Hi everyone. Welcome to our webinar. Today's topic is, do we even need a CMS anymore? An AI experiment in enterprise website generation?

As you know, like we we've been in the dxp space for a long, long time. So I'll give my intro. So my name is Akshay sura. I've been doing this for way too long. I've been working in the CMS DX space for about 20 years now, involved in a lot of the Sitecore community stuff. And with me, we have Kamruz. I'm Kamruz. A lot of you hopefully know me from throughout the throughout the online space, especially, I've been in the CMS space for 1617, years now, so not quite as long as actually I've been primarily in the site course space, but obviously we've been doing a lot of work in other CMSs, like Kentico, Contentful, content stack and other headless CMS over time. And we've also been experimenting, I'd say, over the past several years, in a number of different ways of doing CMS, particularly for our own website right where, when the JAMstack movement, as it was known before composable, we were, we were playing it heavily in that JAMstack space, particularly for content management generation. So the whole thing started with, with frustration, I should say so wanted to redo the site. Was going through a lot of different things, so back in the day when cameras and I first started, we did Netlify CMS with Markdown files. Still had a CMS UI, but it would write to the Markdown file. They get checked in, gets built. Static site worked perfectly. But then we had people come like, oh, I can edit this. I can't add that. Can you do it for me? Because people didn't want to work in Markdown. So it ended up being cameras, and I doing it. So like, okay, fine, we'll find something. So we transitioned to like a headless CMS, and that's been working out. But I know it's a it's a provocative but it's a practical question as well. Do we like in this day and age of AI, do we even need a CMS? You know? Not because CMS platforms are bad, you know, not because of enterprise governance. It doesn't matter or doesn't exist, right? We still need those. But AI can generate, right now. It can generate the page structure. It can do the content models for you. It can create components which are reusable based on whichever framework or UI frameworks you want to use. They can do a multilingual they can do translation. They can build a front end for a page in minutes. So after doing what, all the things cameras and I, and pretty much everyone who's watching this has done. We've built websites. We've built the same heroes, 15 million times, 15 different ways.

Enterprise projects are super expensive, not from like a licensing or publishing content is hard. It's just because of everything that takes you to get to that infrastructure and set up to get to a point right which we all go through. We have to, you know, when we get a project, you do the requirements, you create the content models, you create the wireframes, and then it goes to the front end developers, and they can create components and the themings. And then the backend works on migrating hundreds and 1000s of pages. And we create workflows, we test we go back, we test UAT and do all these fun things before the even the pages go live. So I was like, Okay, let me see if I can do something. It was in my brain. AI was at the level where it was it would help me build something. What happens if I give I built a CMS. So I built it actually overnight, but it took about two to three days to refine it. What if I give it a brand kit? I give some rules and regulations on what it should use, and we can go into it more once we get into the demo. Did not go with the mock up or anything. I was like, here's the Konabos website. Here's my instruction, go build me something. And that's what, that's what I did. And I just wanted to, I'll go through that with you guys and show what worked, what didn't work, what could have worked, and how does that affect our idea of needing a CMS, right? Do? This is not a hype. I'm not promoting this as a product. This was just an experiment. I am not anti CMS. This is mainly like an architectural and implementation conversation to invoke thoughts as well as for us to come up with a solution or the way forward, right? So what is cms actually for? So I know that, you know cameras and I have been working in Sitecore and other CMSs for a long time. Sitecore, AI, which was just launched at the symposium a few months ago. Can do a lot of things, right? It can build campaigns. It can create content. It can push everything that you built for a campaign at a specific time. You can roll back. You have the marketing API that you can use to do things. You can build your own agent. I mean, you can do so much, right? But someone has to do it and but once you put that into place, all the back end systems pick it up. They check your SEO score. They do all of these fun things every CMS is adding their own AI capabilities to help you do much more and not do the repetitive tasks, right? And we all love that. So AI is no longer like assisting developers, so it's generating architecture, components, content, models, front and structure, like we mentioned, if AI builds the first draft of a website, right or landing page? What exactly is the CMS doing? If, if AI can do that, that's the that's the thing we're trying to get to is where do each of these stand in our in our mind?

