Hi everyone, and welcome to today's AI Assistant Suite Launch event. I am Timo and I'm going to walk you through today's session and its speakers. So we have with us Justina, a technical writer who was also the expert on prompting and will help us understand what prompts are and how you can use them to your benefit. Also today with us is Marcelle, the Managing director of Crownpeak Germany and also the head of DACH Sales. And Marcel will walk us through the pricing of the AI Assistant Suite and also the trial phase. Rene, Team lead for FirstSpirit Core is also here today. And he will shed a light on the AI system suite, looking at it from an engineering and tech point of view. And I'm Timo, I'm the product manager for for spirit. And I will walk you through those slides that are about product and business. With that said, front and center of today's session are the content assistant, the analysis assistant, and the image assistant. On the agenda we have three big sections. The first one is going to be all about the AI system suite, looking at its features and functionality. So we will have a hands on demo, and Rene and I will show you what it can do. Closely related to that is of course a discussion about value. So why does it bring value to your company? What are the use cases it covers? What are the things that are noteworthy? Uh, good to point out, for instance. And I touched on that already prompting. And then last but not least, we will help you get started with the assistant suite. We will talk about the trial phase and about pricing. And of course, if you have any questions, please interrupt me, Justina, Marcel or Renee at any point and we will try to find an answer as soon as we can. Now let's be really hands on and start right away with the demo. So I have here a wonderful E-com website in Content Creator. And as you see, we have an entry here for our AI system suite. And I'm going to click on that link. Which brings me to this very page. And in today's scenario let us assume that we are working in a marketing team, and this marketing team has the task to go through that very page. And as you see me, scroll down here, you see that we have tasks. And those tasks of course can get addressed manually. But in today's demo for the AI assistant suite, we want to show you how I can help you tackle those much more efficiently, easier and faster. So I'm going to scroll down and you'll see there's another one here. More coming up. So this will definitely take me, uh, a lot of time if I were to do this manually. But I'm not going to do that. Instead, let me scroll up here again. What I will do is, uh, open up this text, and you see already that we have, uh, two tasks. Uh, one is about making that paragraph easier to understand, and one is to summarize it. And so as I click into the text, you see this bubble show up here on the right hand side. And we have here, uh, shortcuts to AI powered actions. And in this demo, uh, we will show you what they can do. So here this one asks me to make it a bit easier to understand, so I can definitely improve this and click on the simple language hit perform. Let's see what the result is. So we see streaming in the AI generated text. And for full transparency what we can do is click on this very icon right. This is our comparison view. And then you'll see that certain sections are uh marked in blue. Those are the suggestions. Others have a strikethrough. And then we have the black which is kept. And so as I go back I see this is my result. If I'm not satisfied I can also regenerate. And here is the next one real fast. And I can go back and forth. And I think this one is the one I want to keep. So I had the accept icon. Now this task here asks me to summarize which is what I'm going to do. Clicking on that shortcut here. So that is a wonderful summary I think. Let's have a look together. Yeah it definitely gets the meaning of that very paragraph, Renee. Um, as I go through the tasks, may I ask you to help us understand what is happening in -the background? Okay. -Yeah, sure. Um, what are called shortcuts that are our that are our AI commands. And behind every AI command or AI shortcut is a prompt. And um, at the moment we use open AI. So this prompts are sent to open AI together with the content that is selected. It could be a paragraph, it could be the whole text or just a single sentence. For example, it's sent together with a prompt to open AI to do the task we want to do. And, um, then you see the results. This list, um, makes it really easy for editors to use those AI commands so editors can focus on their expertise. There is no need to be to become an an AI or a prompt expert. And of course, this list is not fixed. It's, um, the way we want to provide a list of really good and engineered, well engineered prompts. That helps for a lot of use cases, but you can extend them. Of course, you can define your own prompts, your own AI shortcuts. You can hide existing ones duplicate. So whatever you want to do. And we saw two types of prompts, one with parameters and one without the summary. Just take the text and create the summaries summarize prompt and the the proof one has one parameter. What do you want to do? We want to use simple language and that options are -also prompts. -Thanks a lot Rene. That was very helpful. And as you explained it, I wanted to give the audience visibility on another use case that we've heard about a lot over the past months and quarters, and that is proofreading. We need it every day. It's tedious and nobody really wants to do it, but it's needed. And so as you heard Renee, bring this explanation from a tech point of view. You saw that we covered this very use case as well. And you see it catches all of those errors in spelling and in grammar of that text. So really fast. Hit save. That's it. And what we can also do here not only optimize text but create text. So here our next task is about, um, a list of the top five benefits of AI in the CMS and Dxp domain. And again, this sounds like a hard task. It's going to take me. It would take me at least half an hour. But I think we can tackle this faster. So here I'm going to be um, extra lazy today because