Surfing the AI Tsunami

0:27Janne Parri: Okay Michael so you hold the top position on Webflow's community leaderboard so how is that even possible and what did it take to reach that position?

0:39Michael Wells: Yeah. So when I first started using Webflow, I had so many questions and I ended up relying on the community. Were, there were a couple of phenomenal leaders there at the time that, really sort of inspired this question and answer process to help me get started and help me overcome certain challenges. And at some point I transitioned into helping answer other people's questions. That started about. Six or seven years ago and became like a daily thing. Every day I get up in the morning, brew my coffee or matcha or whatever and sit down and see how the community is doing. Who's having problems? What's happening? Where are they struggling? The reason it became such a preoccupation for me is that I realized that I get so much out of learning where people are having problems with the Webflow platform.

1:34Michael Wells: And the reason is for me is that my background is all technical. It's all about problem solving. So if I know where the gaps are and where the confusion is and where the struggles are, it tells me what I need to be focusing on for my clients and the products that I build. So it's been almost like a market research kind of experience and just so worth the time.

1:57Janne Parri: That is beautiful and there are so many people that you have helped with your posts and your answers. But I have to ask, how do you like the new community thing, Webflow just revamped the community or the forum. How do you like the new version of that?

2:16Michael Wells: The thing that I really love about the team building it is that every little detail I'm like, Hey, could this button exists and do this thing? They're on top of it. So I think everyone's super familiar with the discourse forum and the Reddit forum. The new form has a lot of really nice capabilities that I'm just starting to figure out. Like, I run the, I'm the global leader for the new Zealand, the Auckland, new Zealand area. So they've got an event. Feature that you can just put up an event anytime I want to run something. It's done extremely well. They're really pulling it together. I think it's going to be far superior to what we've had in the past.

2:56Janne Parri: Yeah, yeah. ⁓ And I believe Webflow has some quite big plans for it. So that is also exciting. ⁓ But okay, before we go any further, we have to do little intros. So my name is Janne Parri. I will be your host today. And I'm also a general Webflow dude all the way from Finland. ⁓ And my guest today is Michael Wells, the founder of Sygnal. Michael focuses on broadening Webflow's capabilities with... All sorts of solutions and has worked with huge companies and clients like Citibank and Verizon. As mentioned, Michael holds the pole position on the Webflow's all-time community leaderboard. So if you do not recognize his face, you might recognize the Sygnal logo, which appears ⁓ next to so many posts and solutions on the forum. But Michael

3:47Janne Parri: You have done this for a while. So Webflow is now a little over 10 years old, ⁓ but I believe Sygnal turns 25 or something. How did you end up choosing Webflow as your focus and what can you tell us from those early days?

4:02Michael Wells: Wow. So I started working with the internet before it was called the internet. My first web company started in 1996 and was all about building publishing systems because nothing existed then. So we had to build everything from scratch to build giant city oriented sites, tracking like every restaurant, every every event happening, every sports team, what's going on for them, that kind of thing. So it was a whole new unexplored era. And for the next decade, I kept developing those systems to build client sites, mostly corporate sites.

4:42Michael Wells: Sometimes smaller websites like an individual restaurant or pub or something like that. It was really interesting to see all of the challenges happening in that early decade where HTML is just starting to really standardize in some kind of consistent way. Browsers are getting more powerful and I hit a wall and the wall was design. Everything in my systems was incredibly powerful, but it had to be handwritten as HTML, handwritten as CSS, handwritten as JS. I could put anything anywhere I wanted. I could make changes instantaneously, but it was all code. And at some point it was like, my design skills are really limited with that approach. I can't drag and drop a box onto the page and style it the way I want. And then Webflow appeared.

5:29Michael Wells: And I was sold. Tried it out. Tried out the client facing editor, which was really important for me. It was very important for clients to be able to edit their own content. Once I released the site, tried out at the time, client billing version 1.0 was out. ⁓ That made things easier. I could just host each client individually and have them pay Webflow. That was sorted out. ⁓ It was kind of the ideal platform for where I was at. And once I started building, thought, I'll migrate one site. Let's see how it goes.

6:00Michael Wells: As soon as I had a taste of how much capability the designer experience gave me and how much freedom it gave me and seeing the site as it's being built, rather than edit code, refresh your browser, edit code, refresh your browser. There was no question. I had migrated the rest of 50 sites by the end of the year, and I've been there ever since.

6:22Janne Parri: Nice. That is awesome. I believe your career might be like older than me. Maybe not, but I was just starting to count the years backwards. Okay, that's kind of the history, but ⁓ what sort of projects are you doing like right now?

6:41Michael Wells: So about four years ago, mostly through the forum work I was doing, I realized that, Hey, there's a gap in the Webflow delivery experience for most designers where Webflow handles the design work, Webflow handles. Want to do interactions. You want to manage your assets. Want all the platform solves all of that. But in my mind, it's like having a really beautiful city that has all this great stuff, but it's surrounded by these really steep mountains. And when you hit those mountains, because you want to do something that's not built in, it's a real challenge. And this is the nature of all no code platforms is they've got this solid wall around them. And so.

7:27Michael Wells: Immediately, I found myself helping people in the forums who are like, hey, Webflow doesn't do this thing. Maybe it's a likes and favorites feature, or maybe it's user accounts before user accounts existed, or maybe it is something strange with extremely large media that you can't store in the Webflow CMS, or you want automatic watermarking, or you have weird security concerns. There's always something. And when you get to more professional sites, these problems become make it or break it situations for those agencies.

8:00Michael Wells: So I suddenly found myself helping agencies build solutions like reverse proxies, JavaScript libraries, external infrastructures, anything that could help make those things possible. Well, they still got to use Webflow as the base platform for all their design work. And that became kind of a preoccupation. I stopped building websites in the typical way. I was never a really good designer anyway. And that was great because now I could work with like world-class designers. Who really know how to make it look good and then I know how to make it work good and that's been so much fun.

8:36Janne Parri: Nice. I have to ask, where do you find the designers and all the other pieces of the team and how did you end up working with those enterprise clients

8:49Michael Wells: Yeah, it's interesting. I've never been good at marketing or sales. So I usually get invited. The approach that I've taken is I just enjoy working in the forums. I enjoy solving problems. That's what really gets me excited about my day. And then people would keep submitting problems that there really wasn't a solution for. And so I would I would come in and say, well, you know what? I've got an idea. I'd prototype something. We talk about it. Next thing I know, they want me to build this particular solution for them. A good example. After localization was released, there's a client that I have that is based in Europe and they have a site that is based in UK, France and in Germany. So three languages on their site.

9:35Michael Wells: But due to their history and the way they've always marketed, they always use local domains. So they had a .de domain, they had a .fr domain, .co.uk domain. And they still needed that. All their brochures, all their literature, all their marketing, everything, all their clients knew those domains. So the problem was Webflow doesn't do that, right? Your localized content is under a subdirectory. So they came to me and said, is there any way? Is there any way we can have Webflow, but also make it do this thing where it's actually publishing to three separate sites? So we built it. We built it. We call it locale specific domains and it's a reverse proxy that splits sites. And it was just fun. Was like, this is an interesting problem. Let's build it. And then the experience for me as always, these become products.

10:26Michael Wells: A month later, someone else called me needed exactly the same thing. I'm like, actually, I've already got this thing built. So yeah, my experience is I just keep building solutions and tools and they just find a home. They just find a home.

10:41Janne Parri: Yeah, yeah, that is beautiful. It's always great if you can like do the same thing again, like solve the same problem again. But you mentioned a couple of times reverse proxies. We should maybe visit that quickly. Like, what does it do? I know that Cloudflare is a big player in that and Webflow is in partnership with Cloudflare. So what does it do? What is possible? And should webflowers check in what they can do with proxies? Yeah.

11:14Michael Wells: Yeah, when it comes to site modification, right? Most people are familiar with JavaScript. Drop some JavaScript into the page. It moves elements around. Some things happen. That's extremely limited in terms of the capabilities. It depends on the browser and you're still delivering the full page. So it's not good for things like security as an example or any kind of content mediation. Reverse proxy is a separate server that sits between your audience, the internet, and Webflow.