So cameras jump in anytime you want. So flow, actually, just you could jump in. So like CMS is inherently, there a lot of CMSs that are free, but they have, you know, things that you need to do, right? So, for instance, you can host your site on WordPress. Great. It's a very community oriented. CMS has its own challenges, especially in the enterprise space, and we typically deal with enterprise space, so that's why we tend not to get into other things. But where does implemented in a CMS go? So obviously, the CMS license, that's something that you know whether you deal with vendor a, when the B, you still have to do it, but it's purely like the way you do it in terms of actually implementing the CMS the first time when you do it is because you spend a lot of time on architecture, or how the components work with each other, how They can be reused. What frameworks are you using basically engineering as to how things get done on the front end, how things gets done in the back end, setting up your QA cycles, the CICD pipeline, the migration documentation for not just the people who like content, authors and editors, but also from a development DevOps perspective. So once you do all of that, rarely do you see that producing content and publishing it being the expensive part. It's not like it's the initial build or anytime when you do the big lift, and then and then we talked about this in terms of other things that make it more expensive, in terms of DevOps deployment complexity. So we want blue, blue, green deployments. We want to do this. We want a disaster recovery. And then you layer on top of it, governance. Like every piece of content needs to go through this workflow, some content goes through that workflow, and then eventually you'll end up in a spot where you would want to make changes. It's not like you build the system once and you're done, you don't need a developer anymore. You would still have change requests for features, so things that you need fixed.

So all of this goes into what makes a CMS or a CMS project expensive, whether it is the current build that you do or a an existing build that you take and you add things on top of it. Yeah, yeah. I think in the enterprise space as well, we do see this get expensive with scaling, like global scaling, particularly, we've had a lot of use cases where they need to deploy into certain markets like China, and that makes it more tricky, because it's Hey, we need another instance of the same software. Because. It needs to live within, within that firewall, and new, new branded sites, marketing sites, new micro sites, launching those with new brand kits hits. It's Hey, we need to. If the projects aren't built right, then you can't just be you have to rebuild from scratch, or somebody's got something which is just my my use case is very unique from everybody else's use case, right? So I know you mentioned, like the menus and the banners and the CTAs. You know, we've seen this a lot where, nope, mine doesn't fit this current kit, so we need to build a brand new one, so that, again, just makes things more difficult, more complex, more expensive, as well in long term to maintain. Yeah, for sure. So then coming back to the demo part of it, and I hope the demo gods are smiling upon us. I did pray to them. And then also made sure that all of the API limits, and there's enough money in everything so it doesn't break. As I mentioned, we'll take a look at the experiment that I ran right so let me share my screen. Let me get there. Hopefully no work. See Anton's on the on the stream as well. So hopefully the the the AI endpoints on down, like when he Yeah, cool. So I built this overnight. Megan, I already had this specs I had written, and it was always in my mind. I'm like, let's just implement it, see how it goes, and I'll share how I feel about it now that it's all done, and I've done quite a bit of work on it, but concept is, you could have a multi site. You could have multiple languages per site. You can import an existing site based on navigation or site map a whole nine yards, right? So essentially, coming back to it, what we have set up is, you know, something really simple here, the languages preview or all, could you just zoom your screen? Oh, okay, sorry, all right. So just very simple stuff, plumbing on the background for like, being able to import languages and things we are not going to get into every single thing in here. We do have media. It gets stored, it gets CDN, and then, in terms of providing the guidance, some of it is controlled by, I guess, the admin on the upper side giving this info. What's the purpose? What's the business context? You could if you have products, team members or FAQs you want to include as part of the context, to give it to the AI along with every prompt, then that's something you can specify. We'll come to structure in a second. So from a brand perspective, again, I'm trying to replicate the konabos.com so I put in the color palette, the typography, the tone and the voice that I want to use, things that I want to use. Again, basically specifying what I need when I you know every time a prompt runs, what is it going to use? And then you can also specify a few websites. Here are my inspiration, and it kind of gets plugged into the prompt, right? So that structure wise, again, I didn't want to add anything manually. I pre did the Header Footer just because I didn't have the I didn't want to do it while I was on the demo. But again, from a SEO, geo tagging, all of that fun stuff, you can do that, but from a footer and header perspective, all I did is I just basically said, generate the Header Footer from the Konabos website, and it basically built this for me. What's interesting here is I did not build the structure. I did not give the content type. I did not do a content modeling exercise. It's not the greatest in the world because the links are just text, right? I totally get it. But again, this is an experiment. I'm not trying to sell the CMS to anyone, but it