then also the technical people working with and expanding back what can AI actually do? What does it mean that something is a rack or what are the differences in models and how much can they read, et cetera? This is the next step, right? Definitely. But also, it was, I mean, a little bit of bold step, but also we didn't know exactly what would come out of it. We just try to let it go and see what could people actually use it for. And I think that was an interesting thing in approaching something like that. All right. But maybe we should look at our adoption rate and we can do that by, for instance, also look at our daily unique users of our company GPT. We call it IMGPT. So, we can see we launched it back in 2023. And these are actually mentioned unique daily users, a snapshot of that. And there was a little bit of a soft launch towards a smaller department that have access to it. So, you can see almost not a lot of traffic, but then you can just see how it's growing there. And what we could see that we found out that we launched a tool without really having any training. But then in January, people come back from New Year just excited. No, what happened was we actually launched our internal training into it. And you can see directly landing the unique users in there. So, I think just from a change perspective, this is pretty cool to have the data to back that what we're doing is actually working. And as we can see, it stagnated a little bit, but then we need to do more. And that's why we start building in also these weekly newsletters. And then we can see these spikes every time. That's actually every time we launch these newsletters. And this is just an interesting field to get the data on how is it working and how are people responding to what you're doing in there. Maybe what you can also see, of course, outside of the summer holidays, a decline. But actually, we see a slowly declining right now in it. And we believe this is due to our tools actually a little bit outdated compared to the public model. So, what we're doing right now is continuously improving and releasing new version of it, which will dramatically lead people more into our company GPT. So, I think that's also learning from us. You can really easily lagging in terms of your own development programs. And I think it goes back to my point from before, the number two, the dynamic development of these tools. Suddenly, we have a deep research tool that's super valuable for other people. We don't have that, so we need to do something about that. This is the safe, compliant use. And of course, those declining numbers might mean that people might gravitate towards different tools. Which means there's some value to gain from those. But it's on our agenda. For sure. For sure. What we also did was do a survey after half a year in the organization to see how are people using it, who is not using it, and why is that. And we could strongly recommend you having done that already to do that because we really find some interesting insights and knowledge and also could target your initiative. So, what we found out that 65% was using our tool daily or almost daily. And 17% was using it weekly. That combined, we said we have an 80% adoption rate. And by the way, we had pretty good response rate into the survey here. But then we also find we have the last 19-20% who are not using it or almost never monthly using it. And we're like, how can we get those back on? And that was really interesting to see. And that's who we are talking about now. But what we asked people, what are you using it for? And they say something, of course, is summarizing, consolidating, translating. I think that's obvious for most people. But then what's striking to me is as a sparing partner. And here you found out that this was about actually asking safe questions or you found maybe stupid questions, but do that in a safe space. It's really valuable, especially for the younger generation. And then drafting emails document, of course, goes without saying. But also creative writing and task and reviewing my work is pretty cool in there. So, that just tells us that there are huge opportunities for almost everyone in there. When we ask them what challenges, also what you're speaking to, definitely the full capabilities part. And that's speaking to the training element or capability building, right? So, make sure to invest in that. And then also the unsatisfying quality. And that's correlating by who find there's bad quality and who participate in training. So, I just think that's interesting. So, if you have participated in training, you usually know how to prompt and use it, right? And of course, a lack of time. You speak to the productivity earlier. We definitely also recognize that. But also creativity part of something people get out of it, like the huge creativity and quality gain. And it's hard to measure, but we recognize it. Yes. And when you're doing a business case, this is something that what can we measure and what is maybe hard to measure. Exactly. This can be stuff like, I'm working with a team across time zones. What do I do? It's not something we can ask Excel or we can ask Google, but here it's definitely a stronger context. But the mindset getting us to think about those types of questions, that can be super difficult. That can be super difficult to train as well, I would say. There's something about setting a foundation and then letting people loose and seeing what they come up with. But there's no user manual, right? We can make one and that might help some people, but the capabilities are much bigger than this picture as