11:45Michael Wells: So every request that comes in hits the reverse proxy, which you control or I control, whoever built it, right, controls that proxy. And that means you can deliver anything you want. A request comes in, you can deliver it from the Webflow site. Or if you've got a site merge set up, you might deliver your Webflow site. But then when they go to the slash blog directory, you're delivering your old WordPress blog from the WordPress site live. Okay. So a reverse proxy sort of fronts your website, but it gives you total control. You can change anything you want about the response that's given, how security is handled, what headers are used. If you go back to my analogy of the beautiful city with the mountains, reverse proxy, it's not even grappling hooks, it's an elevator. Like to get you wherever you want to go in those mountains, you can go.

12:41Janne Parri: Yeah, that is beautiful. Very, very interesting. I'm not like full blown ⁓ developer, but you have peaked my interest. Should definitely look into those, ⁓ but I know we are both like itching to get to the like future stuff, like future of Webflow all the ⁓ developments ⁓ that has happened like just in past couple of months and so on. So, where do you feel like Webflow is heading with all of the things that are going on?

13:15Michael Wells: Yeah, that's such a great question. I can tell you what I see. And what I see is mostly some conversations with the engineers. I'm very fortunate to be able to ping some of them occasionally or have conversations when they pass through Auckland. A lot of it is just the product direction, the way the platform's evolving. So if you think about, if you've been watching the forums for the past 10 years, you can see where the pressures are. Developers want more capability. They want to be able to build real applications with more than just JavaScript. They want server side capabilities. They want customization. Now there's always pressures from the design angle, things like new animation forms and new capabilities and new types of video. And more recently in the past three years, a lot of AI pressures.

14:05Michael Wells: I want an AI chatbot. I want an AI to help me with design. I want an AI to monitor my site and tell me what things need fixed. You know, I want all of these things. Because it time savings means money savings, right? And that means profit. So it makes a lot of sense and web flows super aware of these things. But it's a huge ship. So it takes it a while to turn. And it's quite funny to see people go like, we want these things, why isn't my flow listening? It's like, it's listening.

14:33Janne Parri: You

14:34Michael Wells: Definitely listening. Just be patient. So for example, for years people wanted more than 20 collection lists on a page. But the infrastructure they had to change. To allow for the next generation CMS that can do that was so massive and so central to the platform. But they've been working on it for ages and it's just now visible to us as clients. It's a lot of fun to see. But I can see Webflow responding to basically all of the pressures of its own community, the release of Webflow Cloud.

15:09Michael Wells: Code components are amazing. They're one of my favorite new features. Now we've got AI mixing into that with app Gen and all sorts of new capabilities coming into play. It's really exciting to see all these features because for me it's just a toy box. You give me a problem, like, I have all these cool new things that I can do. So I'm building a code component a week just for the fun of it, just because it's so freeing. To have those abilities now and make them intrinsic to the platform and drop it into the canvas and configure it. It's just so much more natural than build a reverse proxy and you can't. Those experiences are very powerful, but you can't see them in the designer. You can only see them

15:50Michael Wells: On the published site. So yeah.

15:53Janne Parri: Yeah, what sort of things are you building with the code components?

15:58Michael Wells: So I think of them in two categories. My focus of code components is always, I wanna build things that are reusable. So I'm not a big fan of building a code component for one client. I wanna solve a category of problems. So half of my code components are widgets. The same way you think of any other element, it has a job and it does a thing and it's visual. So one of the ones that's better known in the community is probably my audio player. You give it an MP3 file and it shows the waveform like you'd see on SoundCloud. And you've got this.

16:34Michael Wells: Beautiful little waveform and you can click on it and it will just start playing from that point and it shows you where it is somewhat styleable and it's got some intelligence like if I have 20 of them on a page representing an album or a series of podcasts or something and I start playing one it will pause all of the others automatically so that I have a cohesive experience so they talk to each other. In a certain way and this kind of thing. I find it really useful because the moment I need an audio player, I'm like I've got the perfect thing. It's styleable. It's already built. I only need to build it once. And I've got other

17:10Michael Wells: Components in that category like things usually things that are code heavy and painful like country code selector for next to a phone number field. You know, no one likes building these things and styling them. So I was like I'm building it once. And then I'm dropping it in and now it's solved and I can tell it and it works really nicely as I start typing it automatically collapses the list and but I don't have to think about it again. It's a solved problem. Then there's another category of code components which is. They're kind of I think of them as enablers. They don't really have a user interface in and of themselves. They affect what you put inside of them.

17:50Michael Wells: So slots are really powerful and code components. Gives me an like here's your content work with this. So an example of that is I've got a masonry component. So I dropped the code component on the page and then I drop a collection list inside of it. And a bunch of images from the CMS, for example, and it knows how to do the masonry, the columnated calculations where things perfectly. And I didn't have to do anything. I just tell it how many columns I want and it does the rest. I feel like this is the future of builders of design platforms is that functionality like that should be super easy to access. Literally drop it in.

18:32Michael Wells: And now I've got a bento layout or now I've got a, you what do I want? I should be able to configure it and do some sliders instead of going, okay, what's the JavaScript and the CSS that I need to make this thing work the way the client wants? It shouldn't be that hard.

18:48Janne Parri: Yeah, yeah. Okay. Those were like code components. How do those tie with the AppGen you mentioned? And what about Webflow Cloud? For those who have not played around with those, what's like the mix and what can you do with all of those?

19:06Michael Wells: So I'll start with cloud. So cloud's quite different. Cloud is, you could think of it as building a full application that's plugged into your Webflow site under a path. So if I want to create, if I've got a real estate site and I want to have a really advanced complex map showing thousands of properties. You could spend a lot of time with Mapbox or Google Maps and JavaScript and try to get the experience you want. Or you could do all of that in React where you have a little bit more programming and UI control, a little bit more interactivity, certain statefulness and so on. And build that page, we'll call it slash map as its own thing.

19:48Michael Wells: And then what Webflow Cloud allows you to do is to host them both together. So you've got your main site, which is all you see it in the designer. And then you've got your slash map site, which is your separate react site that does whatever you want. Spinning globes can do anything. Pulls in data from wherever it wants. Might be coming straight from a real estate database. You have all that app level. Flexibility plus the ability to secure it. If you want an authentication system, you drop an authentication system reacts really good at these types of things. What makes we've always been able to do that, right? With a reverse proxy, I have always been able to create an app of some kind and like stitch it into my site, layer it on top of my site.

20:32Michael Wells: But cloud takes it a step further. By giving you a way to export your design system, all of your components, your variables, your classes, so that the React app can look the same. Pick up your styles, use your same navigation. That's what's really special about Webflow's approach is you get that cohesive approach and your design control is still in the designer where the design team can manage it.

21:00Janne Parri: Yeah, yeah. What about the AppGen thing?

21:02Michael Wells: Yeah, AppGen is quite interesting. It's new. I think version one is kind of a really fun toy with some simple application in the real world. Know, like if I want to build a calculator that has a weird complicated quote calculator, I can vibe code that and create something that looks nice. And it's handling, it's creating a cloud app, sort of. Same idea. It doesn't create code components yet. I think that was promised at the last Webflow conference was that is something that the team is working on. I haven't seen anything yet. That's super exciting. But if you want to build a page that does something special, a map.

21:45Michael Wells: A quote calculator, you know something that you can describe pretty straight in a straightforward way. Give it the data that it needs then. But it's kind of perfect for that and it's generating a cloud flare app. Sorry a cloud app cloud app running on cloud player right and. And all of the capabilities. That you have in your website. So the styling and the variables and your classes and your components, all of those you can bring straight into AppGen so it can utilize them itself. With all AIs, it takes some time before the intelligence is

22:22Michael Wells: There to really leverage it. But the thing about AppGen that really impressed me was it has the ability to use your CMS content as well, which you cannot yet do in cloud. So cloud apps, I don't think there's a way, any direct way for your cloud app to get to the CMS content, but AppGen builds it in, which means. That idea of a map for your real estate clients. If your listings are in the CMS, you can show them on the map. That kind of thing becomes very possible.

