Transcript for "When teams are creating documents with AI, who's in control?": Hello, and welcome to this edition of Templafide Talks. My name is John Taniakos Rasmussen, the host and creator of Templafy Talks. If this is your first Templafy Talks, welcome. I think I've already said Templafy Talks about six times, but that's fine. If you want to see more of the Templafy Talks that we've had, please go to our website, templafy.com. Last time, we were joined by the legend that was Thomas Mazzano. He is an agentic branding adviser, and we looked at how AI is changing brand governance and what happens when content is increasingly being created by AI instead of people. Today, we're continuing that conversation, but we're looking at the problem from another angle because AI hasn't changed how people... How work gets done. It's changed where it starts. For years, employees generally started in the applications their organizations made avail... Available for them. Today, someone might start a presentation in ChatGPT, continue it in PowerPoint, pull all the company knowledge through it in Copilot or another agent, and then ask a to improve it and review it. Now that creates a very different challenge for businesses. How do you let people use the AI tools they prefer without losing control? How do you make sure the right templates, content, data, and rules follow employees wherever they choose to work? And how do you scale all of this without creating more layers of complexity, inconsistency, and maybe most importantly, cost? That's exactly what we're exploring today. We're going to look at the idea of central document orchestration layer across your AI assistants, how organizations can connect those AI assistants to agents, to document applications, and the same company resources, and why control increasingly needs to sit a level above these individual AI platforms rather than duplicating those skill managing efforts. We'll also show you how this looks in practice. So we're gonna... There's no magic here. We're gonna show you, some real examples. And we're gonna show you Copilot, and we're gonna show you Claude and our custom document agents. We are using the web platform Goldcast. You can use the chat right now and let us know where you're coming in from. My team member Jane is in the background. She's going to be answering the questions, because we've got a tight schedule today. So as I mentioned, we're coming to you live from Copenhagen. Let us know where you're coming from in the chat, just making sure it all works. You can also leave questions in the q and a section. Okay. As you can see, I am joined today by three guests. Yes. Three guests. It's a full house at Templafy Talks. Let's find out who they are starting with you. Madison, who are you, Madison? Where are you from, and what do you do? Hi, everyone. Madison Hollali Feinberg here. I am a senior executive client partner at Templafy. So been with the firm now for about four years, and while my title as is might mean a little not so much to to many, I work with our customers. So I work with our customers across professional services, audit and accounting, and commercial real estate. So I'm really here to help represent their voice, their experiences, both with a lot of the AI systems you mentioned and also, with Templafy as well. And I'm joining you from just outside of Washington DC, but currently here in Copenhagen as well. Fantastic. It's a pleasure to have you here. Excited to be here. Been a long time cooking, haven't you, on Templafy Talk, so excited about this one. Same for you. Ayush, welcome. Thank you. Who are you? What do you do? I am part of our solutions team. So our team is responsible to talking to customers, help and understand what their challenges are. And on the basis of their challenges, we try to scope and amplify for them. So help design and just helping our customers. Thank you. Quite a bit. And you're based. I'm based in Copenhagen. Recently moved to Copenhagen, actually. So it's barely been a month. I actually used to live in The States for ten years. So just as adjusting in Copenhagen. Welcome again. So thank you. You're just visiting. You've moved here. And, Dave, who are you... Where are you from, and what do you do? So I'm Dave Jackson from Rich Orange in New Zealand. We're a Templafy partner. Have been for almost eight years now. So we focus on training and professional development and template development. So we do a lot of the Microsoft stack and also within the Templafy stack and, flown in for this, just for this occasion to attend here and really enjoying the time in Copenhagen. Fantastic. Well, thank you for coming all the way just for this. So we'll just... Let's get started, I guess. So we know that AI makes creation easier, but it does increase the need for quality and consistency control. Employees are increasingly starting work in their AI assistance, as we said, rather than the traditional applications. And I think we're going to answer today, what happens to control when everyone can create everywhere? So I guess, Madison, I'll start with you. Are customers standardizing in one AI system or are they managing several tools in your experience? Yeah. Great question. And I'll give you two perspectives, not even the customer one first, but actually my own. So when AI and AI systems were all coming out, I think we all flock to chat CPT. So there's a natural instinct, a natural familiarity with that as a point of entry. But now when you think about your personal tools and workflows and that of your companies, we're