AI for Growth: Lessons from Leaders Howard Zonder

Megan Driscoll: Welcome to AI Growth Lessons from Leaders. Today I have with me Howard ⁓ and super happy to have you on the show. Excited to chat with you more. I'm gonna you just maybe tell a bit yourself a little bit about your business.

Howard Zonder: Sure, thanks. It's really a pleasure to be here. I appreciate you me a platform to talk. So my background is, I don't come from a technical background.

My major in college was East Asian Studies and Chinese Language. And then you might ask, what do do with a degree like that? And so naturally the answer is you go in the Marine Corps. So that's what I did.

And so I did that for a while. Megan Driscoll: love that. Howard Zonder: and got out, went to graduate school, and then got into my career. And I ended up going down a technology path.

I ended up in the telecom industry, probably because I was a communications officer in the Marine Corps. And ⁓ I got hooked up with some very entrepreneurial guys, and we created new products and did some really cool stuff over the course of my career. and then ⁓ ended up in the ⁓ events industry for about five years doing ⁓ running global marketing for an events company, trade shows, ⁓ conferences and things like that. And then when that was coming to an end, I was kind of figuring out what I was going to do next in my life.

And the idea of going back to another corporate job was just, it wasn't working for me. So I ended up looking at some side hustles and I ended up becoming a real estate investor. And I went down that path and I was actually still looking for a regular job because I had a young child and I thought maybe it would be good to have some ⁓ health benefits. But in fact, that didn't work out.

mean, what worked out was the side hustle really worked out. so I started, my business started growing and I got to the point where I couldn't really handle it on spreadsheets and Post-it notes. So I ended up building a CRM system. I call it a CRM system.

It was kind of a workforce management system too, because it basically automated the entire business. And that was my first introduction to building a product of my own. And I started showing it to my friends in the industry and they liked it they said, hey, we want that. And so that was my first software business.

So I built that out and ⁓ the structure of that industry has changed over the last 10 years. So it became a little bit less enticing and I had some time on my hands. so around January, ⁓ Megan Driscoll: Cool. Amazing.

Howard Zonder: early January, maybe even late December, I started thinking, I really, I really need to figure this AI thing out because it's a juggernaut. And if I miss this, if I miss this one, you know, I'm not going to be viable. So for me, the best way to learn something is just to do it. And so I said, I'm going to build an app.

So I just had no idea what I was doing. And I just started building an app. Megan Driscoll: Yeah. Howard Zonder: and ⁓ I built it and ⁓ I liked it a lot.

I shared it with my friends and family and they didn't like it quite as much as I did. But I learned a lot and so that got me started and then it created the impetus for many conversations. And through that process, I came up with another idea, which is now the business that I'm pursuing right now, which seems to have a much stronger opportunity than the first one did. So ⁓ it's geared towards real estate agents.

The challenge with real estate agents and really any independent professionals is that they need to build their authority and their brand in order to bring customers on board. ⁓ Megan Driscoll: Cool, cool, what is the business? Howard Zonder: The challenge with social media is that ⁓ you're really competing for attention in that feed. And that feed goes quickly.

if they're not stopping on it and associating it with the person who posted it, you're really not getting a whole lot out of it. But what really builds authority, and I think you are a perfect example of this, is a podcast. When people can hear your voice, they start to know you. There's this thing called parasocial connection, and they start to build a relationship with you, even though it's in through this digital modality.

And so what this tool does is it... allows a, I'm focusing right now on real estate agents. It allows them to build a short podcast, ⁓ without any special equipment, without a studio, without any technical knowledge, all they really need is their phone, maybe some earbuds. will upgrade everything to studio quality.

It'll do background music. It'll put the intros and the outros on and make it sound perfectly studio quality. And you don't have to know anything about technology to do it. Megan Driscoll: Cool, really cool.

Well, I think that one of the things that I love about doing this show is I get to talk to a variety of different business owners and we always try to find something very specific that I think the audience can benefit from. And I think there's a lot of people who are interested in the app, in creating an app, right? And I think a lot of people have really good ideas, but then they don't really know where to begin. So when you told me about the app, I was like, oh, that's something I think we should talk about.

So I want to talk about... and what I love about it is like, this is also recent, like you decided you want to do this in January. You know what I mean? Like that's how easy it is to do to create something in January.

