Podcast

Dhaval:
This award-winning founder built an AI design product and a user community of half a million users in less than five years. In this episode, we chat about how you can gain a lasting advantage as an AI product creator. We also discuss product development philosophy for anyone who’s interested in building on top of chatGPT 3. Tony Beltramelli is a founder of Wizard, spelled with a U instead of w. Uizard. Helps you build stunning mockups and prototypes in minutes.

Welcome to the show, Tony. Thank you for joining our call. Tell me about your product. Tell us about where you are, how long have you been doing it, et cetera, et cetera.

Tony:
Hey, Dhaval, nice to meet you and thanks for having me. Yeah, so I’m the CEO and co-founder of Uizard. So if you wanna look us up online, you need to look for Uizard spelled with a U instead of a W. and we are basically building an AI powered design tool to make it easier for anyone to basically build products, interfaces for mobile app web apps, you name it. Design is pretty hard. So we we bring AI at the core to just make it easier for everyone. .

Dhaval:
Yeah. Design is the last frontier, right? That’s the part that takes a lot of creativity, a lot of lateral thinking. What was your process, thought process like when it came to building the product and how did you infuse AI in the capability?

Tony:
Yeah. So I’ve been basically like God brought into AI doing my grad studies. I did, I’ve been doing like a few projects back in the days. And then in 2017, back when I was working as a data scientist I was just still tinkering around with AI and deep learning in my weekends. And actually it’s one of these like research project that was laying down the foundation for the company. So the company essentially started as an AI research project, before you even become a product in a company. So it was really like we didn’t have to just back it up, at the end. It was just part of the foundation.

Dhaval:
That’s awesome, man. What was the, when you found a company, wizard, when you define that company? When was it founded?

Tony:
Yeah, of course. So the, kind of like part-time, weekend project that I’m talking about was something I was building back in April, 2017. But it was just honestly like a side project and we only incorporated a company officially in 2018 with my co-founder, so early 2018. It took a long time to build the product, make sure it worked, make sure to solve a real problem, iterate around a customer, and then we launched out of private beta in February 2021. It took a while.Wow.

Dhaval:
Did you launch it on Product Hunt or did you have a list to go from

Tony:
we had both. We had gathered a waiting list of folks that had signed up to our, private beta and Alpha. But eventually, of course, when we were ready to go live, we, we also did a few launches on Product Hunt. We actually won, golden Kitty Awards for best AI and machine learning product in 2021, if I remember correctly. Oh wow.

Dhaval:
Congratulations. And how was the launch? What was the, can you share your, can you share about a little bit about the launch, whether it was sufficient enough for you to bypass the seed round or, yeah, if you can share any of that data.

Tony:
Yeah, of course. So at the time we launched, we already had raised roughly 3.6 million dollar of capital. So seed it took, it takes a lot of juice to just build accompanying product. I think Figma took, what, like four years to, to launch, which is like the same story with us. And so launch, it, it didn’t like it never happened, like the the launch day and then. Skyrocket, Of course, it’s just new spike of launch, you just nurture your user base and it takes time to ramp up. So yeah, it took a long time to just build the awareness build the right network effect in the product to incentivize folks, to invite other folks. But yeah, it took a while. Now we’ve raised, what, like more than 18 million US dollars. And we have, we are serving growing community of more than half a million users. But it, it took a while to get there. Wow.

Dhaval:
Yeah. So that’s half a million users and the whole myth about all of a sudden you are founder of product market fit and you are getting pulled and your servers are crashing. That was like a romantic story that didn’t really happen for you. You had to make incremental improvements that led to finding that fit eventually. Is that what I’m hearing? That’s

Tony:
Absolutely correct. Throughout the. Two years of beta. There was a lot of product iteration. We had to just kill features, relaunch new features, test with customers. It takes a while before you can actually measure and quantify that you have product market fit. , it would’ve, yeah, that’s the road we took. Let’s just measure that we have product market fit before we put this live on the internet. But yeah, even though we measured, we had product market fit, it’s still not an overnight success. Right. It took a while to just get to the first 10 K, 20 K, 50 k, and so on and so forth.