well. I see. For sure. For sure. So maybe what we did here. So what did we learn? So we definitely see having a strong leadership support was a big part of our transformation. We able to push through with initiative really quickly. The next thing, this was not a technology implementation. This was really about understanding how this could be used, identify use cases, make sure that we are tackling their reality and it's like, how can AI be part of that and where can it not? Next was really about saying we need to focus on these use cases. And we hear that a lot in Genitive AI. Use cases, use cases, use cases, but there's a reason for that. And really nailing that really early and continually as they improve these tools, right? And then lastly, I think this is a culture thing because we see many people have this hesitance you were speaking to, but how can we make sure it's acceptable to say, I used Genitive AI for this and be open about it and also we're using together in meetings or whatever it is. But also saying, here we should not use it. And that as a culture, we can survey people saying, did you find it acceptable to use it? And before we did this huge transformation part of initiative, they said no. But today, almost everyone is completely acceptable to say, I've used Genitive AI for this part, right? And I think this is where you also tackle the things about, is using this cheating? Yeah. Well, if you can leave half an hour earlier, then I think that's great, right? So I think this is really key. But also, how do we give feedback? Yeah. Feedback is hard enough on its own. But when we put AI into the mix, that becomes super difficult. But having that courage to discuss it, how can we use this in our projects? Do I think that your AI written emails are maybe a bit on the nose? You know, how do we have those conversations? That can be super difficult. For sure. Yeah, for sure. And maybe lastly, also just talk about what would we do differently looking retrospectively. The community part, we try to have a high cadence almost every month and try to force it in there. And it's like, if there's not enough meat on the bone or people to prioritize it, it will fall apart. So we need to define it as something that is more flexible. And we find actually a little bit antichristian teams channels or a little bit more rarely when we should meet works better. Next is our license governance. We had really a lot of people who are signing up and being on waiting list for a license. And that a little bit kills it for your AI adoption. And also a lot of people were not using their license at the same time. So we found out having a hard governance on people using their licenses was a big part of us leveraging that. And then, of course, as we mentioned here, our AI tools need to be continuously loved and have some improvement into it to make it able to stay in the forefront of the field. And on the final one, I would also note that this is still early days. It's not too late. There's a ton of tools coming out. And maybe it can also be beneficial to try out a few, even if you're just starting out. Let's try some co-pilot. Let's build our own tool. Maybe get a few licenses for a third tool. I think that can be a nice way of exploring in a structured way as well. But they definitely do need love. Did you do? That was a little bit our internal program and what we learned. But maybe we could speak a little bit towards what we see with clients out there. Yeah. All right. So from what we see from clients is that when we need to do generative AI, it's pretty important to go into an experimental mindset that you not will have all the result and impact up front and know exactly what it will do. But going into an open mind about what did you do, but also really involve the right people. We're talking about the right SME, subject matter expert, but also technical people, but also who have the mandate and power. Having that right mix of involvement really early is pretty important to make sure the training material, the tool is fit for the right purpose, right? Next is about setting the targets and measuring impact. So we actually find this is a place where it's both easy directly to track how people are using it, but also there's something that's difficult to track. But we will dare and ask you to actually go in and be able to track and measure your initiatives and have clear KPIs and KPIs on and assess it. And that really would help you steer towards what initiative you should do. And if you don't have that in place, sometimes we see this dies a little bit as we go along the way. So don't be afraid of setting targets and measuring them into it. And lastly, that is really having into here the UX and functionality part that you have the right tool that are meeting the right needs out there. But Tobias, maybe you speak towards more of these. Yeah, yeah. And I think back to the thing that doing structured adoption means convincing potential users about the applicability and the value. Yeah. So if we have a tool that is slow or maybe not fit for purpose, maybe we even build a rocket ship, that can actually also be another danger, right? Yeah. The tools need to be as simple as possible. Yeah. And we need to build from there. And we need to build up the tools while we build out our organizational capabilities. But to me, the UX and functionality part is super relevant, right? If I need to be able to upload a file, we shouldn't acquire a tool where I cannot do that. No. But also, it's nice if this is maybe on my desktop. So I know to click it when I turn on my PC in the morning. This is what