generated all of this based on the existing, existing website, which is pretty cool. And then when we come to the pages, right now, I don't have any page, but what I'm going to do is just so we know I'm running this on my local because I don't want to share the public, public image, public URL. So this is the current Konabos homepage. Has a ton of fin. Four different components. It's built a little bit differently. It's headless, using different headless CMS. What we want to do is, I want to be like, Okay, I want to see if I can generate this right? So let's do 915, and then all I'm going to do is I'm going to provide a prompt create the Konabos homepage, where this what I want to highlight. Just want a better looking component. So while it's generating that, I'll walk you through what is happening in the background, right? So one is from a prompt perspective. We specified what the prompt is. It already has a brand kit. It already has the colors. It knows the site that we want to replicate. It knows the page we need to replicate. But added in the background, what we're doing is, I set up a thing saying, okay, use radix, Shad, UI, Shad cn, and then we have a subscription for shared cn blocks, so use any variant of any component in chat cn blocks, and then it has a coding standards in terms of, here's what I want you to do, here's what I don't want you to do. We typically use this coding standards document on pretty much every project that we use, and it keeps getting updated from a front end perspective, back end perspective, back end perspective, that's also provided in this and there's an architecture Markdown file as well, which kind of tells you, this is how I want the application set up. Here's what I need you to use. Here's how I want the props done. So every single thing from a context perspective, whether it's coming from the brand or the site that we're trying to replicate. Part of it comes from the context of what gets saved in the site settings. Part of it comes from the standards and the MD files which go along with it, right? So what happened just now is it takes about 40 seconds. That's why I kind of knew what I needed to talk about. And it generates this thing in here. So again, trying to look at it, you can delete the thing. It generated some metadata so it can handle anything for SEO, AI or AEO, Geo, you name it. In terms of like, FAQs, it can auto generate all of the regular social sharing. It regenerates the it generates the Meta Title, Description keywords. If you were to change content on the site, you can actually come back here and regenerate SEO based on the current page context. Thought that that was useful. And again, it makes a call, updates all of these. And then this is a basic thing that we had, which is we named something, we gave it a prompt. We can regenerate it. I can translate it, if I want it, and it will create translations for this page. And then we have components. So what it did, and again, I did not create their no content type. There is no content type concept in this CMS, or concept of a CMS, it went ahead and based on all of the components that is has access to, it has access to about 172 or 178 I believe, Shad, CN blocks components and their variations, and it's able to use that and create this by itself. So it created a link component, as you can see, and then it uses multiples of that inside of the hero. And it created this by itself and the background image. And then every single thing that it generate again. I don't want to scroll through the whole thing, but I will. But the easiest, probably first, is probably to do the preview so we can kind of see it. So if I refresh this, you can kind of see that it generated the hero, generated the customer list. It took the information that we had on the homepage and represented in a different way as to how what is done, proven results. It kind of refactored it a tiny bit different way, customer testimonials. It brought it in. Probably could have brought the names, but it didn't really bring in everything I wanted, but it kind of got the component there, right? It represented the white shows, choose Konabos in a different way, use the same content. And that was something which was really interesting. Is over the multiple times that we were iterating through this, I made sure that there was no hallucination, because it was bringing in people from other companies in the Sitecore realm into it and things like that. So I had to redo that a little bit. And if you notice the header, the way it renders the header, the way it renders the footer, and all of the content is pretty much there, which is kind of nice, and I didn't really. Do anything. So then if you go back here, every single one of these is represented here. So for instance, go, I'll do the current time. So people don't say that I didn't do it. Let's do that. So as you can see, you update it. It shows up so it's hooked in. So now I'm like, let's try something else. So there's an option here again. You can add a component, and you can specify if you wanted a specific component or let, let AI decide you could add reference material, as in other pages that this component that you saw looks like another website where you were inspired from, or like a screenshot of something that you want. And then all I'm going to do is I'm going to generate this. So this component technically doesn't exist, because there's no pages on this site other than the one we just generated. So what this is going to do, hopefully, is it will generate a video component. We will have access to the video component so we can kind of see what it does. As you notice over here, created the video component. It decided we needed a thumbnail. Even gave us an autoplay feature. I didn't ask for any of this, but we did get it. So again, I will do the current time just so go top save it, and then if I refresh this all the way down. So you notice that we have a video component for the video I just provided. And obviously the the time that I did provide.

