We Rebuilt Our Website With AI

Speaker A: Welcome to Alpha Marketing, your go to resource for unlocking the full potential of your small to medium sized business. Hosted by Aaron Shashila, CEO of wolfpack Advising. Now let's learn how to dominate, um, your market

Speaker B: and welcome back to Marketing Alpha. Today I have a real fun one and it's been a while since I've recorded one of these podcasts, so I'm happy that you're back. And the fun one is we tell clients not to build their websites with AI. But if you're reading the title of this podcast, you know that we rebuilt our website entirely with AI. Funny enough too. I made a Facebook post that said if you built your website with AI, it's gonna likely crash and burn. So I wanted to take some time to talk about why wolfpack built our website with AI and uh, yeah, talk about some real numbers, talk about what it cost, how much time it took, uh, what we did wrong, what ultimately I did wrong, uh, what I learned from the process and ultimately the one thing that made a huge difference to ensure that the site doesn't crash and burn. So let's get into it. First I want to just talk about like why we decided to go this particular route. Because we didn't just decide to make a website and like redo our website with artificial intelligence or AI, just, just because there was a specific reason as to why. And that is really just having to do with the ceiling of the old setup. The old setup setup was a page builder. It was a WordPress website and it was designed uh, through a page builder. If you're familiar with WordPress, you probably know about themes, you probably know about plugins, and that's ultimately what we use. Themes and plugins, the themes and plugins that we use. The one specifically was WPBakery page builder and another one you might be familiar with within the spaces like Elementor or Divi. It's basically just a fancy visual builder. What that does is it like it you can really visualize and edit the site on the front end and it's like a drag and drop builder. Uh, and it's really nice. It's like WIX or uh, you know, Squarespace, where you can just drag and drop things and it makes it super easy for anybody that you know, doesn't know code to go ahead and edit the website. Which is nice because as an agency, right, we have a lot of people working on our website, but it doesn't scale. Uh, every industry page, every service page, every case study was a manual build. Every single time we had to go and copy a page and everything was a page. Um, and I'll get to what that means here in a second. But everything was a single page and we had to copy that page and then we had to redesign the whole thing with the new content and import it in. And so it took a lot of effort because we had to redesign as well as do the copy for every single time we needed to adjust a service, add a new industry, whatever else. It just didn't scale. It just didn't scale. Um, so for our strategy, uh, in terms of adding new services, adding new industries, scaling our website a little bit further, we um, needed to have like a deep structure to where we have custom post types and custom fields. So what that does is now instead of just pages and everything being under pages, now we have a customer type of page for just services or just case studies and just industries. And within those custom pages we have custom fields. So now the design is pre built, I won't get into that, but it's pre built. And so if you're going to, if you're going to add a new service, if you're going to add a new industry page, the only thing that our team members have to do is create a new service page or whatever it is, custom post type and add in all the custom fields and the design does the rest, the code does the rest. So ultimately the design lived inside the builder and we couldn't even version off of it. So I couldn't even use like artificial intelligence or anything like that to like edit it. Um, because Claw Chat, GPT, um, anything like that, like you can't really edit those kinds of page builders. Uh, and it was just very difficult um, to push any type of content. So ultimately um, we had to decide to make a transition to say, okay, if we're going to scale this website, we're going to add more industry pages and get like a real content engine underneath and not have to worry about designers and being logged or backlogged with designers all the time. Then we need to redo this architecture. So uh, since last year for, um, yeah really since last year we've been looking at rebuilding the site and we've attempted it many times. I've attempted it many times. Uh, there was times where I was trying different content management systems. So outside of WordPress, I was looking at another one called Payload CMS, um, and I even looked at Cloudflare, um, produced and launched their own content management system that was supposed to be, you know, WordPress killer. And everybody keeps on talking about these WordPress killers. Uh, but quite frankly, um, after many months of testing different ones, um, wasting quite frankly a lot of time, came uh, to the conclusion that uh, it was complete waste. Um, you know, I'm sure that uh, payload, CMS and you know Cloudflare has really good systems and everything else, but the documentation, the support around it and everything else just wasn't up to par. And not only that, but I have to retrain our entire team on how to use those content management systems compared to just using WordPress and knowing WordPress, WordPress has been around forever. There's tons of documentation, there's ton of support, there's ton of community