seeing those entry points continue to multiply and explode. So a lot of organizations, a lot of the customers that I work with, being big Microsoft shops, have Copilot as a default agent, as a default AI assistant. So when you combine that with the workflows that are going on the side as well as individual business units or teams that are now adopting Claude for a specific purpose or they're adopting, you know, some smaller AI assistance for a particular document workflow, you have this explosion of AI assistance, all of which have to be managed and end users are kind of trying to navigate what do I do for what purpose. And I have one client that comes to mind, one of the large, accounting and audit firms. You know, they have different business units, different service lines, and they service and deliver results to their clients in different ways. And if you have one firm but with different ways of working, you then have challenges of how not just your brand and your compliance, but also how your messaging and your methodology is then delivered. And so what do customers mean when they talk to you about remaining in control? Well, it depends probably who you're asking, customer side. So I I think you could probably think about this from two different lenses. First, you have your average staff, your end user who needs to create a proposal, a pitch deck, you know, insert your document or presentation type. They wanna be in control of the narrative. They wanna be in control of the outputs that they're ultimately going to present or circulate internally or externally. But on the other hand, you have the organization. You have your your brand. You have your your compliance stakeholders. They wanna be in control of not just how the brand compliance comes across throughout their AI systems, but how every aspect of what the organization controls is manifested in an AI output. So if you have not just multiple AI assistants, but multiple people, multiple teams using different workflows, different branding, different logos, different, knowledge, you can only imagine the compounding effect of how many different and potentially noncompliant outputs you may get from your AI assistants. So if every AI assistant is introduced... Introducing its own controls, what is the big issue with maintaining those? Yeah. If you're responsible for maintaining, you know, even just brand controls as one of those levers, And let's say, you know, you're an organization that has sanctioned use of Copilot and ChattGPT, that means duplicative effort of just managing your brand alone. We're not even talking about the knowledge, the data, the compliance, and all the other aspects. But imagine you're going into Copilot, you have to set up a brand kit or a skill and hope that your employees then use it, and then you also have to go do that in any other AI assistant. So it's not just duplicative work for someone like a brand manager having to manage that brand aspect, but all the other components that come in as well. And how much effort is in... Involved components that come in as well. And how much effort is involved? And it's not probably how long is a piece of string, but in the different tools, is there one that's easier to configure than the other in your opinion that you've come across? In my personal opinion, and and I'm not necessarily here to say that I am, you know, someone who's configuring a brand kit or or setting that up, but from the client side of of the customers that I worked with, I think it depends on, you know, the use case that they're looking to support. If you have a super, you know, let's say, specific, document or presentation use case that needs to be set up and configured through one of these AI assistants, that is gonna be a pretty heavy load, especially if you think... If you have to think about doing it across multiple multiple tool sets. So I can't say firsthand what the actual workload looks like, but I think you're gonna have varying degrees of difficulty and complexity depending on how familiar you are with the AI assistant, what is the use case that needs to be supported, how much variation or AI interpretation should be allowed for a use case versus control and more rules based. So it really varies quite a bit. And, David, you work for a consultancy that advises on these kind of things, Microsoft training but also, like, you know, development of content and templates. How is the AI boom influence the conversations you're having? We've been doing it for a while. So we've been through a few booms, which is interesting. So you can always look back. And people adopt things slowly, and it takes time to build trust and change behavior from, say, searching to asking AI. So we're seeing that evolve over time as they're learning various other tools as they work their way through. And it's take... Taking time, people... But it's then finding out where the where the value is for the individual, and then that translates... Translates through into work. So do you see customers working with multiple or some specific example, right, this is the one we're using? Copilot's the dominant in Australia and New Zealand. And some of that's around data sovereignty. So Microsoft have data centers in Australia and New Zealand. And and so they've got a really high market share in that space. So people are certainly using those tools. But then, you know, for myself here in Copenhagen, I'm jumping between Claude and Copilot for recommendations on what to do. And I'm getting different results. They're good results, but they're different. So