So it's interesting. So as you noticed, I built an entire page, wrote zero lines of code. The page looks pretty decent. It doesn't look so bad. There's a few things I need to do here and there. The components are there. It's just the props that are passed to it. Right at this moment, I spent all less than five minutes on building an actual homepage. And then so I was like, hmm, if I can do that and again for all of this in terms of a page, you can create language versions for this, if you like, you can modify a lot of things. You could build sub pages. So, like, I can create a child page. I can do a lot of things in here. So I went crazy and I built a lot of stuff again. I just wanted to prove out, prove out the thing in my head as to what exactly I wanted to do. And then I also went ahead and, like, you know what I might be able to build the entire page. We have about 500 pages or so, like, oh, 720 pages on the site with all blogs and stuff like that. I can use the site map, or just purely based on the navigation of the site. It'll pull with the header and for the nav, anything that's as a nav type. And I can actually import any of these pages if I wanted to, in a similar structure, right? So it gave us a lot of flexibility in terms of me trying to run this idea through in my brain 100% without me regretting it. So I did. I was able to do that. Let me stop that for a second screen and then go back to the presentation. That's a lot. Sorry. I know it's a lot of information, but essentially what we got is, as you as you saw, we gave it a brand palette, we gave typography tone, some sites which might influence. We kind of gave architectural guidance, coding guidance, and what AI was able to provide and in the background, and I'm sure there'll be questions coming through, but I'm trying to answer all before they come through. It can connect to any AI via API, so like open AI, it can do Gemini for Google. I'm using anthropic API endpoint to build this whole thing and rerun imports multiple times. I basically use $3.60 62 cents, which isn't a lot, if you really think of it, right? So every one of these runs when I'm doing a page and I do a bunch of things, about six to eight cents is what I'm spending on every single one of those, which isn't a lot, to be quite honest, in an enterprise space, anyways. So what we noticed is AI produced the page structure. It can look at a hierarchical site map and create a hierarchical page. It generated content apps. It created front end components, it structured them in a JSON. On output. So a headless CMS can headless front end, can pick it up, right? And it was able to generate components, not just one page. You can add individual components to it. For me, what worked is I got boilerplate components right? So I the library itself had to be pre built for here are all the components I want from chat, CN blocks here built take all of the variants for this component. So for instance, tomorrow we come back and say, I want a Twitter feed, or x feed, or whatever it's called, right now, it might not be there, so that's something which would have to be built. So if it gets built, I would have to then take it and deploy it on the front end website in order for that component to show up. But we kind of pretty much captured most of the components I would use for a site like ours.

It could do a first draft layout. It was pretty quick, as you saw, about less than a minute. You got a homepage, right? It was kind of cool for a marketing person to be able to see a site really, really fast, and they could play with it. They could be like, Okay, I see this site, but can you do this? Can you do that? Can happen? It did modeling. So what was interesting to me in this whole exercise was the content models it generated were pretty decent. They were very close to what I would have generated if it had all of the types and you kind of trained it, I'm sure you would be able to do very close to what you would expect a junior developer to do for you. He was able to do repetitive patterns. It actually could run migration. I've run it about for 50 pages, and it actually does a pretty decent job. And I feel like it takes care of about 20 to 30% of the work that you need to do for a brand new website built, is the way I feel. But this is mostly working on an existing website, right? So you're Yes, yeah, right. You're migrating an existing website as is. So if you've got a brand new website, or new requirements and stuff, yeah, new requirements, or your your I mean, one of the things we do a lot of is, when we do migrations, is it's not a like for like migration, like, yeah. 95% of times we work on migrations, it's not unlike for like, yeah, migrations, because enterprise is complex. Enterprise is built up over a long time, and you do end up with a lot of debt, right? Content, debt, a lot of content that no longer works. We've seen this a lot, obviously, with the new world of AI and geo used to work really well, or kind of well for SEO, might not work exactly the same for geo, and there's a lot of dead content. It's just out of date. It's not up to it's no longer correct in enterprise spaces. So that content cleanup is also a big part of migrations, right? Yeah, yeah, for sure. For sure makes sense.

So being in the CMS for this long, as soon as I did this, I spent so much time, added so many things. I was trying to do a lot of stuff. The first thing I was really pissed off at was shareable content, right? Like, we're all used to having defining news items once using it multiple places or reusing content. Can we build it into it? Yeah, you can, right? Like, I could work more and make it more shareable, as in, I build a hero once I can use it in 15 places, could be possible, but then you have to get into a little bit more of a complexity there. So reusability was, was a bit touchy. All of the things that we do in terms of versioning or governance being able to add permissions. Because the thing is, these components that, at least the way I am doing it, they are so dynamic. There is no content type to restrict someone on right like so it's dynamic enough that governance on something like this would be painful, permissions would be painful, workflow would be painful. Then you're trying to think, put yourself in your client's mind, in the enterprise world, and you're like, Oh crap, this is a bit more complex than I thought it would be. Like, in my mind, everything was rosy. Oh yeah, I can generate it and all of that fun stuff, but the enterprise reality started kicking in. I'm like, Oh, this might work