Name: From AI hype to business-ready PowerPoints
Uploaded: 2026-03-24T16:06:07.034Z
Duration: 46 min 32 s
Description: From AI hype to business-ready PowerPoints

Transcript for "From AI hype to business-ready PowerPoints": Hello everyone and welcome back to Tenblify Talks. With me, your host as always, Jonathan Acros. I am part of the customer success team, customer excellence team, advocate of Templify, enablement of Templify, and essentially marketing. I don't know. I do a lot of stuff these days. For the folks of you that are new to Templify, welcome. For those of you that are returning to Templify Talks, you kinda know what this is about now, so welcome back. As always, we start with the house rules. We use the web platform Goldcast, and it's in there today. We're gonna interact with you live. So let's make sure we test it to start with. I'm, as I mentioned, coming to you live from Copenhagen. Hello from Copenhagen folks. Tell me where you're coming from. Throughout this webinar, feel free to ask questions in the chat. My colleague, Barbara, is behind the scenes. She can answer some of the questions along the way, but, also, we'll take them and answer them live with my guest. Before I get on to the guest, last time on Templify Talks, we showed you how you could connect with Claude, Gemini, Copilot to optimize your enterprise workflows, enterprise workflows. We can't even cut that. This is live. And this time today, we're going back and focusing on Templify document agents and specifically proven that prompt to PowerPoint can be achieved at enterprise level and how to sort the wheat from the chaff. To help me sort through this noise is perhaps Templify's noisiest son, the man leading our commercial efforts on the document agents, and a fellow Essexite, Ed. Hello, Ed. Hello, John. Hang on. John and Edward. It's almost like Jedward, where I have the hair on top and you have hair on the bottom. Ed, if you still got all your hair when you're my age, mate, I'll be very surprised. So well, that's not to soft charge. Before we get into the regular programming, Ed, what do you do? Well, as you said, I lead the commercial efforts here at Tempify on our DocAgent product. So I work very closely with our product team. I work closely with you and our commercial team as well. And also sit and just become a bit of a mad scientist, I would say, with our agency. How far can we really push it? Fair play. Awesome. Well, I'm gonna pause for a second because we have some very important Templify news that I'm going to read. Right. Firstly, if you want more of Ed and I, you can join us in person for hands on events in three locations. So this time, we're going on tour. Now we're going deeper than AI helps you write or helps you build basic documents. So in our document agent workshops in London, Stockholm, and Copenhagen, you can see more of us. You can get your hands on the tools, expect real demos, practical takeaways, and yeah, you can get your hands dirty. So that's very exciting. You'll see Ed and I there too. Awesome. And then huge news. And I'm going to read this verbatim because this is a press release and we want to make sure that the message is right. Semblify just hit a big milestone. We've been awarded our first US patent, which is really exciting. This patent protects some of the core technology behind our AI powered document generation platform. Now if you look at most AI tools out there today, they're great at generating text or images, but they often stop short of creating something you can use in a business setting, like a polished compliant on brand presentation. That's where things tend to fall apart. You end up with messy slides or content that doesn't meet your company's standards. What Templafy does differently is brings everything together. We combine generative AI with rules based automation and enterprise data into one unified platform. The result is a single document engine that actually understands your company's requirements. So instead of just generating content, you can prompt AI to create fully customer ready documents aligned with your brand, your structure, and your compliance rules in just minutes. Very, very cool to be know, a lot of companies talk about innovation. We're actually innovating and painting in it. Patenting it? Patenting. Patenting. Patent. We're putting a stamp on it so you can't use it. And finally, for those of you that want to get your hands on document agents right now, we have just launched our prompt to PowerPoint generator on our website. Now this AI tool turns simple ideas into a fully structured presentation, editable. You don't need any design skills, luckily for me. Unlike most tools, it gives you a real PPTX file that then you can download and customize however you want. It's fast, browser set, and free. Well, I mean, you have to put your email in, but that's as free as I'd like to think. So, yeah, I would suggest and Ed, when we talk about today, you're gonna mention some of the liberate functionality. But essentially, look in the chat now. You can see the links. Don't do it yet. Do it afterwards. You're in class now. Do it later. But so, yeah, you can try the Templify Liberate, the prompt generator, the PowerPoint generator online straight after this. So excellent. So, Ed, what can the viewers expect from today? Well, it depends which viewers are left, John. Because if everyone's going to Templify Liberate, they've probably already bought Templify using it. So no. What people can expect today is two things. And I think one of them is very broadly applicable to all of us, the other is a little bit more Templefire centric. The sort of world of AI, as it were, has been evolving rapidly. And what I think a lot of businesses have seen is that relying solely on one's LLM is quite tricky. You need it to be, you know, kinda get it right every time, but without giving it the right things, it's pretty unpredictable. So from a broad perspective, we're gonna talk about how to orchestrate any sort of AI tool and some things you should think about when you're looking into your own AI tooling. The second thing we're gonna talk about is then how we apply that to our document agents and how we sort of orchestrate different components to ultimately get us to a place where we can help businesses create their documents every time. Fantastic. So you talked about some of the ingredients that you need in order to sort of start the process. And you've got some slides to show us to I do. Walk you through So I'll show some slides just to set the baseline of what we're looking at here, and then we'll jump into some demo action. Fantastic. Super. So with that, to the lovely