⁓ and because it's so quick and easy to do, it is the kind of thing that if you create an app and like you said, like nobody liked it as much as you did, well, you just create another one, you know, like in the process of creating one, you've kind of already like done the hard work of like, how do you do it? So. Howard Zonder: Yeah. Night.

Megan Driscoll: Could you walk us through like, was the thing you were trying, what was the problem you were trying to solve with the app? Like how did it get started? Howard Zonder: The current one was interesting. So I was actually presenting my first app to my former head of technology from my last corporate job.

He's over in London and I could just look, I could see on his face, he was just not all that enthralled with it. And we started talking about it he said, well, I have this idea that ⁓ for a ⁓ bedtime story app. where you could just go in and say, hey, my child is this age and this gender, and this is what they're really interested, whether it's space or dinosaurs or whatever. And this is the lesson I want to inject in the story.

And this is how long I want it to be. And then it'll just write the story. And then you can clone your voice or your mom's voice, your dad's voice, and then that voice will read it. So if you're too tired at the end of the day to read your child a bedtime story, they'll get a custom-made story.

And it could be your or you're traveling, yeah. But it could be like your grandparent, the grandparent's voice, the grandparent's on the other side of the country, but they could still read the story. Exactly. So I thought it was a pretty tremendous idea, but as I looked into it, I realized there was a zillion bedtime story apps, and it wasn't gonna take that long for somebody to figure this out.

But I liked the idea. Megan Driscoll: or they're traveling, know, or they're just, they're not home, right? or the grandparent has passed away. Howard Zonder: And then my background in my career has always been B2B.

And just it occurred to me that all of these independent professionals have this desperate need to build their authority and build their brand in order to bring customers on. I could take really the same technology stack and apply it into this market and there'd be a great need for it. And I think it would serve them well. So that was really the genesis of the idea.

Megan Driscoll: Okay, so then what, tell us exactly, I mean this again, just getting into the like nitty gritty, like what did you actually do to create the app step by step? What did you do first? Howard Zonder: Well, the first thing I did is I tried to build his app just for fun, just to. Megan Driscoll: But even like beyond like the idea piece, like you physically went to what program.

Howard Zonder: Yeah. So, so the first thing I had to do is I had to figure out, okay, how do I clone my voice? How do I, how do I take that and, how do I use AI to write a proper script? Then how do I get the app to use the clone voice to read the script?

And then we got that piece done. And then it said, well, you know, a podcast. Okay. Megan Driscoll: But go back, hold on, stop, I'm gonna stop you.

Go back, how did you do these things? Like, what program did you log into and say, like, were you writing code with Claude? Like, what exactly did you do? Howard Zonder: So I started with a tool called Abacus, something, and it seemed like a reasonable thing to do at the time.

I've moved off of that since. It turned out to not be the best platform for what I was doing. ⁓ It turned out to be very expensive. And it just was, it didn't have quite the same level of capability as Claude.

And I had never used Claude. Megan Driscoll: Why did you not like it? Mm-hmm. Mm-hmm.

Howard Zonder: But I started to figure out what Claude could do and I knew everybody else was using it. I just, it's just amazing. Absolutely amazing. So I migrated everything off of Abacus onto Claude, but the process, it doesn't really matter which one you use.

The process is very, it's similar. Although I think Claude's a little bit better at it. So, ⁓ and this was, there was a lot of evolution in my learning at the time. So my early days, Megan Driscoll: Yeah, and cloud's just getting better and better.

Howard Zonder: I probably took, it took five times longer and cost me five times as much money to do what I could do today for the same, to achieve the same thing. ⁓ It started out as I would say, this is what I want. And I. Megan Driscoll: Okay, so you open up Cloud.

just, again, for people to just, cause I think, I think, yeah, it's like taking the, I, people have struggled to be like, well, how? Like, okay, so you open up Cloud and you basically say, yep. Howard Zonder: Yeah, I don't put up with that. This is what I want to do.

This is... So I'd open up Cloud and I go to Cloud code and I would say, this is what I want you to build for me. And then it would build it. And then I would look at it and then I would say, well, that's not quite what I want.

And then I would go and it would just be this incredible iterative process, which is very expensive if you do it that way. It uses a lot of tokens. So what I've learned over time is ⁓ one thing is you can create these Megan Driscoll: Yeah, it sure does. Howard Zonder: documents for AI, they're called .

mds, which is just MD stands for markdown, but essentially it's nothing but a text document. And so what I do is I would go in there when I migrated the app from Advocates over to Claude, the first thing I told Claude to do is I said, I want you to create a technical architecture document for me of how this app works. And it did that. You know, I'm not technical, but it did a great job at it, I think.