I'm just copy pasting this to my clipboard. And on the right hand side, you see that I hover over this very icon, and as I scroll down, you see that we have the history of today's session in here. And we could go through each and every result, um, and perform certain actions. One thing that I want to do now is send out this very prompt. So again the ask here is to create that very list. And as you see this is happening in real time. And so what I'm going to do now is perform a drag and drop action. Because what we have here let me zoom in a bit, is this very icon. And this one indicates that I can drag and drop this whole element into that form. So let me delete that. I think we can delete the task itself as well. We don't need this very part, but this here is definitely something that is very helpful for us here because as you see, we have formatting right away. This is in bold. We have bullet points and definitely something that we can use in our project. And as I hit save you see it's formatted right away now as I scroll down, you will see that we have more tasks coming up. And this one here is for all the ones working in marketing that are concerned with search engine optimization. Let me close that real quick and then fully focus on this task, because it asks us to generate keywords based on this text here, and for that. Like Rene said, we have something in our list that is the keywords here. So I'm hitting that and this is my list. And I can of course copy paste it, use that um, as I like. Or let me just take out the task itself because we definitely solve that. And save that real quick. Now, what we have here at the bottom is another. And this is the last task for our list. Here is all about expansion. So we have a short paragraph. This one is about AI advances the AI exp. And yes, this is definitely a short one. What we can do with the power of AI is expand on that and expanding on that, we'll make sure that the meaning of that short paragraph is kept intact. So as we spend a second on the result here, we will see that it definitely did the job and the meaning is still intact. If you like, um, for the ones who just joined. Thank you very much. Welcome. Do ask questions. Also in between, uh, the slides. In our demo today, we will try our best to find answers to those. What I'm doing at the moment is I'm going through different tasks on our page here. In a scenario where we are working in a marketing team. Now, we have seen so far the content assistant in action as the AI system suite has two more assistants. Why don't we have a look at the analysis assistant? Now on the right hand side we have the analysis system. So I'm opening that up. And as you see here we have a few results. That is because in my session today I already analyzed this very page. Let me do that again because here we are looking for all the things that we can improve in terms of tone of voice. Tone of voice is something that quite a few of our customers have. Basically, it is the, uh, convention. Looking at your language, uh, the way that the documents you have, the website, uh, is phrased, the wording is, um, and there might be some that need to be applied in a scalable manner. And for the case today, I want to show you how this works with the power of AI. So we spoke about prompts a bit already. And I want to spend a few seconds on that. On the right hand side you see this icon here. So I can click Edit Prompt. This tells you a bit about what is happening in the background. René already elaborated a bit on that. Here our tone of voice has those four rules. So in the results that we are going to go through in a second, you will see how those apply. What you will also see is how a fifth rule, which we are going to add in a second, uh, is applied. So this is what is happening in the background. And on the right hand side. You see this list here. And as I hover you might also see that certain elements are highlighted while others are not. What is that about? That is about the actual paragraph that this is focusing on. So as I as I go through the elements here on the left hand side, we for instance have this one power productivity. So that is that one I can click on this very icon. And again we have a comparison view. This is definitely something that is tremendously important to uh, our idea of AI. We want to make the results of AI transparent to you so you can, as a user in the driver's seat, make that very decision if you want to apply that change or not. So you might remember that, um, we had a rule about counting. and this one is applied here, as you see. So that works. Now, as I already said, why don't we, in real time, change our prompt and add to our tone of voice another rule. And in this case, I think we want to speak of Crown Peak instead of company. Um, I am definitely not an expert, and the rule I'm going to put down here can be phrased so much better. Certain. But let's see how this would work in our demo today. So we could say. Whenever you spot the term company. Replace it with the term. Okay. So um, there are definitely, again, better ways of doing this. But as a first try, why don't we see how this works out so we can save this as our fifth new rule and then hit the analyze button once more. And then see what the analysis system brings. Again, it's going through the whole textual element, the whole textual content. Apologies of the page. And for each and every item it highlights how tone of voice applies. And as we have our item here, let's open that up. And you see we have a hit. Wonderful. And you see, our company is being replaced by Crownpeak. Not even our Crownpeak because grammatically correct is simply Crown Peak. And this of course, is how we want it to be. So we can apply the changes. And then the page is being reloaded and you will see it down here. There you go. Let us now have a look at the image assistant. At the top of our page we have our wonderful logo. But I think we can find an even better image for that. So as I open up this and pull from my desktop, drag and drop this very image, what is happening in the background? Renee, maybe walk us through that. Yeah. Here. While uploading the image, the image is