audience here, I'm not gonna bore you with too many slides, as I wanna show you how we create slides, not how we show them. But as you're all aware, we've been on a rapid journey over the last few years with AI. Now maybe a little bit Templify centric, we we came about in 2014. In 2022, we saw ChatGPT launch. We then saw Copilot a year later. Over recent years, we've seen even more LLMs, you know, come in and and change our working life. And now as I imagine many of you are sat here watching this, we're all looking for places where we can drive real ROI to to prove that the AI value is really there and it's not just something that an influencer said on a podcast once upon a time. But this is a challenging space to navigate. And I'm sure a lot of you are sat there hearing all sorts of buzzwords all the time. That's just creating more confusion and almost is quite scary because things are moving so rapidly. Where do I even begin with this? Now I'm not gonna cover all these buzzwords today, but we are gonna talk about the sort of AI agents portion of this. But what we're gonna be focusing on, as I mentioned, is very much the space of orchestration and how we're bringing structure, content, context to our AI agents to make them as good as possible and to give them the ability to create the best outcome possible too. But let's have a little look at what one's current state could potentially look like. Today, an employee goes into their AI tool and potentially gets an inconsistent output. Typically, the reason why is that they may have all these great systems in the background, but they're not necessarily orchestrated. We're not telling AI how should it work with things, how should it actually interact with these pieces, and what is it doing that for? And ultimately, that leads to a few things. The unpredictable output that I mentioned, a lot of rework that takes away that efficiency gain that you're looking for, and then also trust. People go to an AI tool. If it doesn't work the first time, it's unlikely they're gonna go back and they might revert to old habits. If people aren't using it, it's very difficult to then see the ROI on the other end. So when it comes to actually AI with orchestration, if we look towards the future and hopefully, some of the takeaways here help you get to this place, it's not just about plugging in systems into an LLM or an AI tool that you've got. It's about actually giving it some purpose, not just how a system should be used by an AI agent or an AI tool, but also then in between its usage, what should it be looking for if it's a CRM as an example? How should it be using that data in a CRM? And how can it then bring that to the output on the other end, whether that's a document or any other kind of output that you're looking to do? What do you get on the back of doing something like this and exploring this option? Well, predictable output, genuine efficiency, and adoption. Because the AI tooling will be doing what you need need to do. It'll be predictable so employees know that they're gonna get the right thing out of it, and that will drive efficiency. I think a principle to maybe take away is when you are looking at using any AI tool, try and start with a specific task. Now that could be documents like with Templifier. It could be anything. And make sure you've got the right systems integrated for that process or that task you're looking to do, because ultimately less variability will lead to more predictability, which is what we want. And it's probably the big tagline of AI. It's all unpredictable. So just going into sort of how does that look for Templify then. So this is my triangle. It's my I'm very very precious triangle, as John will tell you as well. And we have an employee that will go into the Templify platform and start working with our agents. And we're giving those agents three things based on how we've worked with the customer or business partnering with us. First is content. We need to know in the case of a prompted presentation tool, what kind of content is actually gonna be used here. And that's not just here's a deck. It's also the templates themselves and specific layout slides. It's placeholders. So where is information going on this slide itself? And what about the images, the icons, the other digital assets? We wanna give that all to the agent and give you the ability to centrally manage that so the agent's doing the right thing and staying on brand. The second is knowledge. We want to be able to give our agents context and information about what it's building so we can elevate the actual information that's being presented. We give it some company context, or we can give it project specific information. And we can also give it best practice examples. So we can go in and read these documents, look at the information, how it's presented, and reflect that in the document that it's generating. Finally, probably one of the trickier areas for many people is data. A lot of the data that we work with is very sensitive. We don't want it going anywhere, we certainly don't want AI taking maybe pricing information and adding a zero on the end. That can cause serious problems. So the way we're working with data is with structured sources and structured integration frameworks that we've had for years now. We are giving the agent instructions on how to use that data. So we're actually telling it very specifically, this bit of data, you must not change and you must place it here in a in a specific slide or in a specific placeholder. And this leads to that deterministic effort. A lot of folks are talking about deterministic. What does that mean? It means knowing where the information is placed will be in the right place on that document without any change. It won't have any hallucination risk. People are familiar with that term. That's how we're looking at it. And why does all of this actually matter at the end of the day? Well, for your document, it means it's gonna be high quality, means it's gonna be on brand, it's gonna be built efficiently, and it's gonna limit your risk. For you though, as an individual, no matter where you sit in a company, you're probably building a document. There are a variety of benefits to you. If you're in sales, it means you can spend more time on speaking with your customers. If you're in IT, it's governed and it's connected to Genetic workflows. It's not part of this sort of shadow AI adoption that we're seeing. For brand, it means everything is gonna be really on brand by default. There's And so many other benefits for different departments. So it really does benefit everybody to look at, at least in Tempify's context this way, but