22:56Janne Parri: Quick side notes here. So I have already booked a bunch of guests for the show. So if you are liking what you're hearing and seeing, please consider subscribing. I am also doing live streams on YouTube. So check the schedule from my channel and come and say hi while I'm live. That's all on with the show. Kind of feel like we have all these puzzle pieces and it's slowly coming together. Obviously I haven't used all of those things extensively. And as you said, AppGen is quite new. I believe it's still in beta and so on, but that's the vibe coding thing. But at the same time, there are like all the other tools.

23:39Janne Parri: That are developing quickly. Like people are trying to like code custom sites from the scratch using Claude Code and all the other tools. And some people are actually having some success with that. Those tools are getting better and the sites that they are producing are are like turning out like quite good in the visual aspects too. So what do you think where are we heading?

24:12Janne Parri: Should webflow folks like check into all the other possibilities and so of that Rambler.

24:18Michael Wells: I'll put it this way. If you're in an industry that touches bytes, you're at risk. Doesn't matter. Even communities like artistic communities that are used to paint and music and live performance, even those are at risk. If it comes to information. You know, AI is changing everything and you have to pay attention to what that means. In a sense, it's the ultimate power tool, but it comes with all the dangers of a power tool. You need to learn how to use it safely. You need to learn how to not wreck your project. One of the things that I find shaping my perspective of where AI fits really well in my projects is it's

25:03Michael Wells: It's phenomenal at the, just like a chainsaw. Can do, I can make big changes with it quickly. But when it comes down to precise work and ongoing maintenance, it's much harder to direct it. It's just not, it doesn't have enough finesse and you have to know how to control it and talk to it. So on the concept of Webflow sites, right? So if I'm building a site for myself, I built a a university map yesterday for fun because I happen to live on a campus. Here in new Zealand, it's just the end of summer. Okay. And all the university students are returning and the new students are completely lost. Where the heck do I find my room? So I realized this. Thought that's very interesting. I wonder if I can make a map really quickly. And so I built a full 3D map.

25:58Michael Wells: Campus. You can fly over, you can zoom in, you can click any building, you can see the rooms. All the rooms are listed. You can search for a room and it will show you walking directions from where you're at on your phone. All of that done in an hour and a half. Now Webflow didn't touch it, right? Because I didn't need any design canvas. Wasn't building a page, I wasn't worried about SEO, I don't care about maintaining this thing. That was a perfect project for AI. You load the webpage, it's just a big 3D map, that's it. For a Webflow project, I think very differently because even if I could build a good looking site,

26:39Michael Wells: That I love the style of it, which is really hard to do with AI so far. It's not great at style, right? Even if I got the customer to love it, how do they maintain that thing? The only way they can change it is edit the raw code or talk to the AI and hope it understands their way of talking to it and doesn't wreck the site. That's not sustainable. So businesses rely on their websites. They need the ability to update them. And usually several people, like most of my clients are, my direct clients are medical centers. So they have receptionists modifying the hours of operation and posting the modals on what's happening this weekend and updating the service fees and those types of things. If they had to do that through AI,

27:28Michael Wells: They're more likely to wreck it than anything else. So at this point, that's the way I look at Webflow's place. They're integrating AI into the platform to assist with tasks like design. Possibly administration. Someday maybe things like content creation. Maybe I'll be able to write a blog article inside the editing interface. That would be great. Maybe I'll be able to create an image as a client. Say I want a hero banner. Create me a hero banner. I think that would be lovely. That's great. But I don't want it to take away the level of control that I have in the site. I'll always want the canvas.

28:09Michael Wells: I'll always want direct access to the CMS and the ability to hand manipulate things. I think that control surface for any client and for any designer will always be important. It's hard for me to imagine a day where I can just talk to my computer and it will make what's in my head. Until it can read minds. I don't know even know how I would communicate it, right? So there's, I feel like there will always be that gap between my vision and the reality. And I need my hands on the product to make that happen.

28:41Janne Parri: Yeah, yeah, that is so true. Prompting is not a good UI for those detailed things. And well, someone might say that, okay, you can just like edit the code, but that isn't easy in a sense that I wouldn't trust. Trust like my clients to do that. And I wouldn't want to be the service provider who says that if you want to change this, have to like code it. So, yeah, definitely I am on the same boat, but I'm wondering all the other tools that Webflow, they probably see the same thing, same insight.

29:24Janne Parri: And I'm wondering if they are trying to build some sort of like editing canvas. I believe there is some, versions of that already seen, how do you feel about that? Are the other tools going to like replicate that or yeah.

29:41Michael Wells: Doing competitors in the same web design space.

29:44Janne Parri: I mean, more like those vibe coding things.

29:47Michael Wells: I think certain functionality, if you pick a specific area like AppGen is a perfect example of this, right? It offers a really interesting value proposition to me as a designer, which is that I can do my design on the canvas. I can get the branding right. I can design components that represent the pieces that I want to see on my site and that the client has approved and their marketing and legal teams have all checked off on. And then I can go build an app without coding knowledge, right? I can build my real estate map. I can build my quote calculator and it's going to use the components I've already got. It's going to use the branding I've already confirmed. It's going to use the navigation I already have site wide.

30:34Michael Wells: That level of consistency, but adding the vibe coding aspect, that's awesome. I think it will take it some time to sort of tighten the screws and get the right model and for people to become comfortable and familiar with it. But that's sort of the ideal path because Webflow's goal has always been to help artists and designers create. Without having to dig into the technical stuff. That's the whole reason the canvas was built. No one likes coding in HTML. I'm a programmer.

31:06Michael Wells: Even I didn't like coding in HTML, right?

31:08Janne Parri: You

31:08Michael Wells: When I saw the canvas, I'm like, yes, please, more of that. Take my money. I really wanted to shift my thinking to something more visual, faster, practical, where I don't have to deal with typos and unclosed entities. I think it's the same with coding. Even developers like me, when AppGen gets to a certain point, I'll go there first for those little features where it's just like, just need, I want a quiz. My client wants a quiz, what can you do for me, AppGen? And I'll spin up a couple of versions, and if something hits, I'm done. Half an hour later, my job is done, and it's beautiful, and it works, and it's brand aligned. That's exciting.

31:47Janne Parri: Yeah, yeah. It's kind of like the question that are you starting from the control and adding the AI and all the possibilities there, or are you starting with the vibe coding tool and trying to add control to that? And I feel that at this point, the control

32:05Michael Wells: You

32:05Janne Parri: First AI second is way better.

32:08Michael Wells: It's both. It's both though, isn't it like I agree like I the way AI works. It's entirely dependent on context. It needs to know. What? It needs to know. You have to tell it what it needs to know in order for it to do anything aligned with your vision. So if I say make me a gorgeous website and I've got a picture in my mind what I want it to look like, not going to get anywhere near that. But if I hand it 50 photos, color scheme, logos, some photos of design layouts I like, it's going to get a lot closer.

32:45Michael Wells: Right? So that's where the control comes in. If I build my design system first, my layouts, my components, my classes, my styles, my branding colors, it has the right materials to work with. And then when AppGen picks that up, it can generate things that are actually brand aligned and are actually much closer to something I can use. But now you see Webflow working on what do they call it? The assistant designer assistant AIs. So that when you've got a blank

33:12Janne Parri: Yeah, I'm not sure.

33:13Michael Wells: Canvas, you can now vibe code and design in that kind of direction. Now that's all early stages, but that's exciting as well. And there it has no structure other than it knows what a component is. It knows what a slider is. It knows how the CMS works. Be interesting to see how people embrace that.

33:35Janne Parri: Totally. I believe many of us are waiting for that. Yeah, that's cool. I lost my trail of thought, just started to wonder what the future holds.

33:46Michael Wells: Tell you. I mean. The experiences that I've had, I built a lot of sites using nothing but AI and they're actually awesome. They're awesome. The capabilities are awesome. The speed of delivery is awesome. The depth of the technical capabilities is awesome. Have to know enough about development to ensure that it's secure and the credit card numbers aren't exposed and you know, all of these things, but still as a dev that's awesome. But the end result from a branding and styling standpoint, the end result from a response and a responsiveness standpoint, tools like localization don't exist, right?