we see going into the UX and functionality part. And maybe just a pitfall here sometimes. Those who develop it are towards a more technological early adopter. So they know all the small features. Yeah. But to get everyone on board, maybe we need to have people to look at it and say, okay, this is not intuitive for me to use. Exactly. I think that's a great point. Yeah. Yeah. We need to develop this with the users in mind at the end of the day, right? Yeah. Yeah. I think also something we see is, of course, there can be some differences across, in part, roles in the organizations, but also generationally. I think it's too simple to say that it's easier for younger people to get started compared to more senior people. Yeah. I do think, actually, more senior, more experienced people have a better better frame into it, right? They know what good looks like, more so than the young people. True. But they may have a harder time picking up the discipline about prompting. Yeah. And thinking creatively. Yeah. So I do think here, and back to your point about community building, this is where that becomes really important. Yes, we can learn by each other, right? Yeah. So somebody knows how to use, but the other person might know a little bit what good looks like. Yeah, exactly. Yeah, exactly. So I think this is something to note. I wouldn't do too much about it. I wouldn't make dedicated trainings. I would say based on age groups in an organization. Mm-hmm. But I would definitely keep it in mind. I think the community part here is super, super important. Mm-hmm. I think finally, there's something about trainings, and there's something about prompting. Prompting is new. Yeah. Yeah. I think, in short, trainings must be practical. Yeah. We can do a lecture. We can, but it doesn't bring a lot of value. Mm-hmm. I would say get people in a room, preferably a smaller room, and have structured playing around. What can I do? If people ask, can I solve this task? Mm-hmm. Then say, maybe ask the model. You know, exploring from there. Yeah. And I think also on the prompting part, it's important to not over-engineer it. I wanted to bring this bullet today. Yeah. Because prompting is just one part. Yeah. And yes, it's nice to have good prompts. Yeah. But at the end of the day, it's a mindset thing. Yeah. More so than having a good back catalog of prompts. Yeah. Again, it's not either or. Mm-hmm. It's both end, I would say. And I would say, as a mature, the model mature, it gets less and less important exactly how you're prompting, right? Exactly. But it's still important today. And it's just a way of thinking and asking good questions and stuff like that, right? Exactly. Exactly. Yeah. All right. Yeah. I think we are on for the next, the final agenda point on our side. Mm-hmm. Just a few slides talking about how we believe we drive AI adoption successfully. Yes. We'll have a little bit of time for Q&A afterwards. So, if you're sitting out there with a few questions, please feel free to drop them in the chat. Then we can get to those afterwards. Yeah. But maybe first call. Yeah. So, what we did after, now we've been doing our own stuff and we're helping a lot of a lot of clients. Like, how would we recommend people moving forward in this space here? And what is maybe a little bit different compared to some other technology implementation is, generative AI could be approached from two angles simultaneously. So, we definitely need to have some pressure from the top-down perspective where we have, what strategies do we have, what tools do we invest in, how do we governance this, where we will not use it, where we will not use it. There's quite some strategic work into whole this field here. And that's important to do. But it cannot happen alone because you're pretty sure it will hit, miss a spot. Yeah. What you need, you need also to activate the grassroots movement. You need to activate the early adopters who are already using it and engaging that community, engaging those passionate people in the organization and saying, okay, where should we use it? How are you using it? Could you please tell us? And what are you missing? What barriers? Should we help you remove? And I think that comes together when you approach it that way, then you will start to see some magic coming out of it. Yeah. So, that being said, we have consolidated all what we have said today down to a model and implement we love around model. Yeah. So, this is our approach to generative AI model. So, in the center here, this is about generative AI adoption for the end users. How can they use it in their daily lives? Yeah. So, Tobias, what is in this model? Yeah. So, I think, and I don't like the word holistic, but I think it applies pretty well here. Right? If we think we can do generative AI adoption by only doing a little bit of trainings, we might, not 100% surely, but we might miss the marks. Yeah. If we only engage leadership, we might also miss the mark. But this is our take on what we believe actually is important. Mm-hmm. We believe, like you say, performance management and setting some targets to follow is key. Mm-hmm. We believe that having the right functionality, tooling, UX, that whole experience needs to be in place. Mm-hmm. That needs to be something to engage the end user base on. Mm-hmm. Mm-hmm. Leadership needs to be engaged. Mm-hmm. I mean, we see that on average across AI projects, having a strong executive backing increases the probability of success by, you know, 80% or something like that. Right. But also because leaders can use this as well. Mm-hmm. And the more they do so, the more they will