also for others as well. If you're using ChatGPT or Claude, how can you give it what it needs to to get the right output out? Finally, before I jump into demo, as I said, I wanna show you how we build, not show you how we show slides. Mhmm. The way a lot of you may be looking at this and going, I've got loads of systems. I work for a company of a 100,000 How on earth do I even get started? It's okay. You don't need everything in place to get started. And it kinda gets driven by context, ultimately. Now if we have sort of limited context, but we know we've got the content available, we can get everybody started. The agent can start building on brand presentations for you, and it's a great place to just get the right template in front of people depending on what they're doing or the correct slides as well for that matter. If we know the knowledge though, we can go higher up. We can start building in more information, more structure into the document. We can enrich what this presentation is, so less work needs to be done. And finally, if we have that data component as well and we understand the full context of a document, that's where we can get to the maximum capacity of what we can complete. So ultimately, either you can ship that document or you can just give it a quick review and then send it off. And this is ultimately where it stands. The more context you know about your document, the more we can help you complete it with our agent. And that's sort of this flow here that I'm showing you, this is what I'm gonna be demoing today live. I love it. So while you've got that slide open, I think it's important to just because I saw that Barbara's actually put the Liberate link in the chat already, so people are probably already looking at it. I would imagine. My phone's going off the hook right now. With with your with your and he does love that triangle, by the way. With that triangle, maybe maybe just put some frame around how Liberate or the the online users that that triangle. Why not instead of me telling you, I just show you? Simple. Let's do that. So when you if you haven't already, when you go into the Liberate experience, you're gonna be met with with this lovely UI. For those of you that don't have Templify today, it's accessible via Chrome. For those of you who do have Templify, you could also access this via PowerPoint itself if you choose to partner with us. And, ultimately, what we're looking at here is the interface that anyone would be met with, whether they were building a complex document or a simple one. But in the case of Liberate, we've only given it some content to work with. That's it. It doesn't know any knowledge because we want it to be used by everybody. And we don't know what everyone wants to come in and do sort of in the 8,000,000,000 people in the planet. That content we've given it is just a template that we've configured. Yeah. Exactly. But there's a lot to come with Liberate, so stay tuned for that in the future. But in the case that I just need to put something together, such as please, I don't know, build me a town hall presentation, I can go ahead, put my prompt in. I can upload something if I choose to, and I can also pick how thorough do I want the actual reasoning to be. So that nice little reasoning tag that comes up in ChatGPT or Chord, we have the same thing in Tempify. If I go ahead and say enter, the agent will have a think. And again, it's just looking at the content. So it's gonna ask some questions to try and contextualize some things. But when I'm getting started, I just wanna get started. So I'm gonna skip and just move myself straight into the outline. What the agent is doing now is it's looking at the slides it's got, excuse me, and it's then gonna have a look and see which slides are the best for town hall presentation. Now as this process is flowing, I just wanna see in the chat if we've got any questions or if I'm John, you have just posting in that we keep using the term liberate, and that's our internal name for it. It's actually the the PowerPoint generator. So, I mean, I actually prefer the term liberate. I'm surprised when I didn't see it the website. No to marketing. But, yeah. We are liberating Templify. We are making it accessible for everyone to use and That's why they called it that. Exactly. So it's really sort of undoing it for everybody to go and use. There's a question from Martin. He said, during this step, is it looking at the slide I have in my library? For example, is it looking at the base template that they have? Well, it actually depends on what you give the agent upfront. So we can either give the agent rich slides that have got some degree of purpose, and we'll get into a bit more of that when it's a custom agent, a bit of a teaser there, or it can just be your base template. So the the one that Martin's talked like, I think because we're talking about the one on the website Yes. This is using a set template just to show the what we can achieve. Yes. If you wanna use your own, as default by default, we do use Yeah. Yeah. A company template that is configured with your font styles, colors, etcetera. Precisely. Yep. And that's what you're gonna show us today, how that's created. Yeah. A bit later on. Yeah. Here we go. Spoilers. Nonetheless, in this experience, Tenem, where we've got access to just that template that we've configured, I get an outline, which we'll be revisiting this flow quite a lot, and we're getting an idea of what the actual presentation is gonna look like. This is what I would call a human in the loop step. It gives me the ability to actually see what am I gonna create and what is actually going to be in these slides. If I click into one like the executive overview, I get some information here. Now we didn't give this agent any contact on the chat, and it doesn't know anything either. So the information itself is quite broad. But nonetheless, I can still change things around if I want to based on the layouts we've given our agent in this particular experience, and I can also change the information if I choose to. I can also go ahead and delete a slide. So in the case of this town hall, I'm not that interested in having that, and I'm not overly interested in having this slide either. So I have that ability to actually do something before it reaches the agent compiling it as well. But even that can be helpful. Right? Some of the things that take the most time when you're building a presentation is actually figuring out the outline itself. Like, in what structure should I have to present this? Instead of having this sort of blank page paradox problem where you don't know even where to start, you can at least