34:29Michael Wells: These things that I'm building only work for me as personal projects. It's like, this was fun. I have a new thing. I could never deliver that to a client. So I really, I'm super excited that Webflow kind of sees that their role in this new world is to be that control surface where that canvas is still the center of your world, but AI is assisting you in everything around that as a partner, as a team member on the design team. As a dev with appgen, you know, they get it. They understand that crystal is it's a question as to whether they can pull it off. The whole experience, I'm sure they're feeling this as well. But for me, as a technologist, I feel like I'm skiing on top of an avalanche. You know, the ground is constantly

35:17Michael Wells: Shifting and I'm just doing what I can to manage the constant change. For a team of Webflow, what do they have, 1,000 employees now? ⁓ For a team of Webflow size to try to coordinate and manage that level of change, just in the underlying tech, that's wild... But they're delivering such great things. I've got a lot of excitement for the future.

35:38Janne Parri: Yeah. And I just had Matthew Munger here who works at Webflow and they are like delivering those updates like regularly. They are not like waiting for the next Webflow Conf to deliver those updates. And you mentioned the CMS updates. We got those and I believe that there might be some classic wishlist items coming up. So kudos for Webflow. For that definitely. But before we go to the bigger picture, had the thought about like components and us, developers and designers building like the systems for our clients in a way trying to build a page builder within Webflow. And if AI could help with that would be the beautiful

36:33Janne Parri: Combination. Feel that the components are like great, but the end result can beautiful and can be like powerful, but there is still a learning curve. I have to really train my clients to use those things and so on, but how do you feel about those components and how do you approach the whole thing of building like a system that the clients can use and build all sorts of things themselves?

37:05Michael Wells: Yeah, so first there are so many different kinds of components you can build. Some of my components are use the standard native component technology plus JavaScript. So if I'm building something for one client, like a fancy quiz, maybe it's got a slider and you answer a question and then it automatically slides and you can't go ahead until you've answered the questions, gets to the end and it shows you a nice result. Where do you fit in the space of other people who have answered this quiz? That kind of mechanic.

37:40Michael Wells: Clients find those really complex to build agencies find those really complex to build so they'll bring me in to be like OK, here's our vision here. Usually say here's what's on the figma. We've promised this thing. We don't know how to build it. OK, so I come in. And I'll usually build it in that particular case, I'll build it as a standard component because I want their design team to have full control over coloring, styling, button sizes, font. If they want to tweak anything, they should be able to tweak it. If I put it in a code component, that's much harder. They have touch the code in order to change something in general, right? So.

38:21Michael Wells: So I'll build it, I'll wire it up with HTML using custom attributes in order to find the right element. I'll build control infrastructure so that I can use the Webflow slider, but it won't advance unless I tell it's okay to advance, things like that. So I end up building all these pieces. What the agency gets in the site is a component that they can drop on any page and just by virtue of the attributes it has, my site level code will recognize it as, look, there's an instance of the quiz. Load this code, initialize it, bring in the content, load it to page one, it's ready to go, done. And they don't have to worry about it. They've got this sort of mechanic that is invisible to them, that's site-wide, that just steps in when it's needed. But from their perspective, they can take this component and drop it anywhere they want on the site. Now,

39:16Michael Wells: The one challenge is there's usually a fair bit of complexity. There's usually, the reason they brought me in is because of that complexity. So there are things that I want them to know. Hey, you've got 20 different options on slide speed. Hey, if you want to change,

39:33Janne Parri: You

39:34Michael Wells: You know, how big the chart is at the end or how the spam detection works before it sends an email, you know, here's where those controls are. So all those custom attributes, the variables that I've used for styling the quiz, even the mechanics of how the quiz work. Where the repo is in GitHub, where the CDN is that's serving the code, right? All of that, I put into separate documentation. So I'll actually build a documentation site for every client, usually. Password protected, so that at any point when they run into, we wanna change this thing, they don't have to call me. They might anyway, but they don't have to. Everything.

40:15Michael Wells: Is there I rely on it as much as they do because a year later I'm like I don't remember what I built, you know, but the documentation tells me everything I need, so that becomes really important. In anything technical is to deliver that user guide adjacent to it.

40:32Janne Parri: Yeah. Yeah. And, I believe you do not have to go even so, so far, like even with the, with the open components that is built, with a goal that the clients can just like track and drop a new section and some content to that. That should have some kind of documentation with it. Maybe just a short loom video or something like that. So that is a very good practice. Yeah.

41:00Michael Wells: I do a lot of Loom videos. I'm probably Loom's number one client, I think.

41:07Janne Parri: I've actually tried Tella. I believe it's Tella TV. Similar product, but I do not remember why I started using that instead of Loom. Yeah, there was something, some kind of editing capabilities probably. But yeah, those are totally cool. Were you about to say something?

41:26Michael Wells: No, no, but I tell you on that topic, there are so many tools that have come out in the past decade alone. Just as the internet matured, Loom is a great example of that. The ability to just record an idea. You're done and you've got a URL you can share. I mean, that's so game changing for us as designers and as people that technical people that need to communicate complexity. I can't imagine how inefficient I was before that having to like hand type every notes and try to take screenshots and compose. I was so agonizing. I can't tell these guys how much

42:02Michael Wells: I appreciate that work.

42:04Janne Parri: Yeah, it's at the same time, like faster, more efficient, and clients find those videos like totally awesome. Like I still get praise for making a short video. It might be just me rambling, sitting on a sofa and saying the same thing I would write on an email, but it's faster and clients like it way more. So that's nice. But yeah, if we change towards the bigger picture, we've talked about Webflows, future and all the stuff they are doing, a little bit of vibe coding. Where do you see the whole AI thing, the internet thing is going? Do we still need websites at some point?

42:53Michael Wells: Man, this is going to be like a crystal ball experience, isn't it? I'll tell you what I expect will happen. And only because I've seen the internet since it was born back when it was called Bitnet, you know, and Usenet and seen all these texts converge together. The driving force behind the internet was that people want access and capability. At their fingertips. It's been the driving force for the development of mobile phones. It's the driving force for Wi-Fi. We want our stuff now. We want to be able to do something now. We want to be able to know something now. So what does that mean in the future with AI? And I think that it's going to be a pretty radical change. Think I was just reflecting on this yesterday, actually. I need to write a piece on this because

43:41Michael Wells: I feel like to most people, including me, for the past 25 years of internet, websites are still mostly a brochure. It's mostly about giving you information, answering questions. What time is this restaurant open? What does the food look like? I want to see photos. And some action, e-commerce, filling in a form, making contact to a business, finding things on a map, booking a medical appointment, those types of things. To the average person, the more comfortable and accessible those processes are answering my questions and letting me do things.

44:20Michael Wells: I think that will direct how the internet evolves. So you can see already, in fact, I was marveling yesterday. I watch on my Webflow sites, I've built some infrastructure, reverse proxy. I can see every AI bot that visits my site. And it's 10 to one, 10 times as many bots as humans. Okay. And it's been that way for a year.

44:44Janne Parri: That's something.

44:44Michael Wells: All my clients, it's amazing. If you wonder why you're getting bandwidth calls. It's not people. But I want to say that's shifted dramatically. So initially it was just AI bots scraping your site for content to feed their hungry model building engines, right? Now it's shifting into a territory I call AI assistance. So user is using a browser, but it's got an AI built into the sidebar. Okay. It might be Microsoft edge with copilot, or it might be any number of things, more AI enabled browsers. And they're actually talking to the AI. Hey, I want to find a nice place to go for the weekend. It's choosing where to visit with the site, what to read, what bookings are available, showing you photos.

45:34Michael Wells: So in a sense, the AI is doing the web visits. It's still doing the traffic. It's still identifiable as an AI, but it's under the direction of a human. Like right now, it's feeding back results. So what I realized was the number one bot visiting my site. So if you look at the graph, you'll see this little rumble across the bottom. That's the humans. And then there's 20 different agents I'm tracking. The number one agent is called ChatGPT assistant. Now don't know where that agent runs. Is it a Chrome plugin? I don't know where it's coming from. But the traffic from that one agent...