analyzed. So in the past, we use AWS Rekognition to do that stuff to get some more information about the image. Now we extend that and the image assistant by analyzing that image via OpenAI. And um, this prompt again is something we provide so you don't have to care about. But if you are interested, you can change that prompt. Of course, at the moment it's optimized to create an alternative text for that image that is uploaded, and that's happened in the background. While uploading the image analyzed, um, and analyzed with the prompt. And here what you see is on top of that because the image, um, when you upload that is not in a context, context less sets in a folder here. This image will be set in the context of this blog post. So we take the, uh, image description together with the topic of that blog post, and let the AI rephrase that alt text. And he will see the result. Thank you very much, Renee. And we have even more to show you today, because why don't we talk a bit about metadata and summarization of content that we have here on the page? And again, maybe, Renee, tell us what the magic behind that is, what we just -saw here. -Um, yeah. Here is a combination of, um, quantum assistant. Uh, you see all those shortcuts on the right hand side and the assistance API that we use here. So we want to be more efficient. We want to let the AI, uh, start, and then we can improve so the editor can improve. And what we see here is we let the AI create the teaser out of that headline. And when we switch to the metadata tab, for example. So three input components are also generated by AI based on the content in the content tab. Um, and this this example, those examples are generated by AI as long as I'm not changing them. So, um, you see that in this little info text, in this little hint below every component, the autofill is active. And at the moment when I focus that component, I put my cursor into that component. Um, Im as an editor, would like to improve or change the text, and then I don't want to let the AI override my improvement. And, um, when I want to re-activate that autofill, I could just clear that field. And then when you click on another. Yeah. A moment later the Re recreates the text. And all of that is based on the assistant API. I can talk about that a little bit later and on prompts that are really project specific here, because they are based on that specific form, that specific -content. -Thank you very much, Rene. Very helpful. Um, as you explained how this works, I saw that we had some questions coming in our Q&A section. I'm going to have a look real quick. So the first one is from Francesco. Thank you very much. The question is are the prompts language sensitive? As in if I were to call a prompt while working in, for example, German with the response Be in German. Rene. Or maybe even Justyna. -Who wants to take that one? -Sure, I can take this one, Renee, if I'm seeing anything or missing anything. Um. just interrupt me. Yeah. So the answer is that yes, they are language sensitive in as much as they will take. They are based on the language of the input text that you're giving. So if your input text is in German, um, the output, the generated output will be in German or any other language should open. I um, actually supports at this -point. -Thank you very much. Perfect. This concludes our demo for today. Let us head back to our presentation. So what is it that we saw? We saw the continents and we saw how easy it is to not only create but also to optimize text. We saw analyze assistant. We saw how it reviews a whole page, how fixes are applied and how easy that works as well. We saw the image assistant, which is here to help you enhance your images with all texts that considers the whole page. And under the hood we saw the assistant API. And really, why don't you help us -understand what this is about? -Yeah, we saw some examples, some use cases. Um, what we saw already the three assistance we provide is that is the idea of how we provide value of AI to be more efficient at so many places by our assistants. But if you want to go further, then you could use the assistant API. We provide the AI to all project developers to help them to bring that the whole stuff to the next level. Uh, to help to extend your project with further AI functionalities, for example, to reduce menu work, to do more automation, to be more efficient. Um, and yeah, we saw a couple of use cases already today that combines our different assistants and sometimes bring that. If it's about your needs of your project then you can extend that and that's possible using the assistant API. One last thing for the developers, it should be as easy as possible so they can just use the I don't have to care about technical details like connection, and they can even use the prompts of the system prompts defined by prompt experts. So as a developer can use the existing prompts and just use them like the way editors use them, but in your -development. -Wonderful. Thank you Rene. Now, a question that we received quite often in the past when it came to I was, what do you support? And we can say that we support SAS scenarios. We support on premise scenarios. And we even support those where you bring your own API key, that is ChatGPT compatible at the moment, or you don't bring your own AI API key because we provide one good news here. We support all of that. Now let us zoom out a bit, because what we see as the three key dimensions of our AI system suite are the efficiency gains that you can get through the AI system suite? The configurability. And we saw a bit of that already and we will see more to see. We'll talk about that topic in our presentation today and the transparency aspect. We saw that when we had a look at the diff view what is AI generated, what is not. And we have more on that topic coming up. Let us focus now on the efficiency aspect. So out there in the market, there are different experts, analysts that tell us, and this is Deloitte, that you can save 11.4 hours per week per content marketer. If you're an early adopter of generative AI technology in marketing operations. Also in that scenario, there is a 12% return on investment