get started this way and know that it's the slides that have been approved by your brand team or whoever else approves the right slides. Nonetheless, once I'm happy, I can go ahead and say create, and the agent will create this presentation. While this is sort of going ahead and creating some slides for me, I'm gonna go ahead and show you this content element, how we're actually orchestrating this. This is the admin center. This is mission control for those of you that use Templify in your company. And for those who don't, this is where things are ultimately controlled and access is limited. If I go into the agent section though, which you can see here on the left tab, and I go into slide layouts, this is how we're orchestrating the template component and giving it to the agent. You can upload multiple layouts in here if you choose to, and you can have more than one layout associated with the same agent or specific layouts for different agents as well. But as you can see, whenever when I uploaded this originally, AI went through, looked at the slide, and gave each slide a sort of purpose and a bit of a use case behind it. Strategic overview slide. Use this slide to present the strategic overview. Mhmm. So this is layer one of the temp the template itself from the content. It's we're inferring to the agent, how should you use this? When I am building something, where would a strategic overview be relevant? So it knows I'm gonna use that one because I need to give an overview of the topic in question. Yeah. But we take it a layer deeper where we can then go in on a placeholder level and then check off this is going to be the topic one title. This is the description of topic one as well. And I have different options here. I can make things text or a number. I can go in and say if it should be a long paragraph, a short paragraph. What this additionally does for our agent is that it lets it use the right slide, but then it also understands where should information be placed as well. Yeah. And this means you can guide your agent when you know very specifically where things need to be placed or very broad if it's a slide that's used by multiple people for multiple purposes. And it doesn't end at just slides. When it comes to content, we mentioned images and icons and things like that too. If I jump into an image, it can give a name. So it's given it a broad name, sidebar image here, because I don't need anything specific placed there. I just wanna give some visual identity to it. But I can say what kind of image this should be. Should it be an icon, a background photo, a logo, or an image? In the case you're making slides with lots of logos on, you could have that, those logos in your Templify library, your photo repository. They can pull them in which they're preapproved. So just while you're looking at showing those, we got a really nice question from Milan, and I think you've answered it. But he said or she said, what if I have different sets of templates and I want to use this tool with relevant slides, images, charts, etcetera based on a user prompt? So essentially, what you're showing there is the back end of deliberate function. Yeah. So in a customer mindset, they would need to do exactly what you're doing there but with their template. Yes. So they can just as you just showed and you're gonna show a bit later on, essentially, you know, this you have a slide, you have a presentation, And just maybe just show your screen again. Yeah. And essentially, what you're saying is this section is a text box. Use this amount of characters. Yeah. This is a title. Use and then the agent will then populate only those particular areas with that content. Yep. Exactly. And never anything else. Never anything. So it's really good control. I hope that answers your question, Milan. I had a really interesting question from Claire, and this is actually I'll let Ed answer this. Claire. Hello, Claire. We know Claire. Claire's a friend of Tempify Talks and a friend of Tempify. How soon could AI agents be available in our tenant? Well, right now. It's it's available as we speak. So, yeah, we can we can look into it. Good stuff. Okay. As you were. Thank you for the questions. Keep them coming, folks. No worries. So when you've gone through our experience that we're that we're offering folks on our website, you can then go ahead and download your document. Like John mentioned, it's all in PowerPoint as well, so you can go ahead and open it and start working if you choose to. This case, I'm gonna open it up, and then we can see what it looks like. And as you can see, just followed the right slides and used the slides we orchestrated in the background together. It's given me a nice agenda, bit of an executive overview. As you can see, broadly speaking, it's followed the structure that we had defined in that, and it's also followed the branding standards that we put in this particular tenant. Now this is good. It's a good starting point. It's on brand. I can start working on it in PowerPoint. It's all very nice and good. But there is two questions there, which is what about the information and what about when I've got data? What if I'm dealing with a more complex kind of document? How do I then start working with that? And that's where these two other layers of orchestration come into play in relation to knowledge and data, and we're gonna split them up. So I'm gonna start with knowledge, and I'm gonna go on to data as well. Just gonna close this down, and I'm gonna jump back into the admin portal. So I mentioned this concept a little bit earlier of custom agents. Mhmm. Now for Templify, you've got an agent that can be universally applicable for, which is sort of what's in the current experience we've got on our website for folks to go and try. The second layer of that is then saying, well, what if I have a particular purpose in mind? I understand this context from that gliding scale. Well, that's where you can create a custom agent, then go ahead and service that purpose for you. We've got two down here, and I'm gonna start with the key insights slide agent. Now if you when you go in and you want to put together an agent, it is as simple as just hitting create agent and you start filling out these different boxes. We've given it a name. We've given it a bit of a user facing description. We've given it a very brief outline because it's only printing one slide. We can give it some questions if we need to. But then importantly, when we talk about knowledge, I have associated some knowledge to it in the form of a call notes folder. So this folder has got different call notes in it. The agent can go in and have a look in that