we... But we are seeing it on primarily people focusing on Microsoft Copilot and Cowork. And how much... No. We talked about this a little bit earlier on, around the personal use of the tools and how that sometimes blends with the business use. Because if you have a... You know, you're using Copilot in your... Sorry. Your ChatcherBT in your day to day Yep. And then you've got a business task to do, Yep. You use Chat JBT. What is the inherent risk of doing that? I think we've been to this radio before. Yep. If you think, back from Gmail or file file sharing activities where you could access your email from your personal email, and people started demanding that in work as well as we moved through the cloud space. I think we're in that same evolution with AI. You're free to experiment and use Claude or Copilot or ChatGPT to how to build a cake, how to do whatever. And now they're seeing the benefits, changing their behavior, which is a big thing. It's a change from searching to prompting, and now they are then bringing those values through to work. And some organizations have caught up and are providing that technology, but we know from when file sharing, etcetera, was personal and not work that if you can't do it in in in your work environment with tools that are sanctioned and provided from work, you'll just do it with your own tools anyway. Yeah. So it's that governance area where organizations, some are providing it and catching up and some are not. But I think It makes sense, and it's funny because it's very similar to what you described there in the, the Gmail, the cloud, who is right, we're not using that for business. And I remember, obviously, working in the customer success org at the time. A lot of the AI tools were banned. We're in a... Where the customers were saying, so we can't use AI. We're not allowed to use AI for this. And now everyone's using everything. And I guess the question, I like the the phrase, employees and customers are not adopting AI. They're reacting to AI. And I think, Madison, a question for you is why why is Templafy different? How... Why is our outlook different to, you know, how do we help these organizations? Yeah. So I think going on the fact that it's not just one AI assistant that's present within an organization and their staff are using, you can think really about Templafy not just as another AI assistant. We're not trying to create another tool that people have to decide how do I use this and for what purpose, but rather think about Templafy as an orchestration layer on top of the other workflows and AI systems that are already in place. So I know we'll go into this more in the demo, and you'll be able to see it live once I I pass it over to my buddy, Ayush. But, if you think about having one central location for all of your approved assets, all of your approved templates, all... Everything that is compliant and ready for use, your admins can manage that from one central source, and that one central source can be Templafy. And then Templafy through our document agents, through our MCP server where we can integrate directly with Claude, with Copilot, with Chatt GPT Enterprise, can really become that orchestrated governed presentation layer. So your staff and organizations, their people, they don't need to change the way that they work. If they love working in cloud or someone over here loves working in Copilot for a different purpose, great. They can continue doing that. But with Templafy layered into the mix as part of that orchestration layer, it really becomes a essentially governed experience for brand managers and then also a centrally governed output that's going to be compliant and consistent time and time again. So as an end user, you prompt your cloud or your copilot you need a presentation, you're not spending time reworking it because it's off brand. By default, Templafy can be that consistent on brand layer. So to summarize then, Templafy follows the employee, not the application. Well said. Much much more succinctly. It's almost like I got given that line. Awesome. So, Ayush, now it comes to you, the the star of the show. I think I want you to show us how a team can generate consistent outputs using Templafy, and I think you've got, a side by side comparison, something you can run together. Absolutely. So let's start in cloud. So, again, to Madison and Dave's point, you can start anywhere you want. And one of the things that we do in our solutions team is to make sure that there is less friction or as less friction as possible So we can work with any AI assistant. So let's start and cloud. What I will do is I want to have, I want to close an opportunity that I have. I know that it sits in Salesforce. I know there is, some proposal that I want to create. But I just... I will just go to Claude, and I will type in my intent. So here, sales proposal presentation that I want to create, and I just mentioned my opportunity. I'm giving my intent. Templafy will recognize that intent. It will also recognize what exactly are the agents that are available for this opportunity. So as I was working in sales, as was... As, it's an opportunity that is in Salesforce, it is recommending me that, okay. You probably wanted to create an audit proposal that lives in Salesforce. So I will just let Claude tell that please go ahead and do that. So it's guided. It's also not going off the rails by just one prompt. It is asking you... It is guiding you on what exactly it will be doing. The second thing is that it is also showing you the design of how the presentation should