for us, right, like for our side, but what about the customer side? I don't see this working for the customer that side. So that was kind of a difficult day. To deal with reality. I think as soon as you she was showing us, and we're like, okay, but what about like, yeah, workflow, right? How do you set up? Okay, well, we don't have workflow because it's just us, right, yeah, what about permissions? Like, you'd want to restrict it down to certain types of roles, and you're like, well, I could do that, but again, it's just us at this point. So, yeah, I really, really care. And then, of course, you've got other features like you and I can work again, like going back to those markdown files, we were like, Yeah, it's fine. It's markdown. It was great by sort of like GitHub, you know, if we need to do a version, look at what, you know, what changed over time. We'll just go back with a Git history. Yeah, no, it's not, it's not that we have those concepts, and then it's like, okay, but somebody wants to visually add components and look at what they're doing. And, you know, we didn't have any of those either? Yeah, yeah, no, no, for sure. I totally get that. And then again, it comes back to like for us, we constantly get asked, right when we're working with with customer everyone in the world is using AI, how much can you cut down the cost of this by purely using AI? And our answer is very, very simple, like we always tell them, Look, I know everyone's using AI. We use AI too, but we use it with supervision. We use it with certain checks and balances in play. And what we've seen is maybe about 30% of of the project can be, you know, done by AI or aided by AI, is the best way I can put it like, even for this, like every output that was generated was run through by one of our front end, senior front end developers who gave feedback, put it back. Rebuild gave feedback, put it back. So that line was already running when I took the coding standards from that line of the project that we have into this. And I feel like there's still place for it. It'll still help you. Like modeling, scaffolding, any kind of migration, first draft generation, right? I think there, there are some real time savings for sure. We have seen it. We do use it. But just wanted to mention, along with all of this, I, however, do not see how like knowing the projects that we've all built how it can replace governance. It can aid governance totally get that can't take over governance at this moment, compliance, I wouldn't give 100% flexibility to AI to do compliance, aid compliance, absolutely. Roles, workflow, integrations into your systems, which are critical, data integrity, parts security, I think those would still need to be managed, and would still probably still fall under the CMS realm, to be quite honest. That's my opinion, anyways, but just wanted to mention that, and I know part of this was cameras like you got a lot of questions on every single slide. Look, I have a lot of questions.

We will have a lot of questions, right? But I think what we again, what we've been seeing with our use of AI, is it's, it's not, it's not chatgpt, right? In enterprise applications, you can't just chat GPT and type in a random bunch of words. It has one of the things that we've and this is a difficulty right with with use of AI, is repeatability, and this is what we keep coming back to, which is why we've had. So when we come we say it's good for about 30% it's because of that repeatability. So if I give it the same prompt, is it going to give me the same answer two times, five times, 10 times, 20 times the last thing you want is it giving you a different answer, and it giving me a different answer to giving another developer or another team member, a different answer, and then we're ending up with conflicts. You're ending up with issues, and you're ending up with a lot of AI based technical debt, which nobody knows how it's got in there over time. And we've been even without AI, the kind of technical debt that can build up, the content debt that I mentioned earlier as well. That can build up over time as well. So having those guardrails, having that exactly that kind of senior team members, you know, treating it as a junior developer essentially, really does add the add the value here, right? Yeah, no. For sure, so in our mind, like, where does that  leave the CMS? We still need the CMS. We we do need content governance. We need a structured way of storing data. We need a workflow engine. We need permissions, and I feel like it's more of the things that we need to control our own destiny in terms of content and all of the operations around it, rather than the UI. I feel like more of the UI is not as like you know, five years ago, page editors and focus was a lot on, here's how you can edit the page. Here's how you can add components to the page. I feel like there's there'll be less of that from a CMS perspective, and it'll be more on how you control the content. Is my opinion, anyway, so cameras, I don't know if you have any any opinions on those. Yeah, I think it's going to depend on the types of customers. Obviously, I think it's going to a it's going to AI is going to aid a lot with migrations. As you have said, we've seen that with a lot of customers, enterprise customers, particularly, they end up with four or five, six different CMSs, and they always try to consolidate, but it's a big effort. So I think that those types of situations, this AI, is going to really, really help. And CMS again, they just say it becomes a system of record, and they can maybe cut it down to a single source of truth rather than five. But, you know, on from that, obviously, just basic CMS usage. I think when we get into the realm of personalization, AI is really going to, really going to kickstart things and really going to make things more, take it to the next level, if you like, right? Yeah, all right. So as we mentioned, and again, I know we're kind of repeating this, but we're trying to drive at the point. It is a modeling assistant. It is a component generator. It can be a Migration Assistant, slash accelerator, right? Really, really good. The other thing I found is when it's generating things for SEO, AEO, Geo, I really found it being very useful to generate all of the content that is needed, not just for the meta information, but for the FAQs, anything that we need to be able to check if the page meets The Standards right? Does it have everything that I need you could run tests to score against, not just the lighthouse scores, but also your SEO and geo scores.