folder and infer knowledge that will then end up in my slide on the other end. So if you have a call notes folder yourself or you've got recordings that are coming in from Teams or Zoom that you store somewhere, you can give that to an agent when one of your colleagues needs to use information and present it back in a PowerPoint format. And when you've sort of set this up and you've put it together and you've given a bit more orchestration in there, so for example, in the outline, you can go ahead and actually tell the agent this needs to be the key inside slide. So we're leaning back on that template orchestration we saw earlier. I can then go ahead and start actually using this. Or if I want to, as I'm gonna show you now, I can give it a test. So the knowledge is integrated. I've given it a SharePoint folder to look at with my call notes. I've given it the right slide from our template orchestration. And I've given it a bit of a name. Once I hit test, we will get into the same view that you saw earlier. And you'll see here this little bar. This is where I can select different agents that I want to use, different custom agents I may have. And then I can go ahead and because it already knows it needs to be a key insight slide in there, we don't have to give as much content. But I'm just gonna say please build one slide for key insights. Once it goes ahead, it'll have a think about the instructions and the knowledge I've given it. It's asking some questions as well, and I can choose to answer them. But because I know the knowledge is there, I can skip it and it can go ahead and build. I always like when you're going through this phase and people always say, like, why is there questions? Why is there questions? But if you were given any task to a colleague, for example, if you wanted to build a PowerPoint together, they would ask questions. Yeah. And the more con more sort of the better you answer those questions, the better they can, you know, complete the task at hand. A 100%. It's like hiring a new employee into your business. If you give that employee brilliant onboarding material, lots of information, lots of context around their job role Yeah. They'll do a better job. It's similar to the agent, and this is where this orchestration I know I sound like a broken record using that word, is really important because we're orchestrating the agent on what it should do, how it should use the information, how it should use the different elements it's got. We know it's got the insights templates, so it knows it needs to use that. We've instructed on how to do that. It's got the knowledge. We've added instruction around that as well of how to use that knowledge. So with all these elements, we can get to a more predictable output rather than just sort of hoping for the best. Makes sense. Once we hit review, it's done as I asked. It's given one slide. But I think if you remember when I clicked into this on the experience version where it didn't have any knowledge, the information was quite broad. If I click into this, we've actually got more information. So it's given me a subtitle of the mark the date of the call I had, which was today. It's given me some information on NVIDIA because that was the case the case study in the notes folder. And it's given some different pieces of information here that are context specific based on the knowledge I shared. I even have an idea here of what's gonna go into this slide before I've even completed it, and I can see it's reflective of the right information too. So I'll go ahead and say create, and it'll create this one slide. So while while you're doing that, I've got some questions. One of them, quick apology, Michael. I wasn't being sassy. He asked for pricing, and I said, how long is a piece of string? Obviously, what Ed is showing, there's the the basic functionality. There's if you wanna connect it to data. It depends on how many users you have. So get in touch with us, and and we can discuss that with you. Ed will be on hand. Victor, this is something that I'll ask you directly, Ed. Thank you for the demo. Very polite. Thank you. Not sure if you'll cover this, but how does the prompt mix master slides, the main template, with isolated slides that are not necessarily built the same? For example, build presentation x with knowledge from master slides with a CVA and CVB isolated slides or elements. When we say I mean, a question for a question almost. When we talk about an isolated slide, is that something that's got pre populated information? I think if I understand it correctly, and this comes up quite a bit, is they want to build a certain document and then there'll be certain static slides. Maybe as Zeed mentioned here, CV slides. It could be an about us slide. The content that shouldn't change but then gets populated into the document during the So the way we can work with that is down to, again, these two components of orchestrating the template itself and what content we're giving the agent, and then also the way we're structuring the custom agent and also knowledge as well. Before I download this and show you the output, I'm just gonna jump back in here. But what you could do is you could add multiple sections in here. Okay. And based on the information that you've given the agent for content perspective, you could ask it to go ahead and use static slides. So within this section, use these slides which are static, whereas in this section, use this sort of more shell of a slide that I want you to populate based on the context. So it all comes down to based on the section, what you want to be static, what you want to be dynamic, for lack of a better term with tempify at least, and then also how you tell the agent on what templates to use on the slide layout section. Awesome. So it's back to your triangle. Back to the triangle. Perfect. Pretty much. There's another question about MCP. Yep. Do you have MCP functionality support and the possibility to use our primary AI tool for input where the MCP connection may provide you with the presentation? Ed? We we have we have got MCP where we've been testing it and using it at the moment. It's something we're looking into more and more deeply as we go ahead, and that means you can access Templafy from your LLM. Good. Well, Matthias, to learn more about that, get in touch with us. That's not the focus of today's webinar, but yes. Correct answer, Ed. Great demo. Might have missed this, but can it access Templify's image library? Yes, it can. So it doesn't generate any images. It doesn't use nano banana or making all sorts of weird images you didn't ask for. It uses your pre approved images. Correct answer. Well