look like. We are talking about being on brand, on being just branded errand. So that is what we have ingrained in this, experience. As soon as you, ask Claude or any of the AI tools that you like, you will actually see that these are the designs that it's choosing, and I can just choose that design number one. Third thing is it is also looking at Salesforce in the meanwhile. So, again, less friction possible. So you don't have to go to Salesforce to look for all those opportunities. It will find that information, by calling the MCP tool, connect it to it, and just give you back that information. So all I have to do is just tell Claude to look for this opportunity, which is Bryte's room materials that I want to close. So I will just select and go ahead, and it will go to the next task. And this is the most important task now that I have given it my opportunity. I have given it the intent that it has to be a proposal, and it has the information that it needs to gather from Salesforce. So now it will just go ahead and do that exact same thing and start, creating that PowerPoint. And in the meanwhile, I will also do the same thing in Copilot as well. So so a quick question before you go into PowerPoint. What is Templafy doing that Claude can't already do? Why would I need Templafy? Because, you know, I can use Claude to build a proposal. Great question. So there there are two things to it. So it depends on the level of prompting That you need to have because every time you're creating, a power, a PowerPoint presentation or any kind of intent, basically, through cloud, the results will be different. It is the nature of AI to be different, just because how it is actually built. You can control that to some degree, but you need to really, really enforce all those things every single time, and you can't depend on every single user within the company to do that. So that is why, it is important that Templafy actually is built within, Claude. As soon as it recognizes your intent, it will actually just follow your instructions and, create the PowerPoint for you. So first, it is brand adherence, so the brand people will love you. The second thing is, it will be faster, and it will also be accurate because it is using your, the rules and the, data that you have already governed rather than creating every single every single thing from scratch. Perfect. I have some more questions for you, but let's jump into the Copilot example. Absolutely. What's happening now is exactly the same process as the... As what we just did in Claude? That's exactly it. Same prompt. Nothing different. The MCP... Amplify MCP is coming in, and it is being consistent across any AI platform. So in this case, again, it is showing me exactly the opportunity that I need, the agent that I need. So I select my agent and, again, submit it. And the next thing, everyone probably knows will be, what exactly is the agent that I have to choose and then the layouts. And so I do that.... The same, proposal template that we saw in Claude. Exactly. Just in Copilot. Exactly. The same design, the same layouts, every single thing. So a user doesn't have to think, and especially salespeople. They don't want to create PowerPoints. They don't... They want to sell As fast as possible. So why not just make that as easy for them? So in this case, I just approved the request, and we move on and Copilot doing the same thing. So Copilot and Claude, essentially, you've just done exactly the same thing. You start with a prompt. You've had an intent. You've now got a consistent output when it comes to this finding Templafy via MCP and then finding the right template for the job. And so that template is configured in... Is it in Copilot? Is it in, Claude? Like, how... Where is the template configured? Very good question. So everything is actually happening, within Templafy. So to Madison's point earlier, configuring brand kits or configuring skills is hard. There can be another AI tool in the market, two months from now, and you probably would have to do that same thing again. So rather than duplicating that, in Templafy, you would have to just do it once, upload your PowerPoints that you have, upload, the brand colors that you would need. If you need to select the data from a different, CRM or any data connected that you might need, Templafy will be there. And after that, everything comes together, and you don't have to reinvent this every single time. So that's interesting. Right? What is reused rather than regenerated? How Does this approach help manage unnecessary AI generation costs then? Exactly. So, if, let's say, part of the proposal sometimes is to have information about your company because you're pitching your business to, or because you're working with some company, you would have to make sure that the information about your company is standardized. Yeah. You don't want AI to actually generate that information or fabricate that information in any way. There should be 100% consistency every single time when people are working on those things. Same thing for terms and conditions. You really don't want those things to change, or fabricate through AI. Maybe AI will do a 100% fantastic job depending on the prompt that you create Or it might not. So why leave it up for chances and actually just, have AI do those... Or Templafy apps to do those things for you? Awesome. And so you have Templafy. You have Copilot. You have Claude. How does it work then with the token consumption when it comes to a proposal like this? Obviously, the proposal is ready now. What will