46:13Michael Wells: It looks like human traffic every week. You could see the weekly bump, the weekends

46:17Janne Parri: Hell yeah.

46:18Michael Wells: Diets perfectly human. It's being controlled by people. Now what's interesting is that means that from a practical standpoint, the human traffic to my website has increased tenfold in the last year. But now a lot of it's coming through these AI assistant agents, right? So what's happening? Well, my site ⁓ is sygnal.com. This is the site I'm referring to. It's an informational site.

46:49Michael Wells: So it's used mostly by people who are designers working on interesting tech problems, wanna learn something. And so they will go to their bot, they'll ask it questions, they'll be researching something like how do I get a grid to do this crazy wild thing? And it'll come across my site and it'll start sending traffic to my site, right? From the designer's perspective, the user who's driving this, they're getting information faster and more conveniently than if they went the Google route. Open 20 sites, read all of them, throw half of them away.

47:23Michael Wells: You remember the process, right? So where does this go? I'm expecting, have you heard of NVIDIA PersonaPlex? No? Okay.

47:32Janne Parri: No, I haven't. What's that?

47:34Michael Wells: So NVIDIA is constantly releasing some pretty interesting edge models that do specific things. PersonaPlex is an audio voice model. What they did was they solved the problem of real time speech to text and text to speech. So it's invisible. Okay. They call it zero latency. So I say something and it responds instantly. What's really interesting about it is unlike you've probably talked to AIs before and it's very like, it's like a tennis

48:07Michael Wells: Game. I hit it, I wait, it hits it back, I wait.

48:09Janne Parri: You

48:11Michael Wells: And then it's awkward. It's unnatural, Personaplex, when it's talking, it's also listening, which means that I can interrupt it. If it's babbling away about something, I go, no, no, I was asking about... This thing and it goes, of course, well then that, you know, it will totally change direction. It feels like talking to a human. It feels a lot closer to talking to a human. The voice quality models are amazing. Its responsiveness is amazing. They're trying to give it also, I think, more personality. It's definitely got different accents. Now I'm talking to an Indian guy. Now I'm talking to an American girl. Now I'm talking, you know, so you've got.

48:55Michael Wells: Just a wide range of capabilities. So for websites in the future where their entire job is to answer basic questions that a chat bot can do, when are you open? Can I make a booking, a reservation at your restaurant? That could be a dramatic change, because why do I wanna click and scroll and read and click? I'm driving my car, I just wanna make a restaurant reservation. I should be able to talk to the site, have it give me the information I need, conversationally, right?

49:28Michael Wells: And I should be able to direct it what I want. I should be able to do shopping. I should be able to be booking my vacation while I'm stuck in the commute on the way home from work. I should be able to do all of that by voice. And then the website itself becomes a database, an information repository with visuals. I definitely want to see pictures of, you know, the resort that I'm booking for my stay. But the concept of a website shifts a lot. Shifts a lot is an experience. So I'm expecting that half of websites where their job is information will start to shift in a direction towards smart chat bots that you can verbally converse with, right? And things like text only and things like visual interfaces will still exist.

50:19Michael Wells: But primarily for accessibility reasons. If someone's deaf, they're gonna have trouble talking to the, you know, so there has to be a backup. So that changes the nature of the way a site works in really interesting ways. And then you'll have other sites where the purpose of the site is an experience. I'm showing off all my art. Here's an art gallery. Or I'm a video producer or a podcaster and I want you to be able to access my content and experience it in a certain way.

50:51Michael Wells: Those will probably remain more or less the way we're familiar with. It's more of a window shopping experience. I'm going there because I want to see and feel what this is like. And they'll probably expand in some directions like virtual reality and augmented reality as those eventually grow. They've been kind of slow. But I've always wondered like why can't maybe I can. I've got a I've got a Quest 3 headset here somewhere and I'm like if I want to buy a piece of furniture from IKEA, why can't I just slap that on and then go swipe shopping and like yeah, that sofa fits. You know that looks good. The tech

51:28Michael Wells: Is all there. But people haven't really embraced it yet. They're just a bunch of threads, including consumer awareness, you know, and tech access that have to happen. But. Yeah, the websites of the future, it's just going to keep heading towards more accessible, easier to use, faster to use, more complete information, less work for the user browsing the site to get from point A to point B.

51:54Janne Parri: Yeah, yeah, I'm wondering about the visual aspects. Like if there are more visually oriented people, let's say like branding people or graphic designers there and they are like scared that scared that their work will be useless in some years because AI just reads the websites and all the content and we need less of the visual stuff, but could the AI like somehow validate the visual experience and take that into consideration so that it could like recommend that, okay, this hotel has a very nice website. I believe that they are quite good at this hotel thing too.

52:39Janne Parri: Like where is where will the line be that how important or how? Unimportant the visual aspects of the let's say more informational sides will be do I get

52:52Michael Wells: I think

52:53Janne Parri: What I'm saying?

52:53Michael Wells: I do, I do. So those are kind of separate questions in my head. The visual aspect will always be essential. And the reason is as humans, visual information, the way our brain processes information, vision is so central. It affects everything, what we want, what we buy. I see a poster, suddenly I'm in the mood for a hamburger. Know, every, this is just so central to our processing. You could say the word hamburger. I'm not going to run into Burger King, right? But if I see it, I'll have a different emotional response. Because of that, visual will always be absolutely central to any interface that humans want to adopt, and especially the web. But I think it's going to change how it's created as the tech changes. So if you remember the time before the web, it was we have print.

53:45Michael Wells: And actually, if I go way back, we had black and white print. And that was what you worked with as a visual designer. Magazines were expensive to produce, you know, and then the internet came along and suddenly everything could be color. And we adapted to that. We adapted to scrolling pages. We adapted to the fold. We adapted to multiple pages. Then we adapted to responsive designs and mobile phone displays. And we kept changing. I think the next phase. Of adaptation will be into new models because code is becoming cheap. And to me, one of the first things that's impacting is that this, I mentioned this 3D maps that I built. Crazy full 3D map of an entire campus, three campuses actually. I can fly between them. You can tell where I am on the phone. It's a wild experience. It took me an hour and a half to build. Very visual, right?

54:40Michael Wells: So now the question is, as a user, what are my expectations going to become once 2D flat scrolling web pages are no longer a limitation? We're at 5G now in mobile access. We can deliver a lot of data. I don't see too many VR experiences on the phone yet. My next experiment actually with this whole map thing, it just started as a wild experiment. Is to migrate it to a mobile phone app. Where you where it's an AR experience. I identify the class I'm looking for by number, which they have on their student syllabus. And then I just point the camera of the phone and it will show me the building is over there behind those two other buildings. There's a building like transparency and X ray

55:27Michael Wells: Type feature. Why? Because I can because it's I want to see how it feels. It's interesting. AI is making this possible. In the past, that's been possible now for almost a decade. I think Google had an experiment in Chicago where they did transparent buildings to find your cafe. But the coding is expensive. The user interfaces are expensive to build. It's hard to test. It's hard to deploy. The bandwidth available on the internet for an average person on a phone wasn't high enough to really make it comfortable. That's all been moving forwards. So now when you're a web designer, and I'd say this applies today, not just a year from now.

56:08Michael Wells: I'm always now rethinking like, should this be a flat scrolling page or should I be using code components to deliver something new in the middle of the page? The whole concept of satisfying that user's need for information, efficient, know, understandable, usable information and controls. If I want to make a booking. We're used to calendars, but what else can we do? There might be other ways to represent that are quick and easy for me to like it looks at my calendar and it says, you're free on Saturday night. Let's grab that.

56:41Janne Parri: Yeah, yeah, but I'm wondering like, will the will the website be the place where those experiences are find whether it is a AI bots or an actual person, could the AI stuff like increase the value of all the social media things so simple example, if there is a video studio and I'm talking to LLM about what sort of video agency I should hire, it could just serve me the videos from, let's say, YouTube. And with all the other experiences, might there be some other places than the websites that the website could be less and less important and all the other medias could be more important?