played. Question from Alex. Are your servers US based or do you have regional EU servers? We have both. It depends on where your tenant would be spun up. So if you're a US customer, you can do it in The US. If you're in Europe, it can be in Europe as well. Awesome. Lovely questions. Thank you, everyone. Brilliant. So I'm gonna download this one slide that it made for me, and I'm gonna open it up. We can have a look. And I'm not just going to sort of show you this slide and then claim that this is exactly the right information. I actually do have the call notes that are in here as well. So you can see here that the call notes that I uploaded for NVIDIA are the software moat is a real story, software and AI is bigger than it looks, robotics is a long game, And there's some other information in here. I'll be honest with you. This is a this a call notes document I got AI to make for me for the purpose of this demo. But if we actually look at what's then been put in, it talks about these different topics. The software mode, the sovereign AI, the robotics. And it brings in specific details from those descriptions. That's all coming from that knowledge association and how we've orchestrated the our agent to work with that information. But as you can see, it's not necessarily exactly as it is presented in this call notes document. And I mentioned in the Ed's triangle that we want to look at deterministic approaches as well and how do we work with data. Because this is great if I'm giving context that needs to be inferred, but it can be difficult when I actually need to go ahead and be super rigid on how it uses specific information. And for that, I wanna show you how we can get to a document being created when the full triangle is in action. So gonna just close the call notes section. I'm also gonna close this one slider it put together for me. And I'm gonna open up again back here and bring us back to the admin center. So we had we just looked at the key insights slide agent. Now I'm gonna look at the proposal agent, and there's this nice little tag that says with brief. And I'm gonna explain what that means in a moment. It's the same UI when it comes to creating an agent with a doc with an agent name and a user facing description of what this agent does. Now this outline is a lot more broad because it's creating a whole document now. We're not just looking at one slide. We're looking at the whole thing. And we've also got a couple of questions or no questions in this case, but you can ask some questions as well. What sort of questions would you ask? What would be an example? So if we look back at the key insights slide as an example, and you've got a knowledge source of thousands of different call notes, Well, we need to know what you're looking for. What's what is the agent? Which call note do you want? A question could be, which call are you looking for? Who is the call? When was the call? And which customer is the call with? Or whatever it may be. Those could be questions that then make it easier for the agent to go into that knowledge source and find the right set of hormones. If we go down to configuration and into advanced, you'll see I've got an asset ID associated with it. And this is part of the sort of how Templify at the moment is approaching bringing data and really rigid instructions to our agent. And I'll give you a bit of a teaser, but I'm gonna show you as well what this looks like. But when you're chatting with the agent, in this case, we actually will generate an automated document that is dynamic, that both has very rigid integrations with Salesforce and other sources, but instructions around those data sources is pulling in as well. What this serves as ultimately is two things. Excuse me. A mega prompt, you could say it as. So instead of you having to write exactly what you need from the agent, it is actually being instructed by this document instead. And number two, the data being handed over to the agent with the instructions so the agent knows exactly how to use it. And that's where some that's where the real sort of power comes in on how you can enrich the content and make sure that the information that's being presented is in the correct slide, in the correct place, exactly as it's presented in your system. Sounds like a really strong idea that we should probably look into, you know, legally owning. Potentially. But maybe we already have. Oh. Oh, is that the patent? That is the patent, but I'll show you what that actually looks like in a moment. Okay. Nonetheless, within this, we've got a lot more information on the outline as well. We've got three slides. I've pinpointed those slides again from that template component. The knowledge from our call notes is still there as well. So we're really seeing the triangle in action now. So as I said before, let's give this a test and see how it runs. You'll notice this is a bit different from what we saw at the beginning. First and foremost, we can have these questions that come up and you can have as many or as as few, sorry, as you would like. And these questions ultimately are inferring to that document that's being generated, what information it needs to pull in from what systems. For this, I'm gonna stick with the NVIDIA. I'm gonna go ahead and say growth and innovation lab for the package. And then for the team, I'm gonna go ahead and say digital and AI transformation, given what we're talking about today. Once I submit those answers though, this dynamic document that I mentioned is now just being quickly built and being passed over to the agent. Once it's passed over, then I can get into the chat. And what you'll see as well, if we go back to that gliding scale of completion and context, this is also reflected in the chat. The context of the questions is a lot richer and a lot more specified towards what I'm building. It knows it's a proposal. It knows it's NVIDIA. It understands these pieces already. Yeah. So you can actually get deeper into my context because that's also what the agent wants to do for you. But for this, let's go ahead and answer these rather than just skipping over because it's proposal, and I wanna make sure I get this right. It's to secure a new contract. Are there any specific challenges or opportunities I want to highlight? Let's say supply chain risk. And should the presentation focus on all three strategic objectives? Interesting that it knows that. I'll get into why in a moment. Or is there more that should be emphasized? No. Just the three. Just the three. And And I'm gonna go ahead and say just build. Yep. So you already know that's a I saw it says there's 4,000 characters can be added there. Yeah. But you