A lot of things can be aided by AI at this moment, and I feel like that's one less thing you need to do so you can concentrate on what matters to you most, which is reaching your customers the way you want, measuring the performances of the content that you're putting out, and it more in that way than worrying about these tiny little things. Did you have something? Cameras? No, no, no, I'll ask for that at the end. No, no, no worries. And then also, like we feel from an enterprise perspective, developers are not gone. They're still there. The dependency changes, right? You need a lot fewer of the builders, but you need more of the architects, more of the people who validate, more of the people who understand the architecture across your company, across your applications, and are able to make decisions and create guidance. But do you need 15 developers? Probably not. You need a fewer amount of developers, because everything your touch touching nowadays, whether it's a CMS, whether it is a source control repository, CICD pipeline, everything has a flavor of AI built into it, where it's making your life easy, including something like Grammarly, right? Like things are all integrated into this. So you're finding efficiencies everywhere. So if you do find a really good set of developers, for instance, they're able to multiply the amount of work they can do just because of the tools they have access to. What does it mean for agencies, right? So it's a it's been difficult, like as an agency, to see the shift and the impact that AI has taken. And recently, I think the podcast is coming out next month with Marcelo, but I wrote an article where.

Was just baffled how certain divisions and one of the customers did just building integrations with a CMS system and their ERP, completely using, completely using AI with one developer for an enterprise the size, I mean, they have one point something billion in revenue. You wouldn't expect that, and that was the scary part to me, is okay now companies are deciding we don't need developers anymore. We don't need agencies coming in and helping anymore, and we have seen a hit in terms of that, of how companies are perceiving AI to be used. But at the same time, when I look at it, I'm like, really, like, you're gonna let everyone wild in your company, whether they are a data analyst or whether they're a marketing specialist, to build sites on their own, if they're brochureware sites, pretty much static sites. I don't see it is what it is. It's not going to affect your company. But if they're building integrations where it will affect your main systems, I mean, that's a I wouldn't get that access to AI any day, right? So there needs to be, from an agency perspective, you can come in and help. In terms of governance, architecture, design, risk management, aid in migration, if that's something that we need to do, and just overall, knowing the architecture and giving the guidance accordingly of where what can be done, setting up that repeatability right with AI, we mentioned that there is quite a lot of templating to do. So obviously, you'd gone through those fields or you put in. This is, this is a theme. This is a tone of voice. This is a brand kit. This is the coding standards. This is, you know, this is what I want to use this and what I want you to avoid. There's a lot of templating that's going on behind that to make sure that the AI is using those prompts, those commands, those instructions every single time. There's a lot of refinement as well that's gone on like you mentioned, you you do something, you cut it out, you do something else. Again, that's experience, because you know what to look out for. But if you blindly taking AI is wide for it, of course, you end up with a lot of those hallucinations. Then when you point it out, of course you're right, yeah.

Cool. And then how does this affect the CMS vendors? Again, I think the focus is more shifted from a perspective of, like, all of the trinkets, I guess. Ai, everyone's like, we do this in air. We do that in air. And there's some of them are really powerful, like we were mentioning site career. AI, is pretty interesting, how they rolled out the agents for the marketers, developers, marketplace, but there's still, I feel like, from a vendor perspective, their focus should still be in terms of governance, workflow, integrations, compliance, giving all the tools necessary for the marketers and The developers to do what they need to in a much more seamless way, making things faster to migrate scaling. I don't think the focus should be on like, the front end the focus, I don't think the focus should be on like, here's a new page builder might be useful. I mean, I'm sure marketers still use it. But I almost feel like, if the structured content is all set up, the UI part is something that you can easily leverage AI and do and say, Okay, you got what you got. This is what it is. Go generate me three options for this hero, right? And let the editor pick it, not worry about, like, I'll put this at this part kind of thing, you know? Yeah, yeah. So they, I think we've been seeing a lot of this, where a lot of the vendors are focusing on content rather than, rather than operations, right? And I want to say content operations, because it's a different thing. But a lot of the customers we are working with, they don't want generic content, right? They want and we keep, I know, we keep talking about brand kits and brand tone and brand voice, but AI does an okay job at it. But it's, it's still quite generic, right? Like you have, it's still, like, the average, you know, the average content you're getting, the mean content that you get you're getting out of the input output. And then we do work with a lot of customers where brand stories, you know, very specific stories. These are new stories. These are one off stories. These are stories that never happened before that you can't really draw in, draw inspiration from other content on the site.