And, and I would always, whenever I have a clog, create a document like this, always make it, I have it create two documents. One is a word document because that's how I like to read it. And then the other is the dot MD, which is how AI likes to read it. And I put them in separate directories and.

Megan Driscoll: So far as you know. Howard Zonder: It's important. is that's it's a little, it's a little bit of a nuance, but it's important because a word document that we would consume is all formatted and it looks nice. Well, all that formatting behind the scenes is a bunch of code.

And so if you ask AI to read that document, it's got to pull the relevant information out from all of the formatting code, which just means you're eating more tokens and spending more money. Whereas if you keep it in a dot MD, it's faster and it's a lot cheaper to do it that. Megan Driscoll: That's actually a really good tip. What you've just said, think, for people to hear ⁓ is because I do think that a lot of this stuff gets expensive and people don't necessarily know that or consider that.

You know what mean? As it relates to, just so I don't forget to ask you this question, because I think it's interesting, ⁓ how much money do you think it costs you to build your first app that in the end wasn't anything you wanted to really pursue? But roughly, what What should someone expect to spend in developing an app? Howard Zonder: Well, if you're going through the whole learning curve, it probably costs between $500 and $1,000.

Megan Driscoll: Okay, all right. If you're going through this iterative process and for the first time doing it. So you could pretty much assume you're gonna invest between five and, and at the end of the thousand dollars, did you feel like you had something that was marketable or no? Howard Zonder: Yeah.

I thought I had, well, I had something that was totally usable and I used it and I thought it was great. And some other people liked it, but it was not marketable, but it was a fully functioning app. Yeah. Megan Driscoll: Okay.

Great. Great. And you could download it on the App Store. OK.

Howard Zonder: Well, no, no, no, no. So I wasn't creating a mobile app. Well, I created a web app that you could, so you could use it on mobile. worked fine on mobile, but that's, it's a really good point that you just raised.

So in my case, what I'm building right now is a, is an app that say a real estate agent or financial advisor, whoever's going to use. So I purposefully built it as a web app that was, and that's one of the first decisions you have to make. Megan Driscoll: A web app, okay. Howard Zonder: is what platform are we going to run this app on?

Is it going to be through a browser or is it going to be for the app store? It's going to be a mobile app. And the reason I came up with or made the decision to do it on a browser is because I wanted it to work really well on a laptop because I think that's where people are going to create podcasts and do a lot of the work. Megan Driscoll: For sure, it's very hard to do that on the phone for sure.

Howard Zonder: And the vision, my vision is a bit larger and I can imagine at a time where I'm going to have an app where people can consume all of these podcasts because they're structured a little differently than the normal podcast. So if I can get the listeners onto my app, then there's going to be a great amount of data in Intel I'm going to be able to get from that. That will be, excuse me. Megan Driscoll: You'll turn it into a hosting platform.

You'll turn it into a hosting platform. Yeah. OK, got it. So you've got a vision for this.

So when you think about, we talk a lot about, and I've actually had some really great people on the show talking about prompting and how important prompting is ⁓ and the structure in which you do it. just being, even if you. Howard Zonder: Yes, and that will be a mobile app. I'll develop that one differently.

Yeah, yeah. Megan Driscoll: Pick any structure that's, know, pick a structure and then use it, essentially. But it's got to cover like a lot of different things like what's the format, you know, what's the, what's the, what are you trying to answer? What's the tone?

All the things. Do you think there's a specific set of prompting that works better for when you're trying to develop an app? Is it different than when you're prompting it to help you build a document or create a funnel or whatever it is that you're doing? Yeah.

Okay. Howard Zonder: Absolutely. ⁓ And this has been a huge learning curve for me. again, it makes a huge difference in cost.

So what I did at the beginning was I said, this is what I want. Megan Driscoll: Great. Okay, tell us, us. I love this.

Howard Zonder: And then they would create something for me and I would say, that's not exactly what I want. And there was this iterative process. Well, every time it has to create a new, a new version, it's just turning through tokens, turning through dollars and every, you know, and all of sudden it would say, sorry, you're out of tokens. If you want to keep working, you know, put more money in the slot machine.

So, and I can't stop, right. It's like, I'm so many hours left in my day. I'm like in the middle of this. So yes, that was.