really keep it so you really keep it simple and it still gives you a good output. Precisely. Because that's also because I'm working with this brief document as well that's in the background being generated which acts as this prompt. Rather than me after spending time writing out the prompt and trying to become a prompt engineer extraordinaire, I can just let our technology do it for us instead. Fantastic. Any questions while the outbreak One, is I can see Barbara's already answered. So the reply is in there. It's basically how do you ensure compliance with the company's design standards? Barbara gave a very concise answer. And then Colin has asked, How do you get content in? Only by attaching single multiple docs in the whole prompt context? Or can I have a content library with hundreds of assets that are being created and updated that serves as a single source? It's almost a leading question. Colin is only still document agents. Colin, you must be new to Templify. That is what Templify is. Essentially, an enterprise wide curated content library for your entire organization. Document agents now is the layer on top. So if you've already got a Tempurify library, which as you see in the chat, lot of people do, we connect to that so the work is done. If you don't, lucky you. Because when you have the agent, you also have access to that. And maybe Ed can, give you a quick look at that at some point. So, yes, Colin, get in touch. I'll be happy to walk you through the platform personally. That's a that's a great experience to have actually. I've had a demo from you before. Yeah. A lot of people have. A lot of people some of them, not long no longer with us, but, you know, I try my best. Might be something about the brightness. I don't know. I don't know. I mean, I'll be honest. I've been struggling sat here this whole time because the beams of light are shining from John's head into my eyes. That's the second time you've come up and had a go at my head. And I think this is something that we can get to HR. I think we can have a look. Live now. That, but also, just in case, put them on. Thanks, John. Anyway, I can see everybody now. Hey. That might be better. So and I can also see the light of glory in our outline section as well. So with the outline itself, you can see that we've got a lot more structured information and a lot more slides that we're gonna work through here. But if we go in and actually look at all the descriptions, we've got a lot more information as well. We've got different strengths that are coming in from that brief I mentioned. We've got constraints, summary. We've got different context pieces of how NVIDIA is defining infrastructure, tailwinds. I mean, the information that we're now reading in our content brief has been elevated massively. And this is then how we get into a place where these documents be created a 100% every single time with minimal need for review. If we go through here, we can see all these wonderful slides have been populated with implementation timelines of different steps and quarters, when they're going to be happening, different metrics here, which I'll get into them in a moment. But I would remember these numbers for when we actually show you the brief itself and all the way down to the thank you. The only one I don't really want though is six strategic pillars. I think we have enough on the strategic pillars front. So I'm happy. I'm gonna hit create. Now I mentioned this brief. What does this actually look like? What is the document that's being generated in the background? Well, I can show you. Here is the brief that was generated in the background of it. And you can see it says here this content was generated by templify.ai. Now some of it was, and some of it was pulled in from a specific data source. But depending on what I choose here, brings out different pieces of information. I believe I can actually show you two different examples of a brief, but this is the brief that we're working with for the documents currently being built. If we go down, you'll see that some of the information like that sovereign AI infrastructure and such is being pulled from the knowledge. But if we scroll down, I asked to remember the numbers here from the data that was being presented on metric side, and you'll see the same numbers here. I also mentioned this component around instructing the agent itself. Now to us as humans and not agents yet, at least, you can see it says metric number one and metric one title. Well, what's that what that is actually doing is it's pointing to a placeholder within the metric slide. So when an agent looks at that, it goes, hang on a minute. I know there's a placeholder with metric number one in. And I know that now I've got this data asset for metric number one, so it can understand predictably and draw that red thread. So it puts that information into that place in the document. Furthermore, because this is being pulled in this 130.5 number through one of these dynamic bindings where it's super specifically pulling out this information from a CRM in this case, that's the information that's being pulled. So then we have that deterministic data capture and the instructions on the other end to tell the agent where to put it. And this is then how we can get into a place where we have super rich instructions. As you can see here, it's a lot longer than just those pieces of information, But we can capture the right information from the right systems and put them in the right place in the document. Fantastic. With that, what I will do, we'll just have a quick look as the other thing to note is when you're building a 20 slide proposal document, it does take a couple of minutes, but the amount of time it would take me at least is a matter of hours. So while this is just building, I'm gonna pop open my files. So I'll ask Isabella, our behind the scenes camera woman, to to not share my screen while I do that. I'll show you another brief as well. Are they Ray Bans or AliExpress? They are Ray Bans. But I can say that because I you'll never know. I'll never know. So I've got another brief now up on my screen, and this one's for JPMorgan Chase. And you'll see this is different information. And this is different information because when you are putting in that information in those questions, the brief is changing dynamically based on what I'm actually selecting. So it's not staying the same every time. It really is actually leveraging what your employees are looking to do. So if they pick a specific Salesforce record, they can go in and then get different information on that record versus what somebody else is doing. So it really is this dynamic instruction