and every single like pretty much any digital marketing professional at Some point used WordPress. It's kind of like you know, WIX or Squarespace. If you're used to that ecosystem, that's what you're used to, that's what you know. And so it makes it super easy. So ultimately for that time, for like basically last year, November all the way through up till like July of this year, I was just testing, I had um, tons of different times that I was trying and failing and trying and failing and just not really getting anywhere. Um, and it wasn't until July 2026 that we committed with a particular direction and that was using WordPress. Um, we used WordPress but we did not use a, ah, WordPress theme or standard plugins. What we did is we uh, we used a, what's called a headless WordPress system. So with uh, headless WordPress you basically use WordPress as your database. So you use WordPress to actually edit the content. Um, and that's where I'm talking about the custom fields and the actual front end of the site, the where it deploys and the design of it is actually just JavaScript and that's hosted somewhere else. So that front end just pulls from the database which is WordPress and hosted somewhere else on this backend. And what that gives us is that that model I was talking about earlier where I can create custom post types for you know, service pages, industry pages, uh, industries, um, and case studies, whatever else. Uh, so I was able to create those different custom post types, I was create, able to create custom uh, what we call taxonomies on uh, the back end of WordPress, uh, which gives us more flexibility and of course all these different custom fields apply applying to different custom post types which just makes it super easy. So at the end of it, by July our team was um, starting to design everything and um, we Actually right before we decided I was, I was still leaning in the direction of using WordPress and the traditional setup but using Elementor. But again like the design was just taking the too much time, too much time. Uh, and, and this year in 2026 it's, it started last year, November. But November, November of 2025 was when you know, really Claude uh, code really took off and so you could really scale a lot of stuff. And so when we talk about 2026 this is a first full year that people are able to really iterate super quickly using Claude code. And I quickly realized that um, their traditional design time just was not going to cut it for the type of speed we wanted to get this done because we had already like, like it was nearly a year of working on this particular project. So um, even going the standard elementor route with page builders and everything else, we backtracked from um, we backtracked from that uh, and ultimately we went in that headless WordPress setup and the um, big quite frankly the biggest unlock on why we were able to go in that direction is Claude design got super solid. If you're not familiar with cloud design, if you have a cloud subscription, they offer Claude design. It's Claude AI forward slash design. Uh, and it's a real awesome system today. Uh, back when it first started it wasn't great. Uh, it was like a ChatGPT version or a cloud version of just designing things and it would just make the design and then when you would ask it to edit, couldn't edit it, it couldn't do anything. But now today, presently, uh, you can ask it to start something, it can start it and then it has a full blown editor, full blown editor. So you can really get the refinement on the padding, the sizing of elements. Um, change it however which way you want just like you would in a traditional like Adobe editor or like in Figma if you're actually a designer, you know about Figma or Canva for example. Canva is one of those ones too that some people might, might use and know about. But quite design got super, super solid. So this transitioned us into a setup where we have a design system in cloud design. So it has our entire brand kit. Um, we started a folder for just this project and we're able to create one design per each page. So one design for a service page, one design for an industry page, one design for you know, a case study page, um, and continue on. Then we take the quad design and when you go to share and export you can Go to like other options and there's a way to share it with CLAUDE code. So basically we would take that cloud design after iterating it on a few times, working with our designers, getting it just right, then you take it, you export it and you give it to CLAUDE code and CLAUDE code is able to deploy it. So once we had the infrastructure set up, we were able to move very, very quickly. So by September 13th of this year, this was when we were fully committed to this type of stack. Um, where July we were kind of trying out Elementor and doing the page builder. And then we did that for a month. And then August we were doing this transitionary period and still testing out cloud design in this new setup. But by September 13th, midday, um, I pretty much had it set up to where we had a JavaScript, it's uh, using Astro Astro, um, front end on Cloudflare and the back end was WordPress. And I had that set up through Kinsta, uh, which is a nice hosting place for, uh, WordPress websites. That's what we use as an agency. And I had a cloud design folder set up to where we had the brand system all in there. So the process became super quick. Ultimately this process totally won. Um, it was super, super quick once we had the design system and the infrastructure set up. Um, and it was super familiar because we still could use