57:29Janne Parri: What do you think?

57:30Michael Wells: So I think this is going to be a point of pressure that the whole industry is facing. And it's always been the case that you've got aggregator sites that take all your content. So I might be a photographer. Should, even 10 years ago, 20 years ago, right? Should I create a site to show off my work? Or should I upload it all to Instagram or Picasa or somewhere else, right? Where there are people. So the internet's always kind of worked like that. And I think that's going to increase. Should I build a site to sell my furniture or should I use eBay? Right. It's always that kind of challenge. Now, when it comes to your average site and information, let's say I'm doing e-commerce. There's now a new protocol Google has just released where

58:22Michael Wells: It's an agentic e-commerce protocol. Have you heard of it? Do know what I'm talking about?

58:28Janne Parri: Probably not.

58:29Michael Wells: So the idea is that I'm doing Christmas shopping end of 2026. I'm really excited to see what happens this Christmas. It didn't happen last Christmas. I was watching, but I think this Christmas we're probably better positioned where you could, know, it's September and you're like, I want to prepare for my Christmas shopping and order my stuff. And you hand it your wife's Instagram feed. And it figures out what style does she like, where does she like to go, what kinds of things is she interested in, what books is she reading, whatever is in that feed, it can pull the information.

59:05Michael Wells: And now you've got, let's say, ChatGPT or whatever is the agent you're using. It can now come up with some ideas on what does she, what would she enjoy? It can even direct you, hey, go take a photo of her wardrobe, show me what shoes she has, you know, and it will figure out like what's she likely to enjoy. And then the agent goes and does the shopping. Now what's interesting about this is every website owner is now not just building your typical e-commerce storefront, right? They've also got this protocol that allows an agent to come, shop, query what do you have, look at photos, and check out and purchase products automatically. So now we're building our sites in a different way, right? Humans are always gonna be important.

59:54Michael Wells: But that's gonna start shifting into where the humans have contact with the site and where the agents have contact with the site. And as that gets easier and easier, take grocery shopping. Right now, I do my grocery shopping, I have it delivered, all of it's bought from one store. That's probably not the best deal. All right. And probably not the widest range. If I said here's my shopping list, here's what I like and have it be able to talk to 20 grocery stores, figure out the right pricing and then organize delivery, which is a logistics problem. Right. Suddenly my options open up a lot. So I think websites will always be relevant, but the shape that they will take is changing.

1:00:38Michael Wells: The agentic interface will grow a lot. Another area that's happening is WebMCP, which now lets you interact with your website using that assistant bot in real time. Like I could have it browse through, let's say we have your history of podcasts. I could go to your site, talk to my agent in the sidebar and say, browse through the list of podcasts. I'm interested in things that have to do with IX3. Find me those and it'll look at the transcripts. It'll find the right pieces and it'll literally just mine all the data for me. But I did it on your website with your website's help.

1:01:14Janne Parri: That's cool. Add to those people who are now trying to look at my website and hold the list of the podcast episodes. Those are coming as

1:01:24Michael Wells: I'm

1:01:25Janne Parri: Always. Every designer is in the process of updating their sites as and my. But do you think like that there will be like this general for a lack of a better word integrations like the e-commerce thing or will the AIs be able to navigate the sites that are already there as a human would do.

1:01:54Michael Wells: It already can. In fact, I've just started building a product for my own clients initially, but I might be able to make it available to the wider community, which is I instruct an AI agent to completely scan one of my sites and find problems, broken links, typos. It is amazing how thorough it is. I found things, I just unleashed it on one of my medical client sites. It found hours of operation on one page that disagreed with hours of operation on a different page. Just because we forgot it was even there, right? Things change and then we didn't know where it needed updated. It just found so many things. And now I find that capability, the fact that agents can real time navigate and browse a website is itself useful. But

1:02:45Michael Wells: When you give it a specialized interface, web MCP, the Chrome agentic shopping protocol, I don't remember the name, MCPs in general, I feel like the stack changes. We tend to think of our sites as, today, we tend to think of them as CMS. And then HTML and then the users here, right? But now I think it's gonna be CMS, HTML, MCP that lets it browse everything. And then we've got humans and bots interacting, both interacting with the site in different ways. Efficiency is better, especially because when you try to scale this, when you say, I want to buy some nice shoes for my wife.

1:03:29Michael Wells: The faster and more efficient the communication between the agent and the websites are, the more sites it can efficiently check. And I really like this because it has the potential to sort of democratize the internet again. Instead of having like Amazon owns all the shoes, now my agent's like, no, I like Bob's shoes. Those are nice. He's done a good job. Check those out. See what you think. That's a different experience and it cuts out the middle man and that makes me happy.

1:04:01Janne Parri: Yeah, yeah, but what about all the well in the past we have had SEO optimizers and in the future could we see some not so good actors like trying to optimize the sites for the LLMs somehow to trick them to choose them or something like that. Is that in your crystal ball?

1:04:23Michael Wells: 100%. Every tech is a knife and it has good uses and it has bad uses and people will use it badly. In fact, were just talking about this the other day. There's this fantastic technology that Cloudflare has released where when an agent comes to your site, Cloudflare can recognize that's an agent, let's give it clean, simple markdown instead of all the HTML and images and CSS and JavaScript, which is a hundred times heavier, slower, costs more bandwidth, has all sorts. And like I said, if 10 times as many agents as humans are coming to my site, it just makes sense that 90 % should be efficient.

1:05:09Michael Wells: So the problem is, I think it will take a little while before the makers of these agents trust that because it's easy to, it's not easy to hack. But if you're interested, you could definitely say, you know what? Humans see this. Agents get this completely different thing, which has nothing to do with reality. It's kind of like the problem we used to have. Back when Google used the meta keywords. People would do keyword stuffing with unrelated things and just completely abuse it. At some point Google was like, we give up. We're not even reading it anymore and just ignored it. And that was 20 years ago. So yeah, people will abuse these interfaces. Definitely.

1:05:54Janne Parri: Oh yeah, and it goes, the practice goes even further than that. Like I remember when I was little, there was like those yellow pages. The phone book had all the companies listed and then there were a lot of companies with AAA this and AAA that like just to get to the top of the list. Like, yeah. It has been around and it will be around. Obviously I'm excited to see what kind of tactics will be used.

1:06:25Michael Wells: There are always going to be challenges with tech no matter what we do, no matter how far it advances. I'm still super excited at the benefits that it can bring as well. So I think even this year, but especially the next two to three years, we're going to see transformations that make us rethink. Everything about how we shop, how we get information, how we get entertained, how we choose things, how we communicate with our friends. You know, it's like, it's very much like the feeling that we had at the beginning of the internet where it's like, wow, everything's new. What's this Instagram thing? How does that work? How does it

1:07:04Michael Wells: Affect my life? You know, we're in that all over again. And I'll tell you what, I'm lucky enough to experience that twice in one lifetime. That's kind of wild. It's kind of wild.

1:07:15Janne Parri: Obviously the people that are living right now always feel that big stuff are happening and we have never been so, so out there it's the experience of being alive, but I really feel that there are some like major things shifting in our lifetimes. But one thing still about the AI stuff, it. In the past, there was a lot of talk about AI hallucinating stuff. And at least in my bubble, it is not as present as it used to be. So do you think that we have kind of solved that or where

1:07:54Michael Wells: Now.

1:07:55Janne Parri: Are with the trust thing.

1:07:57Michael Wells: You yeah, AI hallucination is really interesting. If you have done any research into how LLMs work, the current wisdom is that the smarter and the more advanced LLMs get, the worse the hallucination problem becomes.