Megan Driscoll: addictive. Yeah, no, you can't stop once you've started. It's impossible. Howard Zonder: So then let me just start with some important, I think, philosophical foundations of prompting.

⁓ AI is, ⁓ it will do whatever you ask it to do and it will find a way to do it. And if you haven't given it enough information, it will make up the difference. It'll just make it up. And so when we talk about AI hallucinating, A lot of the time, the reason it's hallucinating is because you have not provided adequate context and it's just filling in the blanks.

There's a story, which I believe is true. There were some hikers up in Canada and they were trying to their way back. It was getting a little later in the day and they saw weather coming in and they needed to get back to the trailhead. And they went on to AI and they said, what's the best way back to the trailhead?

And AI answered them. and answered them in the best answer possible. But what they had not done is they had not told them what season it was. They didn't tell it what time of day it was.

And so it ended up taking, it turns out that they were in like fall or winter, and it assumed that they were in summer. And they didn't tell it was the end of the day, so they figured it was noon. It ended up sending them down a path that was completely impassable, and they had to get helicopter evacuated out of there. So it's really important to provide adequate context.

And when you're creating an app, that means that when you talk about what it is you want, you have to be as specific as possible. So for example, my app is built, but what it doesn't have right now, and what I'm going to probably work on after this call, is I'm going to put a support capability into it so that the user can reach out for support. So what I'm going to do, and I'm not just going to say, at my app and I need a support function. That would be a disaster.

So what I'm going to do is I'm going to say, I need a support function. And what I want you to do is I want you to put a link in the left-hand menu above the feedback link. And when you click on that link, I want it to open a modal, which is just like a little window that is on top of the primary window. And these are the fields I want in it.

And when they hit submit, I want it to go to this Google sheet to populate. And I want it to send me a text notification that I've got a support request. And I want, then, so, so I'll start with that, right? But I won't even do that in Claude.

I'll actually go and do that in Chat GPT. Megan Driscoll: Okay, why the switch? Howard Zonder: because I don't pay for chat GPT. use the free version.

So I'll go in like that and I'll say, I want you to take this and I want you to ask me questions until this is a fully formed feature and every permutation is figured out because there's things I probably have missed. then we'll all iterate with on the free version. And then when I feel like I've got the feature completely figured out, then I'll say, Now I want you to write me a prompt for an AI coding engine that will build this. And so then ChatGPT builds my prompt for me for free.

And then I bring it ⁓ into Cloud. But then I add a few things. So one of the things I always add are ⁓ Don't build this yet. Take this prompt and build for me a plan of how you're going to build this.

And if you are not 100 % certain of what it is I want, I want you to continue to ask me questions until you are 100 % certain. Megan Driscoll: That's actually a really, so just like I tell my, when we talk about this, make sure you tell Claude, tell me all the reasons why this won't work. Tell me all the things that are gonna go wrong, right? Because I think that obviously it's designed to give you answers in the positive.

So it doesn't really want you to, it doesn't really want to tell you that things aren't gonna work. It wants to tell you things are gonna work and that it's telling you the right thing. then, so I think this is a really important piece to kind of add in there to say like, Howard Zonder: That's right. Megan Driscoll: If you don't think you can do this, you keep asking me questions until it is very, very clear.

And I love that. That sounds like that's great. OK, so when you do that, yep. Howard Zonder: So I do that.

And then I add one other piece, is now because it's in Claude, it actually knows my app and it knows my branding and how the app is branded color schemes and all that. So then I'll just say, make sure that the implementation is consistent with the branding of my app. And then I always add this line and make it incredibly intuitive, fun to use and Apple like. Cause Apple to me is the creme de la creme as far as usability.

And so now it understands, it's got all this context to understand not just what to build, but how to build it. so now, and then even then it's not building me the app, the feature yet. Now it's just building the plan. And now I look at the plan and I read that.

And if it's fine, then I say, go ahead and build it. Or if it's not, then we go back and we iterate a little bit more on the plan until I feel that the plan is perfect. And then I hit the go button and then it builds it. Megan Driscoll: Mmm.

Howard Zonder: And now I've eliminated all of this iteration and I've done a lot of the heavy lifting on a free platform. Megan Driscoll: Tell me this because I think people, again, I did this podcast exclusively almost to kind of demystify AI, to kind of take it and say like, AI can do everything and then really nothing at all if you don't know how to do it. You know what I mean? It's only as good as you're willing to use it and understand how.