Dhaval: Yeah. That’s interesting. There is a, it’s a gradual process, right? It’s not something that happens overnight and a lot of people have, all of a sudden you’ll find the fit. .

Tony: No, completely. But then when it works, you can actually really see that it works. In the past, like six months, we’ve acquired and served more users than we’ve had in the, in the first year after the launch. it compound. And when the compounding effect works, you can definitely see it. It’s a no-brainer.

Dhaval: Yeah. And the distribution and the right amount of, capabilities creates the pull in the market. So that’s great. Thank you for sharing all of that. Let’s dive into, Your, AI capabilities. Tell me how is it different than other design tools that are out there and, yeah.

Tony: Yeah, so our features are honestly quite unique. You can. You can of course assume that I’m just trying to market our own product here, but, you probably won’t find this anywhere else on any other product. For example, we leverage AI to just enable our users to import a screenshot. So let’s say that you are a product manager at Airbnb and you want to revamp that, onboarding flow when you get new people to sign up. If you were to just open Figma, sketch, any other tool, if you don’t have anything already designed, you have to start from scratch. So what we’ve done is that we enable you to import a screenshot of anything, and then we use AI to just recreate what’s in screenshot, so you can actually then go ahead and modify it to your liking. So you overcome the white blank page problem in just a few. Drag and drop your screenshot modified. There you go. You have your design. These other places where you, we use AI as well. For example. Let’s assume that you are, like me, you know what you want to build, but you’re not the greatest designer. So you can tell our ai hey, you know what? This is great, but it looks pretty bad. Can you please just copy and Paste the style of, I don’t know, Twitter and make it look like Twitter? And so that’s also a place where, our AI can just automatically do this, pull the style of anything and then apply it to your project. . Another example would be to, you, you can brainstorm ideas on paper or on the whiteboard with your team, and then you can just snap a picture of whatever you sketched. And then our AI will transform this automatically into a design that you can then modify. So it’s really all these features are, is all about like, how can you, can we just simplify the ideation flow to make people. Focus on the core value, right? I’m trying to solve a problem with this design. I don’t want to get lost into the weeds of like how many pixels to the right, how many pixels to the left? Should I move that button? It really doesn’t matter when it comes to ideation, and that’s kind of like what we’re trying to do with ai.

Dhaval: Wow. I see you have a lot of differentiation within the product. One, one biggest challenge you are solving is the initial creativity block. That Some designers experience when they have a blank page and you saw that biggest problem, right? So now you have some something to play off of. And that’s, that’s awesome. When did you build the foundational AI model or ML model that serve the backbone of the product? Like how did you decide the use cases that, yeah, we are gonna solve for this use case and we are gonna use this AI capability. What was that process like? Tell me about the first use case. You decided to tackle with, ML?

Tony: The first use case was really like , the foundation project that I mentioned earlier that kind of gave birth to this company. And it was so just context on this when I was like my first tech job, when I was still an undergrad was front and developer. And I just hated the part where you need to take the design work and transform it into code with HTML and css. And so back at that point the idea was, hey, can we use AI to just take a picture of the design? And then turn it into code automatically. , the developer can just focus then on implementing the business logic. And of course, this feature was launched and eventually not used by our users, so we just killed it. But the core foundation of trying to understand the user interface, buttons, forms, what is an image, what is text field, what is a login screen? All this core foundation is the same that power or tech today. It’s all about okay, we are gonna be operating within design, within user interface design. And if you . Teaching a machine, the basic concepts of that domain, then you can reuse it for all sorts of application. Like all the features are ultimately based off the same core whether it’s for styling page generation, rename it, it’s kind of the same core. so as a general principle, I would say if someone want to build a product for, I don’t know, the automated industry, focus on what are the overall concept. Of the automative industry that you wanna work with, and if you can kind of like nail learning these concepts, then you can actually build multiple application layer on top without having to redo the core every single time.