1:08:15Janne Parri: You

1:08:16Michael Wells: And that it's unsolvable. It's currently thought of as unsolvable. What I think has happened, and by the way, you can see hallucinations in weird places that you wouldn't expect. So for example, I use when I'm programming all day long, I talk to my computer. I tell it what I want. Don't want to type all day long. So I built a piece of software using vibe coding, right? That allows me to just hit a button and start talking. Then there's a product called Wispr Flow. This is similar in concept to that, but it's just built into my computer, runs a local model, doesn't cost me any subscription fees. And it does a really good job. And so anywhere I wanna type, I just hit a button and talk. Now that...

1:09:03Michael Wells: Tool uses an internal model for the speech to text conversion. It's called Amazon Whisper. That's where Wispr Flow name comes from. It's a model that does that transcription. And it can hallucinate too. And the way it hallucinates is you see the transcription and it's where it's normal. And then it suddenly goes off the rails and it'll say something like. I'm not sure I'm not sure I'm not sure I'm not sure I'm not sure for like 30

1:09:33Janne Parri: You

1:09:34Michael Wells: Times and then it goes back to normal and you're like what was that? That's

1:09:38Janne Parri: You

1:09:38Michael Wells: Not what I said. I can check the audio. It has nothing you know weird about it. This is what hallucinations look like. So you wonder about like how does my Tesla self driving hallucinate? What does that look like? What happens internally in its information processing? But all LLMs have this issue. But from a practical standpoint, when it comes to things like coding with Claude Code or using Claude co-work to generate a slide deck presentation for, I run an AI group in Auckland and every week I'm generating decks of slides for the classes.

1:10:14Michael Wells: And I don't have time to type these things, right? So ⁓ you're right. I don't see hallucination ⁓ surface there as much. So I think what's happened is that the savvier vendors like Anthropic have gotten good at detecting and suppressing hallucinations at the intermediate layer. So the model hallucinates and they're like, that's not right. Check it out and redo it. ⁓ And yeah, something's happening there... But ⁓ from an experience of using the LLMs, it's so much better.

1:10:45Janne Parri: Yeah, yeah, I feel that too. But what about, well, we visited, optimization stuff. If we bring this back to Webflow, how do you see Webflow stands ⁓ with all the AI, AEO stuff at all ⁓ of those optimization tools? Do you feel that Webflow is in a good place?

1:11:05Michael Wells: Have you used optimize?

1:11:06Janne Parri: No, not I have just used the analyze not the optimize

1:11:10Michael Wells: It's amazing. It's amazing. It's very well built. I was in the beta and it really impressed me with how much control you had over creating your experiments and how they realized ahead of time. This was the key. Positioning of the idea of the Webflow experience platform was that you create these different experiments, different experiences, different CTAs. You want to know, would this work better? Does this color change things? What if this wording or this video or this whatever, how does that impact our conversion rates? You can set that all up.

1:11:48Michael Wells: What's really genius about it, in my opinion, is that it can run all of those itself and switch automatically. So unlike a normal A/B testing platform, you set up your experiments, you run them for a period of time until you get enough data, and then you stop the experiments, and then you declare a winner, and then you implement that as a permanent fix. But what Webflow realized and the company that they bought, I can't remember the name, the company that had originally built the AI based A/B testing technology, they realized that, throughout the year and over time and based on public sentiments and news and politics and whatever's going on, people's interests and tastes constantly change.

1:12:36Michael Wells: So even though you can run an experiment today and declare a winner, two months from now, it might be a different winner. And that was the results of their studies. So they realized that, hey, you know what? When we create an experiment, we should be running it forever. And if the system decides that, hey, this one's the winner now, it now shows that one more. To maximize, to automatically maximize your conversions, which is what we want. At the end of the day, it's all about squeeze that lemon as much as possible. I want as much juice as I can get. And that's, what AB is for. Yeah. I love that

1:13:14Janne Parri: That is.

1:13:15Michael Wells: They're really kind of breaching new territory. So WebFlows always has this mentality from my perspective of we see other people doing things and adopting ideas, but we want to go beyond that, right? We want to take a step further. Everyone's got AB. What can we do that's better than AB? Automated real time ongoing AB and maxing out your conversion rates. Intelligence as to where users are coming from customization, personalization. So they build some pretty amazing things and that's why I have a lot of excitement about.

1:13:47Michael Wells: Things like cloud and app gen and other things are building is like they really think of it as you can't just build a feature that's the same as everyone else. That's boring. It doesn't win you anything. You know, they want to go beyond that and shatter kind of new territory, which is why it takes a while to build and release and perfect these things as well.

1:14:06Janne Parri: The optimization thing is fascinating. It brings to mind like Google ads and their automated bidding. Obviously Google has a lot of data and they can optimize the ad bids to individuals. But what's your sense? How much data or how much traffic the Webflow optimize needs? In order to be effective.

1:14:32Michael Wells: I think it, I don't think it's any different from other AB setups and platforms, but for AB, if you ever done AB testing in general, it needs a lot of data to be effective. And that can be challenging. Think that's why I know a lot of people, including me, were shocked like, oh my gosh, optimize is not cheap. It's really, it's compared to the rest of the product. You know, it's expensive. But then I went and looked at the alternatives. I was like, I get it now. They're all enterprise products. It's nice to at least have this as an option, sub enterprise level for a medical center that can afford it and wants to maximize enrollments or something.

1:15:12Michael Wells: It makes sense. Whereas if you were. I don't even remember the names of them, but a lot of the products out there for proper A/B testing at a high level, they don't even list prices on their site. Like, call us. We'll quote something for

1:15:25Janne Parri: Yeah, I...

1:15:27Michael Wells: You, just for you. I don't like that at all. It makes me very uncomfortable.

1:15:32Janne Parri: I know the feeling and I have checked out those pricing pages too. But yeah, I believe there is one at least one like Webflow add-on or plugin. Was it Optibase or something like that, which is way cheaper and you can do like simple A/B testing tests if you wanna like try it out, but I have the experience of doing, trying out some, AP tests with a site that had some traffic, but not much. And all my, all my tests were like insignificant, like the data that I got, I could not do any decisions on that.

1:16:10Janne Parri: And as you said, the situations might change and the result might be different within a year, within a couple of months and so on.

1:16:19Michael Wells: What I what I really liked about because I used to use what was it called Google optimize or Google optimizely Google optimize. That was free. That

1:16:28Janne Parri: Yeah, something like that. I used that too. Yeah.

1:16:32Michael Wells: Was free for ages and then they shut it down. I don't know why Google likes to shut things down randomly to just annoy everyone. I had some clients depending on it as well and then. About a year after that, think, was when ⁓ Webflows optimize became available. And I started looking into it. And what really impressed me was the way the experiments work. If an experiment is not active, it's nowhere in the page. All of the decision making and all of the rendering is handled server side. So the page that's delivered

1:17:07Michael Wells: There's no problem of like I've got a white screen. Wait three seconds. OK, now it's made it's

1:17:11Janne Parri: You

1:17:12Michael Wells: Decision. The JavaScript has run. The right thing is showing it could pull back the curtain. You know that was killing your SEO because the time to first. Know, paint was like four or five seconds in. It was, it was really nice to see that they really thought about those things. And I also really liked the fact that from an optimization standpoint, in a sense, because everything's rendered server side, you could potentially also measure how the SEO impact works.

1:17:44Michael Wells: You're not delivering any duplicate data that's just sort of hidden in the page. It's very clean, which means you could run SEO analytics against the different experiments and kind of go, which one seems to win? You could get all sorts of other intelligence out of it as well. So after playing with Webflow's Optimize, I actually haven't been interested in looking at any other platforms when I need it.

1:18:09Janne Parri: You

1:18:10Michael Wells: It's like, it's the obvious choice. If I'm on Webflow, why would I use anything else? Unless budget is just extremely limited, then you have to look around.

1:18:20Janne Parri: Yeah. I'm just waiting for the client that has the budget and I can really try it out. Well, Webflow partners have the sandbox and we can like check out some demo data and those functions. So that's kind of cute, but I'd like to try it out in the real world. But you mentioned the server side rendering. And that is also a big thing with the new component visibility things. I believe that ⁓ the folks that are excited about that ⁓ tend to be the more tech savvy, which means usually the bigger clients, maybe the enterprise clients. So do you feel that a Webflow is going towards like the enterprise clients with all of these ⁓ advancements?