To tell these stories so and people are able to tell a lot these days where it's very generic, AI generated content, and you know, companies don't want that generic, genericness. Yeah, cool. And then again, migration in the AI world is become a lot easier. I would say, like it can crawl sites with a site map, or just crawling real links. It can map a structure. It can generate models, components. So the first like time to first draft is done much a lot with the use of AI, of course, making sure it doesn't hallucinate, making sure you give it checks and balances. Balances definitely makes sense. But this changes the migration and economics, for sure, it's it's made our life really easy, and then redefining the CMS in the 2026, world, again, our take is cms becomes the actual infrastructure to use that provides governance and all of that fun stuff. And AI becomes the interface, and it becomes, this is the shift that we see currently, or the way we feel, anyways, about the shift from a key takeaway perspective, yeah, I mean, AI removes a lot of the friction, right when you're implementing A CMS solution at an enterprise scale, the CMS value is mainly governance structure, anything that will help you build a better set, help you build better content, like cameras was saying. So yeah, I mean, those are and then the enterprises must rethink how digital platform economics work in the modern AI world, finally, do we need a CMS? Just to reiterate, I'm not anti CMS, yes, but not for the same reasons that we used to even like a year or two ago, right? If AI can generate pages in minutes, then CMS is no longer the builder of pages. We build content, and thus that content can then be used for AI to generate the pages. CMS becomes more of the governing body, right? The organizations that win will not be the ones with the best page builder. They'll be the ones with the best control over how AI builds those pages. So CMS is not dead. If it cannot justify its existence beyond like scaffolding and page templates, AI will expose that quickly, right? So we've seen how fast AI is going. I don't think the question is, is it CMS versus the AI? It's never a competition. It will never be a competition. We do need a CMS. It's so CMS is the infrastructure. AI is the builder. And that, to me, changes a lot of things in this, in this modern world. So with that, I wrap up. So I don't know cameras if we have, I know you have questions, but we'll get there's a few questions here, as we'll cover that questions or comments that people have posted, so we can also go through those. Yeah, so before we get to the questions, we do have a webinar coming up in a couple of weeks. Pretty interesting, a similar line of thought. So we have Ashwin, who's super, super smart, who we worked on a couple of deals with, with Sitecore, George, again, smart. They bring their we're doing it as a panel. And what if you know this the last website you build, and it's mainly manufacturing focus, because we were working on a manufacturing bid, and this topic came up, and we ended up getting into it and some really interesting thoughts. So if you know of any manufacturers, anyone who uses sitecore to build enterprise websites, please use the QR code. It will be streamed as well on LinkedIn and and YouTube. Let's go to our comments. Let's see so Natalia says interesting discussion, even in the AI era, CMS is still needed to organize, control and securely manage content at scale, while AI simply helps create and optimize that content more efficiently. Totally agree with you. Don't deny it makes absolute sense. We feel the same way too. So for sure, Zach, so the. Argument I have is, do you build this real time with AI? Totally get it. So, yeah, this is real time with AI, right? This is not cloud code running on a terminal, so I totally know the difference in in how it builds. So like our usual AI pipeline, we use Cloud code and others to from like a build perspective. This was mainly an experiment from a CMS perspective, how I can do it, so all the prompts and everything goes straight through the API to anthropic, and stuff gets generated so it does not cloud code by any means, different use case, right? So if it was just again, you and me building it, we probably wouldn't need this to begin with, like we did before. But, you know, we could also just run that from our own from our own computers, and, you know, run it with plot code in the background, yeah? If this was really for customer use, then that's not going to be possible. So it has to be done through a web interface, essentially, right? Yeah, yeah. But yeah, that just on, kind of related to that question. I just wanted to, kind of just ask you clarify, I guess, that the CMS you built, how many lines of code did you write to build the CMS itself, and how long did it take you? Because obviously we saw you generating the pages from the CNS. But how long take? How did you build the same How long did it take you to write the CMS? Yeah, so, so I had a set of set of requirements of what a CMS should be, right? So what I did is I took that it took about three to four hours to build the base CMS. So I use Cloud code for it, and then spent another six to eight hours refining it over a period of two to three days. So within three days, I had the CMS built. It was last week when I did that. And I don't know how many lines of code, because I don't need to know how many lines of code did it for me. But yeah, and I know you had another question regarding, why didn't I use an existing CMS was one of cameras. Is questions, why didn't you take there's lots of open source, right? We've used a number of those over the years. We even have customers right now that are using a number of open source CMSs, and they do a very good job, right? Yeah. So what I found this several really good open source CMS is that you can host yourself like strapi or WordPress, or whatever, right? What I found was it would be like me tacking things on and fighting with the admin UI to get what I wanted, as opposed to this way when I rewrote everything, where I controlled everything, every aspect of the admin UI, every aspect of how it gets generated, the way the components are stored, structured back end in the database and stuff. And that was the predominant reason, is not being able to control the admin UIs in those open source CMSs, as I would want to. But that was you would have cut down the amount of