Dhaval: So a new AI product creator or a product creator who’s thinking of adding AI to her portfolio or existing product, the advice will be to learn the concepts, the industry, the inner workings of a domain. And then once you understand that you can turn that into a use case and add AI capabilities to those use cases very easily with today’s technologies. Is that, am I hearing that right?

Tony: Pretty much think about, GPT 3 the reason why GPT three is just so amazing at, you know, being versatile. Is that ultimately they didn’t try to teach GPT to, I don’t know understand emails or understand customer support tickets. They focused on training a machine to understand English or language. And then once you kind of like nail that core foundation of understanding language, you can then do other things like customer support, email, copywriting. So like the same principle, like what are the building blocks of the domain you’re trying to target and automate with ai, nail this. And then the application layer will almost, become, a no-brainer on top of that core model.

Dhaval: Would you say that with this new transformational technology that we have found ourselves being able to play with, would you say that for a product creator, domain knowledge has became. A lot more important now than it was before because a lot of generic stuff, all the typical product work that people used to do has been commoditized. And what remains is the product creator’s ability to understand the domain and tease out the relevant use cases. Would you say that’s a valid hypothesis?

Tony: I would actually agree. , and I would even say UX, the quality of the product around the AI is probably even more important than the core ai. I mean, if you think about chatGPT, which we just mentioned, OpenAI launched GPT3 in what, like 2020. There was like no traction. A bunch of, researchers, engineers were playing with it and had fun leveraging it for different products, but it’s only when chatGPT came that, you know, in five days they had a million active users. But it’s the same model. It’s still GPT three under the hood. The reason why it’s scaled so rapidly that the quality of the product. Leveraging this core AI was just mind-blowing. You could interface with it, chat with it as if it was like a human. And that makes the core difference. So as you said, like for new product creators, it’s all about like the domain, the product that’s gonna be, that people are ultimately gonna interact with. They’re not gonna interact with the ai, they’re gonna interact with the product. That’s the ai. And if you nail this. Then you win basically, it doesn’t matter , how smart or, it’s secondary. the quality of the secondary, we even say,

Dhaval: Yeah. Let’s change gears a little bit. Thank you for sharing that. One question I have for you is that you seem to be fairly technical co-founder. How would you recommend people who don’t have that sophisticated technical background, what do you recommend them to break into the AI product creation space? Or is there a learning curve there? Can we use the existing LLMs, the large language models, their APIs or do you recommend fine tuning them? And then if there is process of fine tuning, what is the learning curve there? Like for non-technical person?

Tony: This is where probably gonna have, very different, opinion. Yeah. , but what I would say is that, Ultimately, it’s very easy to build something around, on API, as you mentioned. Well, easy in terms of getting something up and running. It’s really hard to build great product around it. , but so I, I guess the core, whether you are technical or not, is what’s your differentiator that makes this product unique? Because they’re gonna be like so many other products. And if yours was easy to build,then how do you defend against copycat? Right? You need to find a moat, whether it’s, the quality of the product that’s just so much better, whether it is a community of users. , so and ultimately the, the way companies win is not just by having great products, is also having great distribution and a way to just put this product in the hands of users. So, If you’re a non-technical person, you might actually be really shining at the distribution part. And ultimately this could enable you to be successful, even if you can’t, master the higher level of technicality that the your product offers. , so that would be one distribution ultimately decides who win and who doesn’t. I would say

Dhaval: that’s great. So distribution is one of the most underrated. Pm , product creation capabilities. And once you find the distribution, the important thing is to be able to iterate on the feedback you get from the users who are experiencing your product. For your situation, did you have a big team to support you with all of those iterations? How did you find that initial fit when you were going through those iterations?