be the difference for Claude... An organization using Claude, not to just use Claude to generate this as opposed to having Templafy and Claude together? How does that... What's the impact there? Yeah. So it's a good conversation. It's a good question because we see this question coming up every single time when we especially, we've made, meet with, IT teams managing AI. They love using AI, but, when it comes to token consumption, that is also high because it is so hard to predict what the users will actually add in terms of intent. So the token consumption can be changing every single time. So enterprises love unpredictability. And if I may chime in here, I work with, a a client who is also a client of of Claude and Copilot, and they've been testing the waters, particularly the Claude, for presentation generation. And and I think the result of the output itself is, you know, this looks and feels like it came from our organization. But the quote of it being 20 or $25, and that's USD, American here, per presentation generated is just unattainable for an organization. And when you think about how AI continues to grow and develop and sometimes we're just using it as our personal playground, if you have those unpredictable costs for someone just playing around or going through multiple generations to get the right output that's going to actually be suitable, ready, consistent, governed, and high quality enough to put in front of a customer as an example, your costs are just gonna go through the roof. So with Templafy, of course, there are a couple of mechanisms as to how we can help influence and and help govern more consistency with your token cost and consumption. But the the net net is because you're not having to rely on building everything from scratch for every individual that's asking for a presentation or a prompt, and that's natively coming through Templafy. I'm seeing with some of my customers who are seeing those, you know, $10.20, $25 per presentation with Cloud as a stand alone, you know, come down upwards of 40, 50, 60% of that when they have Templafy layered in. Is that okay? I think it's just scary for IT managers that they're running an open bar tab, and and they don't know what it's gonna be at the end of the night. Whereas... And that's not a good place to be. Right. And so anything you can do to put some controls and guidance on choice and agents and and the outcome of that is reliable presentations of the same consistency. And one more thing to that. When people actually work with AI, they want it to be super fast. They want something that is super, super quick and actually just gives you all the results and a perfect presentation every single time. And it is not... So it is possible to do that, but sometimes, if you actually let some things not be governed with AI and just let those things be added as is, things become faster. The case studies that you might have or the information that you need to about, need to add about your company, those things doesn't need to be generated with AI at all. They already exist. They already exist. Exactly. Okay. So both presentations have now been generated. One in Claude, one in Copilot. Let's have a look. Absolutely. So on the left hand side, this is coming from Claude. On the right hand side, this is coming from Copilot. And if they're on brand and if there is control and if there is consistency, you cannot tell what was the AI that was used, but you actually get to see just the output, and it looks fantastic. So if I go to slide one, slide two, slide three, as you can see here, it is actually going through that structure. It is actually... If you see the sections that are getting created, there's an introduction, there's a problem statement. And if I scroll down further, there is audit objective and the solution overview and all that. So that is the structure. But let's go a little bit deeper, into, let's say, bios that are generally a very good, slides that are getting added into these, proposals. So if we go down here, we will actually see that these are the people that were part of that opportunity Within Salesforce. I didn't have to think about it. There was information already in Salesforce, and my Identify agent was able to just bring back all that information and figure out that Jordan Lee is actually a person, and he is, working within the company. And he he... We have information about it already. So if we have information about Jordan Lee here, why should we actually create that same information again? So here, as you can see, Jordan Lee's present slide is here. And if I go to the other, team members here, they are also here as well. And same happens for accreditation. So really, really important stuff because this is what, gets you so many other companies as well. So accreditation. So our soft tools or soft trees and all the certifications we have, you don't want to mess up that... That AI. So as you can see here, all that information is right there logged and just getting added into the presentation. I didn't use the word generate because it is actually not generating. It is just taking that slide from the Templafy library and just So what parts of it are AI? I mean, this looks to me that it's so consistent across the presentation. Yeah. How... What is the, what's the benefit of having Claude and Copilot in this case? Absolutely. So let's go here. So every AI actually works differently. So for example, slide six here. As I can see or as everyone can see, there's a different choice of slide layouts that were used. The before and after