likely also bring the rest of the organization with them. So, that's super important. Yeah. You don't come around to trainings. Trainings are important, but they are not the big, like the full picture, right? So, I think being creative on what that looks like. Yeah. It can be a 45-minute setting in a room where you show a few cool things and then go back to your day. That can be a training. It can also be a two-hour structured virtual training. And maybe just saying, yeah, this is not a good idea. It can be a lot of one-off. So, you participated in a genitive AI training one year ago. You need to go do something again, right? Yeah. Yeah. At least if you want to stay up to date. Yeah. I agree. Yeah. And I think also then when we do structured work around adoption, it's also a lot about communication. Yes. Why are we doing it? Yeah. When are we getting this new tool? What are the next features coming? Why are we doing this? That's super important. I think both in those adoption type projects, but also as a continuous thing, keep communicating. What are we doing? What is coming? What's on our agenda, right? Yeah. I think that's super important. And if I just may chip in, I think what sometimes can be challenging is this can be quite technical really easily. Yeah. So, how can we communicate something that is technical without making it too difficult to understand? That's really a challenge to the communication partner. 100%. Yeah. And again, I think that's why you will likely have people driving AI initiatives that may be a bit AI-heavy. Mm-hm. That's normal. Just like you want ERP-heavy people doing ERP. Mm-hm. But I think again, thinking about the end user base. Mm-hm. It's super broad, right? It's a full organizational thing. Mm-hm. And the communication, the tooling, et cetera, needs to be tailored for that, I would say. That's super important. Yeah. And the final thing that's super hard to work with, right? That's something we create along the way. Mm-hm. But the culture, the ways of working, the ways of using these tools. Yeah. Yeah. I think that's also something that begins coming to life when we start training, we start engaging in the relationship and we start communicating. Yeah. So, I think this is our best bet of doing a full model. How does this look like? What should be on your radars up there? Yeah. And then you can apply this on a department level. You can do this on an organizational level. But I think it's pretty important to say there are many facets to succeed with this. And if you only play on one of them, it's better than nothing. But you might miss something if you're not. And I think this is also where the target setting comes in. Right? So, if we have an ambition to train all people, let's say that's the ambition. Mm-hm. Then we also know where we are going, right? Yeah. So, I think setting those targets, talking strategy. Mm-hm. Setting targets and then doing initiatives to back up those targets. I think that's the way to go, of course. Yeah. All right. That actually also marks towards the end of what we had planned for you this morning here. But we actually have time for a small Q&A here in the end. So, if you have any questions, please put them in the chat. Then we will try to speak to them. Yeah. As we see anything coming in here. So, that was a little bit what we had planned to be is what do you see here coming up? As we see if any questions come in here. But the next couple of months, what's on your radar here? Yeah. That's a tough question. I think as we see more models and also we're seeing more types of software coming out. Deep research being one example, for example. Yeah. Other agentic based systems coming in. Yeah. Not just being on a use case thing. Mm-hm. But also actually being something an end user can interface with and interact with. I think that's coming. And I think actually where we are is in most organizations, we have a lot of people that are quick to follow on. Mm-hm. And to get started. We have some that we are, you know, not dragging into it, but we need to prove the value. Yeah. And I think as we then start adding even more on top being deep research agents or whatever it might be. Mm-hm. We need to actually do the same part of convincing. Mm-hm. We need to again convince people you also need this. Right? Yeah. It's another feature. Yeah. And I think those new features fit the exact same rules as just getting people into a classic. Yeah. AI tool, if you will. So, training, communication, target setting are not one-offs. Mm-hm. This is a continuous journey, I think. And what I'm focused on is how does that journey look like. Yeah. As we build even better tools. Yeah. There's a risk that we might leave a lot of people behind still. Yeah. And that's what I'm looking out for. Yeah. I can echo that. We have had my technical people developing our internal GPT. It's just like, we can do all this thing. I'm like, could we hold back just in terms of not overloading people too much? So, I definitely recognize also we need to dose and get everyone on board before we just hammering. Yeah. I agree. But, okay. Should we, end the session and say, if you find this was interesting and you want to hear more, please feel free to write us an email and we would love to drink a coffee and hear where you are and maybe we could share even further on what we have done today and also what we see. Yeah. Yeah. I think we can share more and have coffee. That would be nice. Yeah. But, thank you so much and thank you for joining. We hope this was valuable to you and we look forward to getting in touch or seeing you out there. Have a good day.