time, I think if I spend time on one of the CMS, but also, like, they probably have restrictions in terms of the type of data I can store. So I don't know, it was just a give and take, and that was the my take on it. I mean, it's, yeah, it is a give and take. And then, given that it only took three days to build this thing. It's like, do I spend three days having my code, code analyze an existing code base, yeah, and then they'll try and build something on top, or do I just start from scratch and have it go through, right? So, yeah, yeah. But yeah, I mean, three days is, obviously, is impressive, and this is why. But, you know, we obviously see the limitations. We know what we're looking for. We know what we want, so we can kind of push it and ask it and prompt it. But this is also one of the reasons that the saspocalypse, they keep talking about, yeah, right now, and in the AI world, right? It's, it's, it is possible to do this, whereas if you were to do this in the past, I mean, you can't estimate it, as you know, said, but for sure, to get to where you have got, got to right now without AI, it would have taken you weeks, if not months. Yep. So that comment is actually from Anton. So estimation is now quite tricky. You know how much time it'll take without AI? What about AI? 3010, so, yeah, that's the tricky part. And I think you aren't on you guys use AI quite a bit with your innate and all of that fun stuff. I think after you do like a project or two, you kind of get in. Uh, understanding of, yeah, here's how much percentage we could probably cover by, and that's what we're kind of doing. So that seems to be working. So Zach, again, yes, junior developer, it arguments you. It's still built projects exactly like I did before, just like a Password button, yep, code generation button, great for stubbing code. Yeah, you're right, right, like you have to, I think as you use it, you know where it's useful to you, the way it's useful to you, might be a slightly bit different the way it's useful to me. Because one, I might not be using the way you're using it. I might not be giving the prompts the way you're doing it, so it's a tiny bit different, but I think you kind of get used to it. And again. So Zach is a super smart guy. I've worked with Zach, what, 13 some odd years ago at NTT and revere, he's on the job market looking for a job. So if anyone has a job, Zach's the guy for you is a Sitecore architect, super smart, like I mentioned, but he had been posting stuff, and as he says, the job market is strange right now. Seems like there are more junior developer openings and architects and leads. Feels like companies are walking the wrong Yeah, I totally agree with you not to put junior developers under the bus, but I feel like couple of senior developers slash architects, managing a couple of unitives, but also using AI to augment is more the direction which I would take personally, but totally agree with you. Zach, there's a lot of people out there looking for jobs, and it's just a bit frustrating. Totally get that. I hope you find something really, really soon, and then we have a bunch of others. Sorry, I'm trying to go through all the comments again. I think this is Anton, just a thought about the future, probably the whole components idea is not more relevant. With CMS and AI, you need components that are repeatable blocks that allow you to cut costs using uniform elements. Now adding new components and component variations is cheap. Each of your pages might be unique, but not be just a set of components. Totally get that yet makes sense. Once you build up that component library, very rarely do you find, oh, I need to add another variant. And even if you do it super easy to do that. Yeah, totally agree with you on that. Adipo wrote, I think that the valid mentioned that your AI account is learning how we do things and likes things, what it what is different from someone trying to start something from scratch without any developer knowledge? Totally. Yeah, I agree with you on that adipo, as you're doing more stuff, the context is there. It understands it better. You understand it better of how you work with AI. So totally get that. I'm just trying to see who this other and done again, let me run this one. Do you have a feeling that sometimes CMS might slow down the work with AI? You get an extra complexity that doesn't allow you to move as fast? Yeah, I think depending on the use case, absolutely right? Like, we'll run into cases like this all the time. We're like, I wish I could do this, but I can. So you have to find a way around it. And I don't think that's specific to ACMs. Any system will do that. You're trying to do it with an ERP. Like, damn, they don't have an API for me to push. I have to pull a JSON, I have to sing and dance, and then it will do what I needed to do. Part of it life, I think, but part of enterprise, life, right? Then, choice, yeah, and then the last one, I think this is the last comment. Let me just make sure. Sorry, one second.

Yep, that is the last comment. All right, so this one is from Nick thinking outside the box here, but I don't see any reason why the CMS can build itself on demand to achieve the requirements it has been asked to complete. Totally right? Well, you did, you did kind of audio CMS on demand, right? Retrofitting this for existing CMS is, can they do, if I can do it, they can do it so any of the CMS vendors can absolutely integrate this concept. And like I mentioned, this was just a POC. There is no intent of putting this to sell as a service or anything like it was just mainly an experiment for us to kind of push the limits and see a concept come to life and generate all the questions and look this, this webinar, generated that for us, which is we have. Questions we had to implement to understand the answers. And then now we have more questions, but we kind of understand the next direction, so  totally get that. I really am looking forward to this year and seeing what innovation any of these CMS vendors come along with. If you have any other questions, please post it on YouTube or LinkedIn. I believe the YouTube framing wasn't right in how it was streamed, so I have to double check that. But other than that, have a wonderful day. Thank you so much for watching this webinar. And yeah, let us know if you have any questions and join the next one in in march right just a few weeks away, sign up. Thanks, everyone. Bye.