that is reported again on generative AI investments from those early adopters. If you're interested in the details. Have a look at that wonderful study that they produced. You find the URL down here. The load, of course, is not the only one. IDC in the same manner says that they estimate 40% marketing productivity gains until the year 2029. Boston Consulting is phrasing this a bit more broadly, and saying between 10 and 20% of productivity gains can be achieved with generative AI tools and processes being deployed. As I see it, McKinsey is actually making a great point by stating that yes, those AI based productivity productivity gains are achievable, but they do require one thing and that is investments to support your workers. I think one could spend hours discussing those studies, and numbers are high level though what we learned in the past quarters. And thank you very much again for all those who participated in our beta phases. That was tremendously helpful. What we learned is that things are good when it comes to efficiency. There is a certain level without AI, but things are even better as soon as you start implementing AI. Now zooming out again and spending a few minutes on the configurable dimension of the AI system suite, I'd like to hand over to -Justyna. -Thanks, Timo. Yeah, so we've already heard quite a bit about how the AI Swede was essentially made to suit your needs. Which means not only can you use the fully functioning and already pre-configured prompts that are provided by Crown Peak, but you can also create your own prompts, and that means that you can freely adjust the content assistant and the analyze assistant to tackle any task that might be project specific or specific to whatever it is that you're looking for, and complete this task quickly and efficiently. So any large language model essentially uses prompts to generate output, and a prompt is nothing more, nothing less than, simply put, an instruction that is given by the user that guides the model on the content that it should generate. So in other words, in the prompt, we just define the task that we want the model to perform. However, that also means that the output and its quality is going to be dependent on the prompt and that quality. So I want to talk a little bit about what are some things that we want to be looking out for when we are creating prompts. And one is that you want to be as specific and concise as you can and avoid vagueness at any cost. And to show this today we brought an example. And you can see an example of a very vague prompt on the right. It reads right a product description of our new AI powered feature in our content management system. Here you can see it's very general. We're not giving the model a whole lot to work with. Um, and that might lead to output that might not be desirable or not exactly what we're expecting. So rather than doing something like that, we want to be as specific and concise as we can. So we want to write something along the lines of write a two paragraph product description for a new feature for a website. The feature is AI powered and aids customers with text related tasks such as shortening, summarizing, and reviewing text in our content management system. Here you can see that I'm giving the model context. I'm telling it where I want to use the text that is going to generate, so that's going to help it understand what kind of text I'm looking for. Plus, and providing information about the feature that it's supposed to write, about which the model is then going to use in its output. Well, you might have already spotted on the previous slide is that I also specified the output on the previous slide. I said I want a two paragraph text, and that is super important for any prompt that you're writing. You want to tell the model exactly what you want the output to be and what its format format should be. So again, there is an example to make this really clear. On the right we have an example that doesn't give a lot of information about the format or the output in general that we're expecting. So it just says list advantages of using our new AI part feature for text related tasks in our content management system here, we're not really specifying what lists it is that we're looking for, so it's going to be up to the model to decide that. And that is not really what we want most of the time. So rather than doing something like that, we want to be as specific as we can. So on the left you can see a prompt that is very clear in what it is expecting the model to to generate it says list ten advantages of using our new AI powered feature for text related tasks in our content management system, I'll put a bulleted list. Each bullet point should be no longer than two sentences. Do nothing. Do not output anything else. So you can see we're not only specifying the number of points that we're expecting, but also what kind of list it is, how long each bullet point is. And then lastly, we're telling the model to not output anything else, which generally is good practice because the models tend to be quite chatty and juicy. Now, since we had a question from our audience today about language, I see that those here are in English, which you recommend writing them in English or in the language that the text is in. So I would generally recommend writing prompts in English just because for now, this is the most used language and you will most likely get the best output doing so in English. There is a possibility to do this in other languages, but the output might not be as desirable. Thank you. Right, so once you've created your prompt and you have gotten some output, then how do you go about using it? And I believe that there is some best practices we should keep in mind when we look at any kind of AI generated output, really. So firstly, you want to make sure that you verify any critical information that you yourself have not directly provided in the prompt. That is because the output may include some statements that to you might look like they're facts, but they might actually be completely made up. These are usually referred to as hallucinations. And then secondly you want to look out for biases. These