are actually a little bit different. In this case, it is using the, yin yang design. And on the other side, it is actually using small circles to just represent that whole point. So that is why you can actually say that this is AI generated because there are two comparisons that you're doing. But in a real world, you probably will just have one proposal, and it would be so hard to figure out if it was actually created through AI or not created through AI. You touched upon it in the start. I think you almost overlooked it. The structure in... Is exactly the same in both. Like, the AI hasn't created that structure based on, like, its own interpretation of the task. Exactly. The structure is set in both. Is that done on the Templafy side as well? It is done on the Templafy side as well, and it will follow not just that, but also the brand every single time. If I look at just the layouts here and the presentation that I have created, it is actually using the designs from Templafy, from the brand that has to be used rather than fabricating that information when you normally work with cloud or Copilot in that case. So I think that's great. Like, inserting the bios is always a pain and often is recycled from the previous deck that you've used. So this is doing it immediately but in a reliable way. And then you can leverage your token investment on the model you want Exactly. To generate the important stuff in the presentation and not have to spend that same expensive tokens on generating bio. So there's some really neat stuff in here Yes. Which is good. Exactly. Does this reflect how employees actually work, Madison? Definitely. Both speaking for myself as an employee and, from the customers that I have the pleasure of working with as well. I think this demo is so awesome because it illustrates so many things that are important for your average employee who is using AI to create a presentation and that I can prompt, you know, I need a presentation for a proposal or whatever the business need is. I don't have to think about adding what brand should it be, what slide should be pulled in for proposal because that's already been configured. So when you think about the configuration aspect, that's also hitting on the governance of what someone in a brand or compliance angle cares about. So you don't just have 10 people creating proposals and getting 10 materially different things, but you have a unified structure, a governed template, and you have, like Dave was saying, AI where it's suitable within this template and and other rules based automation and and other components. So I would say absolutely, definitely speaking what the customers are looking for. That's great because then the person presenting... Putting the presentation together can focus on their customer and values to the customer. They don't have to waste mental effort putting in bios and terms and conditions and all the rest of the stuff, which they care about the customer and the results of the customer rather than the rest of it. And that that just happens for them, which is great. And it's not just about not having to put them in, but it's also knowing that if terms and conditions or things that need to be reliable and rules based and not AI generated, They're just automatically put in in the way that they are supposed to be not AI generated, and that is a huge, huge time saver because the amount of rework that goes in if something is completely AI generated to say, does this term align with what is sanctioned by the company? Like, you can't expect your individual, like, employees to know that. Yeah. And you don't want them to focus on the... Focuses on... For the customer and the results on the pricing. Exactly. And to add to that, because this is a sales proposal and because this is information coming from Salesforce, you need to get your pricing in an order. You cannot mess up the price line at all. So just giving you one more example where, again, information just was just, pulled from Salesforce other than getting, generated through AI. So this is one other example, where you wanted to make a million bucks, but life happened or AI happened, and there was one or more zeros. So... And I would really underscore this point too as well because we've talked a lot about brand governance and compliance governance and really wanting to produce a presentation that looks, feels, and is like it came from the company that's, ultimately, yeah, producing it or that the user is producing it on behalf of. But I also think the data integrations and connectors that you were speaking to before is another huge one that I see in my customers because, again, you wanna meet people where they're working. And if they're working in an AI system, they don't wanna go to Salesforce. They don't wanna go to this knowledge source or this other repository where they need to get information, and they certainly don't want AI generating the data. So if you can connect not just the AI assistant where people wanna work, but also the brand, then the compliance, then the data, and you can output that is marrying everything together. Like, how how could it not be the Nirvana that everybody wants? Nirvana. Nice. So this is that this is that orchestration, ma'am. Exactly. Whatever I said. And I think the way I really like to describe it when I'm talking about it is that Claude is still working in the background, but it's working on only what it needs to work on. Right? It's pulling that information. It's it's putting it into the right parts of the template. It's not building a slide