the page builder and designer. Like you could use it in cloud design. It's the same thing as like a page builder we use in cloud design. Then you export that, you give it to Claude code via the terminal. I was using the terminal a lot to communicate. Um, as long as you have it set up properly to be able to configure and work with WordPress as well as configure and work up with Cloudflare. Ah, and makes it super quick. You don't really need to know code. I mean, you do need to know code. You, you need to know what you're doing. I think we'll talk about that here in a second. About like, you still need to know what you're doing, but, um, you don't have to waste time writing code anymore is ultimately what it is. So given the speed and everything else, we were able to go full, like sprint mode. Um, September 12th was when things were like, the foundation was set. September 13th was really when we pushed, um, literally, I'm telling you, like within the next week or two, the whole thing was transferred over. Um, so by this week it's October 1st. Today, as of recording of this, which, uh, it might Go up the same day. Last, uh, week was when we deployed it. Uh, so it was, you know, the second to last week of September. So literally this, this process took us a week or two. So you're talking about a very large website, very complex website in a sense, uh, very nice design. Realistically, like, actually it took us like nearly a year to complete. Right? But once we have the design system set up and the infrastructure and we're able to choose on a particular direction, it took two weeks, less than two weeks to do this type of build, when normally this type of build would take three to six months. Um, traditionally, uh, especially a lot more time, a lot more people, a lot more effort. So it was, quite frankly, uh, a game changer, uh, for us to be able to launch this quickly. And now with the setup that we have, we can just have Claude or ChatGPT connect to our website to do any updates we need. We're not bottled down by these page builders that don't communicate at all with Claude Code or ChatGPT or any like that. You could just, boom, connect to the website and say, edit these pages or edit this design in this way and it's all done. So the way that we were able to actually work very quickly is, uh, what we did is we set up integrations with cloud code, uh, to be able to access the back end of the site as well as the front end of the site via cloudflare. We had that set up and then like I mentioned, we started in cloud design. We shared that page with cloud code, we gave it a cloud code and it was able to, boom, push to the back end, push to the front end, deploy it. And our team was able to jump in there on the back end and actually edit the content, refine the content. But quite frankly, we had a lot of the content already drafted up within like, basically Google Docs. It's, it's our project management system called ClickUp. We had it, uh, built out within ClickUp Docs, but it's basically Google Docs where we had every single page and all the content we knew we wanted. So I pretty much just take that, gave it to Claude code as well. Claude just talked to the back end and added it all in as well, which is why it was like a week or two and literally only a couple people were involved in this project. So how much was it? How much did this cost? Um, to be quite honest, um, the week that this site went out, one or two weeks, you're using a lot of usage and a lot of tokens within a very short time frame. And um, AI models don't like that today. Um, they try to space things out. They limit your usage over a certain period of time, especially if you're on a subscription. We're already on a subscription. We're paying over well over a thousand dollars a month for our subscription, for our agency. Um, and then usage credits just within one week. $1000 in usage credits just in the, in the one week. So realistically, you're talking about my seat to do this. The usage credits as well as all the extra time we've done for like deciding how we're going to set up this infrastructure and playing with payload, CMS, playing with, you know, Cloudflare's version of WordPress that is supposed to be the WordPress killer, and then ultimately deciding to go back right, when you're talking about this rough in, um, you're talking probably two, $3,000, uh, in just literally AI spend, uh, not counting and not considering the time and the expertise needed from my time, uh, you know, Jenna's time, who's in charge of all of our operations and reviewing and helping out, as well as our designer's time as well as our content people's time to be able to write out a little content. So that all in with just the tools, you're talking probably close to $3,000 plus everybody's time over this time period. So still, you know, five, at least easy, five, six, $7,000, uh, for the amount of cost all in for this type of site, if you were to talk about the normal way, oh my gosh, way more, more, um, money, not only time. So let's take away the, the factor of time because time is different. What we've learned here is yes, the factor of, uh, cost goes down because you're spending less time. Right. However, the cost still remains high because of the tokens you're using, ultimately because of the tokens that you're using. And you still need an expert's time. You still need an expert's time. Now, not only like you need an expert that knows design and knows, uh, let's say WordPress and knows content, SEO content and is able to work with that, but you also need an expert in AI uses or use cases. There's a lot of people