Tony: well initially how did we find our first customer? How did we find our first users to get feedback from? Because we had basically built some very transformative tech. Just having a few demo videos online was actually enough to just initiate enough interest that some people will start signing up to a waitlist and from that waitlist, we could just then go and recruit people to just access the beta, the alpha and learn from their feedback. so I guess there is always like early adopters that are looking for trying new novel tech. , and so AI has this just magical effect that people certainly wanna try, right? So you can totally leverage this to your advantage and collect, signup, email and eventually get the first 10 or 100 people to try your product and iterate from.

Dhaval: Yeah. That’s awesome. In your case, what was that moat when, earlier you mentioned that you had to find a moat, whether it’s a technical advantage, user experience, community or distribution. In your specific situation, did you find that mode, through your product, through your distribution? What was that like?

Tony:
Well, I guess the moat is something that we are constantly building, building and refining, right? Because as a company grow there, the moat changes over time. Initially it might be the. The deep knowledge that the founding team has for a specific problem that no one else understand. Like this is your unique selling point that no one else has and can replicate, but eventually the quality of your tech, quality of your product, quality of your community. But , early on we had no moat I mean, when we were, first building this company, if anyone had the same core interests about a specific problem, we’re trying to solve it would’ve been fairly easy for someone to replicate it , if they started at the same point in time. So early on, I would say your moat is like how deeply you understand the problem you’re trying to solve. that makes it hard for anybody else to just, solve as well as you would but over time, for us it became clearly the product and the community. We’ve had really put an emphasis on community led growth which ultimately it’s hard for other companies to replicate. plus the core technologies, which is still not available in any other design tool. , we still are the only one with the features that, I mentioned earlier.

Dhaval: Yeah, so it looks like you found moat in two places. One was in the community led growth, and the second was in your core technology advantage of understanding the problem and identifying the most foundational use cases like initial writers or initial creative blog. Looks like you really have an amazing roadmap ahead of yourself tell us a little bit about where does, where is your vision? What is your vision of the product?

Tony: Yeah. The vision of the product is really to enable anyone within the product team to take an active part in design processes. Design is increasingly important. You can probably, like, every product you use, you will immediately judge this product based on the quality of the design and the UX of this product. You might actually just not even look twice at that new app you just install if it doesn’t feel good to use. But the challenge is One Figma is really hard if you are not the designer. The number of time I meet a manager or team lead that doesn’t know how to get around Figma is just insane. And two, even the few folks that understand how to use Figma well, they’re already pretty busywith all sorts of design work and product discovery. So our vision. How can we just empower all the other people in the team that don’t have the time or the skills to use Figma to still take an active part in product design? Because chances are is not just the designers or the product team that have good ideas. The customer success are team is hearing feedback from customers every day. They have some kind of nutrition on how to fix it. And so if we provide them a tool that enabled them to. Materialize the vision that then ultimately they can help the product team, move faster towards solution and everybody wins. So core vision is again, like put design in the hands of as many people as possible so companies can build better product faster.

Dhaval: Thank you for sharing that and we’ll end on this question. What advice do you have for product creators in new to the AI space or existing product creators in the AI space who are trying to excel

Tony: it might not be the best thing for a lot of folks to hear, but you need to get deep into the technical details. You can build a lot of things around API, but ultimately you don’t wanna build a product that would just die if OpenAI decide to just turn off the api. And so if you wanna build something at last, you really need to understand what’s going on under the hood. And so I would say, Get at least an intuition of what’s going on. So you understand at least you know how to move things, ar how to move forward how to recruit the best engineers and how to have more proactive process around improving your product. A great place to start would be the FAST AI course, which is free maintained by Jeremy Howards which is an amazing place to just you. Get into ai really quickly and for free. So go ahead and, and learn the deep stuff.

Dhaval: Great. Thank you so much for being on the show, Tony. We learned a lot from you. Thank you once again.

Tony: Thanks for having me dhaval