1:19:10Michael Wells: Some of the advancements for sure. So for example, with enterprise clients, you've got issues like. This customer wants to migrate a site to Webflow, but they've got 3 million CMS items. How do we deal with that? Right. So those were pressures for building the next gen CMS and it provided some of the funding and the motivation to make major infrastructure changes like that. I think the development of optimize was probably also driven heavily by those top tier clients who have the funds and really want the capability integrated and complete.

1:19:45Michael Wells: So that makes a lot of sense. But I've heard people in the forums complain like some new AI feature is released or something and are like Webflow only cares about the enterprise clients, but I don't see that at all. I mean features like appgen. I don't imagine any enterprise client ever using that. It's not for enterprise clients. They have an IT department. They have a dev team. They live and breathe, react, right? They can spin up something that they absolutely control and integrate it through cloud much more effectively.

1:20:17Michael Wells: AppGen tools like that are for the studios, the freelancers, the regular site users who want to add capability but don't have the funds for an IT department or even the contracted developer to build this piece, right? It's likely to put me out of work someday, but I'll find things to do.

1:20:37Janne Parri: You yeah, yeah, some of those features, the more developer focused features feel like that they are, especially for the IT departments, like checking the box that now you can do this, and the IT will be happy and obviously there are like features in the enterprise plans or plan that are important for the for the folks. But yeah. There is the sentiment that workflow should not be all about the enterprise clients, but I've said it before that it is good to know that there's room to grow for the smaller clients too. And if the big boys can do it with this platform, maybe I can too. And the prices of the smaller plans are on the level that I can definitely lead it.

1:21:27Janne Parri: Do that so I'm not worried about that. Yeah.

1:21:31Michael Wells: Yeah, that's one of the things actually I'm so pleased with. I've worked with a lot of companies in the past where you just watch even Shopify, right? You watch the prices just go up and up and up. I had one experience where I brought a client in. And they were using the lowest level plan. Was like the Shopify light because all they needed was the buy now buttons. I was using a Webflow site and I had a shopping experience built into it because they had some weird requirements that Webflow's Ecom couldn't do. And I was paying $9 a month and then a year later I went and checked and I was paying $50 a month.

1:22:10Janne Parri: You

1:22:10Michael Wells: What happened? They didn't even tell me, you know, they just kept eliminating lower plans and elevating my and I didn't think to check it didn't even occur to me that could happen. But it did. And Webflow, by contrast, you know, we've had the last few years, we've seen huge inflation, particularly in the US. We've seen hosting costs and data costs increase, oil costs increase. Everything has been increasing. But the base plans, the entry point for your base workspace and your base site plan, they haven't changed at all. And instead, Webflow's, instead of raising the whole, you know, pricing curve, they've steepened it. They've made it so if you use

1:22:51Michael Wells: More then ⁓ you're getting value and you can choose how far you want to go. How many designer seats do you want? How many client seats do you want beyond the free ones that you get? That's really nice. ⁓ I really like that and that awareness that. Most of the community is made up of designers and developers and freelancers and small businesses that rely on their platform. And so their intention has been like, how do we make that easier? We'll reintroduce client billing and better. We'll make seamless transfers a thing. Enterprise clients don't need any of that, right? But Webflow invests a

1:23:29Janne Parri: Have you?

1:23:30Michael Wells: Ton in our community.

1:23:32Janne Parri: There are like, levels to the business plan too. So before you are forced to jump to the enterprise plan. Have the bandwidth tiers and probably some CMS tiers or something like that. So in a way, you are not forced to jump into the enterprise plan. So fast so that's a good thing too yeah I guess I'll have to let you go but is there anything you'd like to add tell or advise to the listeners

1:24:07Janne Parri: Any things that we did not discuss this far.

1:24:10Michael Wells: I have a lot of people ask me in my AI meetup and in my Webflow meetup that I run here in Auckland. They ask what they should be doing to prepare for the future. And I think my answer is still. Keep doing what you're doing. Keep working on your design skills. Keep working on your site engineering skills, but also do your very best to stay on top of the advances in AI. If you're not using AI, you're falling behind. So, so before I use the skiing on an avalanche visual, which I

1:24:49Michael Wells: Feel personally, another one that I like is surfing a tsunami. Right now,

1:24:53Janne Parri: No.

1:24:54Michael Wells: I feel like I'm barely holding on. I'm at this level of steepness that the reason I started my local groups is because I can't keep track of the news. I can't test the new models. So I'm outsourcing that. I'm crowdsourcing it to my team that I'm building. And I think the goal is, if at all possible, we want to stay on top of the tsunami. If we're in front of it, we'll get crushed. We're behind it, we get left behind. Right. So the goal is to try to stay on top and it's not easy to do, but I think for anyone who's in an information industry, especially, which we're right in the center of that space as Webflow developers, I think it's a key point of the strategy is just know as much as you can. And even if you have a specialization like animation or video banners, just so much happening there. You got to know.

1:25:48Michael Wells: It and use it. And then that keeps you on top because your clients are like, this guy knows how to do really nice videos at a third of the cost of anyone else we could have produce it. That's awesome.

1:25:59Janne Parri: Yeah, yeah, definitely. Obviously that might feel overwhelming at some point, but you do not have to spend all your nights like trying out new tools, but every now and then when something comes up, just have a go. Five minutes with playing would be nice.

1:26:19Michael Wells: Definitely.

1:26:19Janne Parri: Okay, Michael, we have one weird section at the end, but before that, if people want to follow you, check you out, maybe contact you, where should they go?

1:26:29Michael Wells: The best place is probably in the new Webflow forum. You can find me as memetican and you'll see my little signal logo with the bars. Easy to find. Reach out to me there. Of course, if you ever happen to make it to new Zealand, come visit me. You can be a guest member in one of our groups. And then my website. So it's just signal.com, but signal is spelled with a Y S Y G N A L. Find me there, drop me a line and we'll connect.

1:26:58Janne Parri: Yeah, yeah. I will put all the relevant links in the description. And there's a, there's a high chance that if you ask something on the Webflow forum, ⁓ Michael will answer. But yeah, the last section, do you happen to have an idea what the last section is about.

1:27:16Michael Wells: No, I don't. Surprise me.

1:27:18Janne Parri: This is beautiful. So it is a funny little weird Finnish section at the end. So I will give you a Finnish saying. I will translate that to English and your job is to guess what that means. So it can be a metaphor or idiom or something like that. And I will try to explain, translate and explain.

1:27:50Michael Wells: I'm ready. Let me get that GPT out.

1:28:14Janne Parri: In Finnish, the saying is niin metsä vastaa kuin sinne huudetaan. In English it is the forest answers in the way you shouted it into it so what does it mean if I say to you that the forest answers in the way you shout into it what could it be about?

1:28:28Michael Wells: Hmm and it's a finished it's originally a finished saying

1:28:33Janne Parri: Yeah, I believe so.

1:28:35Michael Wells: And you definitely have nice forests in Finland, so it probably means forests and not something like the crowd of people or something like that. Probably. I would say, I have no idea. You gave me the hardest one, didn't you? Yeah, I'm curious though. Give me some.

1:28:54Janne Parri: This is quite hard. Give me something, then I'll end your agony.

1:29:00Michael Wells: I would have thought it has something to do, I have a lot of affinity for Japanese culture, which has its own sayings and so on. And a lot of them are tied into nature. And so my first thought would have been that something almost like a Zen kind of like life responds the way that you treat it or something like that.

1:29:24Janne Parri: That is actually quite good, quite on point. I would translate that to just you get what you give in a way similar to like you reap what you saw or something like that. Yeah. You were not that far away. Yeah.

1:29:37Michael Wells: Yeah, I like that. That's nice. It's a fancy way. That was luck.

1:29:44Janne Parri: But that's just a funny little Finnish thing I have to do at the end of these episodes. But okay Michael, this was a blast. Thank you so much for being a guest.

1:29:54Michael Wells: Thank you to be on.

1:29:56Janne Parri: And for all the listeners, watchers, we'll see you in the next one. Bye!