today, most people today are not able to use AI like wolfpack is today and like we are able to use it today. So that just adds another layer of expertise. When it costs, when you talk about the normal way, tens of thousands of dollars from another agency is what you'd get a quote for at least $10,000, $15,000 for the type of site. Um, it's a headless WordPress build. So you got JavaScript, you need a developer there, you need somebody that's able to work a backend infrastructure as well. Two different types of developers. You need a designer, um, at least one designer, likely two. Um, and then you also need SEO experts, you probably need some, some SEO specialists, at least two, as well as some type of SEO team lead that is actually running the show and kind of controlling everything from an SEO perspective. Um, I mean, even just let's. I'll hit one particular item, um, that most people forget and why I mentioned if you, if you use AI to build your website, it's likely going to crash and burn is the URL migration. We had 600 URLs that we needed to redirect. 600. If you were to do that the normal way, yeah, you can get fancy with spreadsheets to make things super quick. You can know, export it into a spreadsheet and, and say like, hey, redirect this URL to that URL, uh, and then import it into some type of plugin within WordPress or, or give it to developer to work through. But, um, if you're just talking about that in the redirecting that you have to do from old URL to new URL, that in of itself is a huge pain in the butt, um, and could take a week for one person to just work through. Um, but with Claude code, it's one prompt, two prompts, that's it. That's all it took. So, um, ultimately the real cost is still the experts. The real cost is being able to know how to use AI for connecting to the backend securely. You know, mind you, we're not even pasting our secrets, uh, meaning our API secrets, you know, tokens, um, access keys, things like that. You don't paste that into cloud code. You have secure ways of sharing passwords with Claude, so that way it doesn't appear in the actual chat history and it's encrypted. So you still need the experts. Um, you still need the content team to write the SEO, uh, content. You still need a designer that knows proper design. Uh, you still need somebody to lead the project. And you can't forget the migration part and talking about migrating old content, old URLs to the new piece. So honest take it might appear as like it could be super cheap in dollars, um, which theoretically, if it's a small site. Yeah, you don't need to spend as much um, in cloud credits as we did, especially if you're saving a lot of money when it comes to the design. Because we're using cloud design to help us iterate very quickly. Um, and if it's a small site, you can do things pretty quickly. But when you start talking about getting things customized and the branding piece and everything else and really making it not look like a Claude or a code or an AI generated site, you want some customization, um, to it and real expertise. So what did we learn? Well, AI made it super painless to start over. Once we committed on something, AI made it super painless. And it was two weeks, so that's super nice. When we had a commitment, the sprint happened and we were able to move so quickly. Second is design first, Design first. And we already knew this, right? But you have to go in with a brand system first, like a brand kit or something. And you need to design first before you just go straight to the coding. Before you go straight to like using Claude code to like actually create something, you need to know what your color scheme is and everything else. Then you need to know like how you're going to be setting up the site too. So you need to know, hey, am I going to list my services? What about my service area? What about testimonials or anything like that? You kind of need to go in knowing how you want this website to look in the infrastructure. Now the other thing that you need ultimately that a lot of people forget is the copy and the images. That is ultimately why I've said if you built your website through AI and it's an AI generated website, it is likely to crash and burn. Uh, and is because of the copy and the content as well as the migration. If your copy is all AI generated and you don't have any expertise or personalization into it, it is likely to get flagged by Google. They've done tons of spam updates as of late. And it is likely that you crash a lot of your rankings and your visibility that way. You still need personalized content that's uh, like highlight your expertise, highlight your authority in every way. And you absolutely need custom images as well. Real images, not AI generated images, not stock images, real images of you on the field, whatever it is, your people in the office, real images. It uh, goes way farther than the, the generic route. Now the final thing is the migration. If you go to just redo your website with AI and you don't think about migrating your old URLs and pointing them to the right place. So if you have like a hundred blog posts, let's say, or Even not even 110 blog posts, but you don't redirect those old URLs to the new place, you are going to lose all that traffic. Same thing with your service pages and anything else, any other page, every single page needs to have a redirect from here's the old place to the new place. When you do this migration before you do it. So that way when Google crawls it, when they're still showing your old links within Google search results, that doesn't immediately fall before and