from scratch. It's not trying to build icons or tell a story. That's taken care of. So not only is it simpler and quicker for the user, it's a lot cheaper for the organization as well. Exactly. But I know how I work with slides is sometimes if I'm working on a presentation that I have already generated Yep. Sometimes I wanna jump back in that presentation And then I want to maybe edit a slide here and there. Yep. What would I do with that? Would I use Claude Co pilot for those as well? You can. But, again, that level of prompting that you would have to add to that will probably be, it will be dependent on who the prompt... Who that user is. So you can absolutely do that, or we have the agent that can also take care of all those things within Templafy as well or within PowerPoint. So let's show you that. So I will actually maximize one of the PowerPoints we already created. We have a nice little add in that can just, look up the whole PowerPoint. It has the context of the PowerPoint that I just created. It knows that I actually went to Salesforce to get all that information For me. All I have to do is add a little few things and just ship it to the customer. So this Turnify agent Is connected to the same... With the same MCP that what we just saw. Correct? Exactly. Okay. And we work in that exact way. One PowerPoint that is, let's say, getting created through cloud or Chargebee or any other tool. It is not 100% ready. It can be in some ways, but, generally, what we see is you want to proofread things. So let's actually look at that right right now. Fantastic. Perfect. So I have a prompt here that I want to use. So... And this would be the same... Like, the prompt you're gonna use now is is it... Imagine if you was asking a colleague to do this task for you. You just describe what you need. Exactly. Or it could be, we actually see this with consulting more that, let's say, an associate or a senior associate was working on a proposal. Now when they want to ship it to a customer, they wouldn't actually do that directly. It'll probably go to a partner. They will actually put some comments in, and then you would actually get back that presentation, and then you will be making those edits again. So this agent can also take care of those things. So if you have comments or speaker notes or anything like on the presentation, it will be able to capture that information and just Get on with it. So in this case, what I'm asking is create a new slide that just justifies the fee that I'm adding and just capture all the context, from the presentation created. And just give me a nice looking slide, and you will actually see what the agent is doing every single time. So first, it will, look at the intent, of course. The second thing that it it will do is it will try to find the slides that are actually already in your Templafy approved library. So if there was any slide that could have been used, it will actually try to tell you that as well. In this case, I can see here that it looked up the presentation. It found the original brief. That's fantastic. And it is asking me, which baseline should I use. So I will just tell it again to just use the figures that are already available, and, it will start doing its thinking again. And we want to have that same experience as well where it is guided. The user is instructed what it is doing so that, it is not left for just one prompt. It will tell you what it is doing. And while our our handy dandy AI, edit agent here is thinking and doing the work on our behalf, Another use case that I'm hearing really commonly within my customers where this is particularly applicable is not just, you know, adding a slide that maybe the initial generation missed and, you know, I had in my head I wanted in there. But, also, if you look at the slide front to back... And I have a client who who works out of Canada, so they're often doing, presentations in both English and French. So if they went to Claude, as an example, as a stand alone and wanted to create an English version and a French version, again, going back to token and cost consumption, you're running that back twice. So you can simply, with Templafy, not just create a new slide, but you can have it effectively take your entire presentation in English, turn it over into French as an example. So it's not just about adding where the first draft maybe didn't have everything that you saw, but maybe turning it over or changing the tone of voice or in this case, like I was mentioning, the language as well. And if you're sometimes lazy, you can just change the company name and just ask the agent to do that. Permission to the The way I use it the way I use it the most is if I have a box... And it sounds like a simple use case, but it takes time. If you have, like, six boxes on a screen And they're telling a story and you think, oh, I need a seventh. Instead of resizing every box and maybe those headings, you say, add me a seventh box, and it will do that for you and make everything else neat at the same time. It's... I mean, for me, it's it's saving me a ton of time. But, but let's see. I think Martyr not harder? Yeah. Let's go with that. Let's go with that. I've never been accused of being smart. I understand. So it is about to actually create the slide. It gathered all the data that was needed. It is looking up the designs that are available in the Templafy admin center or the brand, that has been approved. It will look for that and just add that slide content. It will fit in nicely and just create that slide for you. So, hopefully, in a few