biases might come from training data and could then be amplified by the model which potentially could lead to skewed or unfair output. And then lastly, just understand the limitations of what the models can do and also the limitations of the training data, because those limitations may lead to the model not always providing contextually appropriate responses. Thank you very much, Christina. I'm going to give the audience here a few seconds if you have questions. I would like to, um, connect the dots once more. So as you saw in our demo today, we have those shortcuts to certain actions. And behind those are, well, prompt, engineered, uh, prompt that we know bring consistency and high quality. If you, though, decide to cover your use cases in your specific project settings. Go for it, copy paste ours and then alter them so they make sense in your scenario. As we've spoken about the efficiency dimension and the dimension of configurability, let us spend some time on the transparency aspect with our AI system suite. We want to address certain pain points that we know you have. We know this from interviews, from polls. Thank you again for that input. Not only with functionality that we provide, but also with what we call a starter pack. On the right hand side, you see a few screenshots of documents that are included in that starter pack. You see that we have one that is on prompting and in that wonderful document there are even more ships than just, you know, spoke about today. You see a solution brief on the AI system suite that nicely summarizes the functionality. And you even see a document there in the background that helps you get started. One topic that was highly discussed was legal. And so with the AI act in short or, uh, the European Union Artificial Intelligence Act, full name, um, being in force already. And that is from the 1st of August this year. Um, those questions come up, uh, even more so. So what is the European Union Artificial Intelligence Act all about? Why does it matter and how can you prepare for it? The act applies to all AI systems that impact people in the European Union and to all organizations, including those the employers, distributors, developers and importers. And it aims to do one thing, and that is to ensure that AI systems in Europe respect ethical principles, safety and fundamental rights. This might sound scary, but the good news is we got you because we have a material on that that helps you prepare for the AI act. One point to consider here, for instance, is the inventory of AI systems. So it is advised to have a list of AI systems that your company has or plans to have, and then to determine looking at each and every item if it falls under the scope of the AI act. It is also recommended to classify those eye systems. So you categorize them and you look at the risk levels and the company's role. It is also advised and this is not an exhaustive list. Right. These are just three points we want to highlight here. To have an AI governance framework. This one would include an overview of your legal risks, how you implement safeguards, and what the mitigation measures are that you might have or that you might want to have in place. And we have more, much more information on that very topic in our startup hack. That startup hack will get you started with our assistance suite, help you get the most out of the functionality. Contains documents and videos, for that matter. And those videos are on functionality in such detail that we didn't cover even today. Plus, it comes with a trial phase. Now, the trial phase I mentioned at the very beginning of the session, the trial phase is also something that is important in the context of pricing, and that now is something that Marcel will talk more about. -Thank you very much, Timo. -Thank you very much, team. Yeah. When thinking about pricing next slide please. Um, was very important for us to make that as easy as possible as, uh, predictable as possible and as, um, customer centric as possible. So from a, from a from, from a metric point of view. Important here is the number of users. Number of named users is, um, one component of of the pricing, the major component. So that's the that's that's that's the case. Um, you are able that's another point. Um, to bring your own API key. Yeah. We have seen scenarios where companies have well, they've, they've dealt with uh, with AI for quite some time, and they already have this in place, but you can have ours. Yeah. That's, that's one important component to and we are happy to discuss our offering. We're happy to show you the value of our offering. Um, so we are here to help. Yeah. The team is here to help Timo. Uh, and the team everybody wants you to help. Next slide, please. We know. Well, I this may be brand new for for some people. It's been it's been for a while. And you might think well can I try it. Yes you can. Yeah. We've spoken about the trial phase. Well four weeks. Yeah. You get four weeks and you will get our support. We will guide you through. We take you by the hand. We will have a definition of clear goals together with you. Uh, together with your teams. We will have check ins, regular check ins. So we will get started. We will have, um, several meetings. Um, and we want you to feel comfortable. Yeah. And, well, as every trial phase has a beginning, it also has an end. Yeah. And then we will obviously, uh, speak about numbers and, uh, value. Apparently it's free of charge. Yeah, that's the best. That's the best price, I assume. Um, so last night for my from my side, please. It's well reached out to me. Um, just in case I don't respond. Don't don't don't worry. It's at someone else. Um, we will take care of you, and, uh. And, uh. Yeah, let's let's get -started. Yeah. Thank you. -Thanks a lot. So very exciting. Now, today we saw the AI Assistant suite. We saw the streamlined content creation with content assist. And we saw the rigorous content review with analysis. And and we saw the powerful image optimization with image assistant. And as Marcel said, we are here to help. We have to answer your questions. And we are excited to see what you're going to do with the AI Assistant Suite.