you don't lose all your traffic and all your rankings. So ultimately why I cringe when I hear clients rebuild their sites with AI and why I posted, it's because of those things. The content still matters, the images still matter, the redirection and the expertise still matter. AI still heavily says, like, you need this type of schema markup, which is totally not true. Yeah, schema helps. And if you hear, you know any, any AI model, say, like, put this schema in your website. Yeah, sure it helps, but that's not a game changer for SEO. It's not a game changer for AI visibility. Schema markup is just structuring your content in a way that makes it easier for AI or any type of crawler like Google to understand your content. But it doesn't change the content. It doesn't make your content awesome, it doesn't make people want to read it. What you need is true, authentic, real content that shares your expertise. Like I mentioned before, the generic copy, the generic images, the stock photos, everything else is what is limiting with AI, uh, generated websites. The templated design may look fine, but really ultimately can fall apart when you try to scale it. And how everybody's site starts to look the same. So the uncomfortable truth in all this is AI can produce really clean websites and make things super quick. But what I would say is if you're doing it like by yourself, I'm all for it. If you want to do it by yourself and maybe, you know, you don't quite have the budget and you want to spend the time to do it yourself, uh, especially if you don't already have a website, but when you're talking about you already have a website and you're ready to upgrade or something else, the digital marketing expertise that you need hasn't, hasn't changed, hasn't changed. It comes down to the expertise. If you're bad at copy, you're not going to be able to steer the AI, especially if you're bad at like understanding the SEO and the type of content needed, it's not going to fix that. And the same with design, uh, AI can go in with some design best practices, but really understanding conversion rate optimization to steer AI in the right place and marrying the two together is where the expertise comes from. Now bottom line here is I still love AI and I still would encourage anybody, uh, to utilize it within their own company in every which way. We are trying to push it as much as we can to our clients and embed it within our tools as well. I am just using this as a word of caution to not try and think that you're going to be able to use AI to just magically poof and migrate your website over. I want you to be cautious about it because of a lot of the issues that it comes with. So if you're going to do it anyway, lock the design first in the stack. I really still recommend sticking with WordPress and not trying to go with a hard coded like fully coded custom site just because WordPress is very well document documented. So lock the direction, the stack first. Even if you're going to go just WordPress and you're not going to go the headless version like we do, you, you could still do that, get a real design system or a brand kit. So understand like what you want people to feel like when they land on your website. What is your brand, what is your color scheme, but what do you want them to feel when they land on your website? Once you have that, build some type of like sample use cloud design. Play with cloud design. There's other ones out there like lovable that I know people use. I'm not a fan of lovable. You know, might as well just go to cloud design. But start with the design first and really hone that in once you've decided on your stack. Now finally, it's just make sure you have like if you're migrating from an old site to this one, make sure you have just some expert that is able to help you with security issues, um, that understands the SEO side, um, and make sure you add in your own personalized expertise in your own images, even videos. Ultimately, last year we tried for nearly a year to generate our website with AI and by the time we actually committed, it only took two weeks and it took in a hard cost within those two weeks. 3,000, roughly 3,000, um, dollars when you factor in everything, $3,000 worth of just AI spend, not including time. So the takeaway I just say is AI is another tool. It's another tool within our bucket that you can use and it changes how fast you can actually work. Just going from something that took or, uh, usually would take three to six months to do to how we actually were able to do it within two weeks is massive. The final thing that I'll mention is because we were able to do this and now we're able to actually scale our content correctly and making sure that it's personalized. We got an email today. Our website's been out for less than a week. Got an email today, um, from our SEO software, uh, that basically says that how much our AI visibility has increased in the last week, um, and has increased significantly across all platforms. Anyway, if you love this podcast, if this was helpful for you, leave us a rating. If you have questions, reach out, leave a comment. Whatever it is, reach out to me. You know where to contact us. And until, um, next time, Peace.

Speaker A: Thank you for listening to Alpha Marketing. If you enjoyed today's episode, be sure to subscribe, leave a review and share it with other business owners and marketers. Your support enables us to continue bringing you actionable advice and, um, the tools you need to thrive. Join us next time for more insights with Aaron Shoshella and keep striving to be the Alpha in your market.