seconds, it should be there. But what I like about this, right, is this is... We're showing this in real time. And I think that's something that, you know, we always... That that comes up is is like What about this is faster? What about this is faster? What about this is faster? I think the choice we've made is that we want to aim for accuracy and consistency Yes. Over just creating something as fast as we can That is then needs to be proved and checked and reworked. I think that's important where speed is good in some areas, which is just a revival of consistent terms and conditions and other stuff that just happens, you know, to do it. And then areas where you want to spend the time to invest to get... Make sure the proposal's exactly right, then, yes, it takes time for the agent to get there. And that's that's fine because you're gonna get a better result. Exactly. And, one thing I didn't know if I mentioned, all the connectors and all that information, is also, preconfigured here in this agent as well. So if you had your connector working, in the create phase or let's say when you were working in Copilot, it will also work in this, experience as well. So you're not... We are not taking that experience away. We are just making sure that that consistency remains across the whole journey when you're creating that PowerPoint. Following the employee. Following the employee. Yeah. Exactly. Alright. So now we have the slide. We have the slide. It is justifying all the information that was added in this proposal. I can, of course, look back and just see what it was. But as you can see, first, it is actually finding the layouts that were approved. It is finding the context that was already available within this PowerPoint and within Salesforce, of course, and just creating something that is on brand. So all the little things, for example, my footer, my header, my brand logo, my colors, I don't have to think about that at all. It will be just there for me. I just have to provide my intent and create something which looks nice and, presentable ready. Fantastic. So as a recap then, Simplify helps employees create compliant on brand documents wherever they work while the organization stays in control. So what is one takeaway you would want the audience to go home with around that? Use any AI tool that you want. We, in our solutions team, have different kinds of AI solutions that we work with every single time. But when it comes to actually creating an important document, don't leave that for AI. Don't leave it for AI alone. For AI alone. Exactly. Let AI actually do the important bits. Let AI actually do your research, your analysis. Let them actually figure out who this customer is and what are the, acquisitions and the things that they have done to understand their business. But when it wants to create a PowerPoint, you will be just spending more and more time and just scratching your head. So rather than use Templafy as that source so that you can save a bit of time on that. Thank you, Harish. Madison, it'll be interesting to hear your your one key takeaway, would you like to hear? I think I have if I have. I'll pass it to Dave. I think... Look. First time, in Copenhagen, so I'm using AI tools, and I'm getting a different result from different ones on where's the best place to have breakfast. And that's fine. But then I'm using just one tool to get me to locate there with with Maps. Because it's the same world here. You can use whichever AI tool you want for the deep thinking, but then you wanna pass to a consistent approach with the agent to actually generate the presentation. And For me, I think it's about enabling employees. Like we talked about following the employee. And there's been this argument for a long time around shadow ITs. This department is using this and this department is using this. And there may be teams that want to use Claude. Others prefer Copilot. We have Gemini customers. How ChattyPT, how do you then control the outputs and derisk the business? And that is essentially having that orchestration layer in the middle. So whichever tools are being used, Templafy will be there with the important parts, meaning every output is consistent and it meets the user where they are. So as a recap, we're going to follow the employee, not the application. That's what we want to get across. And there's going to be consistency across every, AI system that you have by having that orchestration layer in. We want to ensure that you reuse content wherever possible instead of unnecessarily generating and increasing your token consumption, and we want to orchestrate those document workflows centrally. Don't let AI build your proposal workflow. Have that in a central location and distribute that to your AI systems to use. So I think that we've covered the orchestration layer of simplifying how via MCP, we can support your use cases. I guess, before I go, I wanna say thank you to all of you for joining. I wanna thank my guests. I wish you live here now, so you don't get as big of a thank you, but you get a huge thank you for the demonstrations that you've done. Madison, thank you for calling all the way from New York. It's a pleasure having you on. Dave, you've literally come from... There couldn't be a further place away on the planet than New Zealand. So thank you so much for calling over. I'm looking forward to spending more time with you this week. So from the team here live, from myself, from the team behind the scenes, thank you so much for Telify Talks. We'll be back soon. Thank you for watching. Thank you.