booklet for the agent to follow and get us to maximum results. And as we talked about at the beginning on Templify news and as you're now seeing in action, this is what's been patented. So it's really exciting for Templify, and it's also something that we are experimenting a little bit with of how could this be done not just for documents, because we're talking about orchestration for AI tooling, but any kind of AI tool. What if I could create this instruction booklet for whatever I'm doing in Claude or ChatGPT or if I'm building an app or a website or whatever it might be? That's what we're now looking at too, which is super exciting. I'm gonna take these off now because now I'm struggling to see my own laptop screen. I see. I think I'll have a go. There we go. Fantastic. John is now officially cooler than me. I think that was never in question in the first place, Ed. Well, at least I have hair. Three. Three strikes. Three strikes. Does that mean I have to leave now? No. No. Get to the end. It's alright. The presentation is almost done. Then you can then you can kick me off. Are there any questions in the chat, though, other than how could we look in sunglasses? Yeah. Exactly. There's a a good GIF. Oh, it's of me. Sorry. No. It's Tom Cruise. My bad. Well, yeah. It's confusing. Exactly. But one again, one of the things I like about the the flow there is you do get to see every step, what's being done, like adding content for each section. It's not like it just sticks and you just have this spinning wheel. You know exactly what's been Where's what's happening. Nonetheless, let's go ahead and actually download this and see what the output looks like when we have the full triangle in action. Now in this case, the agenda item eight hasn't been filled, but that's only because it's seven. So it knows when to fill things in and when not. And we can already see here, we've got our executive summary leaning into some of these concepts that we saw in the brief document, sovereign AI, the robotics and physical AI that we saw in the knowledge as well. I am just gonna jump a little bit ahead to the strategic objectives, full stack AI platform, sovereign AI systems, robotics and physical AI with specific points being brought in. Hopping back into the brief, and I apologize, I'm jumping a little bit back and forth. We can see these strategic objectives have actually been brought in and have been placed correctly in the document within this section here, which is really showing how we can bring in the right information and place it in the right slide and right placeholder as well. If we scroll a little bit further down to the objectives, we've got these numbers. 130,500,000,000.0, 74.6% gross margin, 88% data center share, 2,800,000,000.0 market cap. That's what Templify's market cap is actually after agents were released. But if we go down and look at that, they're the same numbers that have been placed here exactly as presented correctly. And that's because this information is coming from CRM. It's been deterministically instructed, and then it's been placed in the slide correctly as well. And this really is what happens when we can get to the full triangle effect where we understand the full context and we're fully orchestrating what content is the agent using, what knowledge does it have, what data are we bringing in, and how are we instructing it as well. And I think this is something that everyone could, I hope, can take away today, which is when you're thinking about your AI tooling, if it's prompted presentation, if it's anything, think about what you're doing in the background. How are you pulling the buttons and the levers to get your AI tool to do what it needs to do with the right systems, the right instructions, and the right task as well. Good stuff. But that's what I had. It's fantastic. We've got some great questions along the way. I don't know if it's going to automatically kick us off in one minute, so I do want to be conscious of the time. There's been some questions in the chat from Patrick and Susan. We'll be in touch with the answers for those if they haven't already. On the next Templify talks, we have an AI advisor who has worked for McKinsey in the past, has been in the industry ten years, and he's gonna be here to share some of the fundamentals of AI, what you should look into. And I think Ed's triangle will really help with his advice, knowing exactly what parts you need to focus on. Ed, we have thirty seconds. Final words from you, sir? Nothing other than if you're interested in agents, if you're interested in bringing AI to your presentation creation, get in touch. Love to hear from you. And honestly, also, good luck. It's a difficult world to navigate all this AI stuff. I'm sure you'll all do great. And if you've got any questions as well, feel free to bring them to Templify too. We wanna help you in any way we can. Well, as just said, we could've got another couple of minutes, so actually not gonna jump off we can. Oh, so but not to that point. I think where to find us? Obviously, you're leading the charge when it comes to the commercial parts. Yep. Where do we find you? Well, I'm not gonna give out my home address, but the there's the Temple of Fye Copenhagen office, and then there's That's your behavior today. I think I'm gonna give out your your There is the you can come and contact us, and our our commercial team will be in touch. And I've I've probably met some of you in the audience already through our wonderful account management team. But, honestly, get in touch with us via our website. Have a go with the sort of Templify Liberate that we've been telling you about, but the sort of open the open platform that you can try and and there's get in touch buttons everywhere there. Yeah. And then I think John might leave my email after this as well. Yeah. Definitely. I think one of the things I would suggest is have a go. Try out the the PowerPoint generator on the website. You'll get a feel for how it works, the flow, the ability to update upload your own contextual document there and what can be generated. And then just imagine how powerful that could be with your control in the back end. If it's just a demonstration you want, if you wanna learn more, then get in touch with your customer success manager, and then they can bring in Ed and or take you through the back end themselves. But but, essentially, get in touch if you want to learn more. And sorry we couldn't answer all the questions. We'll get to those on the side. But from Isabella behind the scenes, from Ed, and from myself, and everyone at Tenrify, thank you for joining, and have a good afternoon. Thanks, everyone.