And so I think a lot of people talk about this in sort of big ways. And I love that we're getting really granular here. I want you to really lay out for me How did you take it from Claude Code, right? And I mean, I know how you did it, but I want listeners to know from you how you did it.

How you took it from Claude Code and how you actually got it onto a website where it was live. Howard Zonder: That is a great question. It's a really important one. And it's one that I had to figure out when I did the first app.

I was very excited. I have an app, right? And then I have all these other people that are using it, but I didn't have an app. I had a, I had an app on a development environment on my computer.

Well, it was actually running on abacus, which was very expensive, but it was really run out of my computer. That is not a business. It's a hobbyist tool. So if you want to take an app, when you take it from ⁓ something that you've built that works and you say, okay, now I want to turn it into a business, you have to think of things like security.

Megan Driscoll: Yep. Sure. Howard Zonder: and scalability. Okay, it works for one person.

How's it gonna work when there's 100 people or a thousand people or 10,000 people? Is it still gonna work? Okay, I have a database. Do I need to back up that database just in case something happens?

⁓ Am I compliant with privacy requirements with any data that I'm collecting? There are a lot of additional issues that have to be considered and... in order for it to run 24 seven and be available to people with reliability, it can't be running on my computer. It's got to be on something else.

And so this is where we talk about developing an app. If you want to develop a app for a mobile phone, fine, develop an app because what you're going to end up doing is it's going to be the self-contained thing and you're going to push it through the Apple store or the Google Play store and that'll be fine. Megan Driscoll: Right, right, exactly. Howard Zonder: But to do an app like what I'm building is bigger than that.

And so at that point, I had to recognize my own limitations. I'm not really a technical person. So I actually had to, I had to partner with somebody else who was much smarter than I am. And we built out an environment on Amazon web services and with, with proper database setups and things like that.

Excuse me. Megan Driscoll: Okay. Great. And how do you like Amazon Web Services?

Just curious. How do you like them? Howard Zonder: ⁓ I have no real opinion about it. I just went with the advice of my developer.

And I mean, I've known about it forever, my, from all my reading, it's a more expensive option, but that's because it's a very rock solid option. Megan Driscoll: Mm-hmm. Yeah. Yeah, I mean, I've heard, I mean, there are other places you could go to do this, what I call like the hosting of it, if you will.

But ⁓ I have also heard good things about it, but it is expensive. Talk about the expense of it, the ongoing. Howard Zonder: Well, right now it's not that bad because I'm right at the beginning of launching. I mean, for me, it's probably I'm paying maybe $75 a month in that general area.

And I've actually created two separate implementations on AWS. So what it is, so on my computer is my development environment. Megan Driscoll: Great, just to host it. Okay.

Howard Zonder: Then we upload it into what we call our test environment on AWS because so when you implement it on AWS, it's a little different. There's things that have to be done that are beyond me in order to make it work. Then we do that and then we test everything in that environment. And when that works, then we push it to our production environment.

So two environments on AWS and all of that's right now only costing me about $75 a month. Megan Driscoll: Great, but you're paying a developer to help you do this stuff. Okay, and where'd you find your developer? Howard Zonder: Yes.

What are my best friend's songs? Megan Driscoll: Great, super, that's even better. I was gonna say, I always send people to either up work or fiver if they have something, that's usually a good, but if you know somebody, obviously that's great too. And I do think that this is something that people are talking about how hard it is for kids who are graduating today.

think, I mean, my daughter's graduating next year, so I'm not quite there yet, but I do get like a little like, what she, I mean, she can come work for me. And she may very well actually, giving her, doing a summer, she's gonna have summer internship with me this summer. ⁓ And because basically what I do is, you know, there's so much you can do it with AI. But what I essentially do is I kind of help companies.

I basically they get a comp, they get a company like in 90 days, they have like a fully functioning company. Essentially, I build their CRM for them, create all the automations. And then I load it with like about 5,000 contacts that we've already done lead gen on sort of ideal, their ideal potential partner or event or clients either B2B or B2C. And then we kind of like they.

are able to kind of go from there basically. So there are usually companies that have been existing with a small number of clients or they're a growing business and they need more clients or they need a place where they can mine their clients 24 seven. So I have someone who works for me who's an expert in go high level which is my favorite Sierra and that's the one that I like. So we work together to create like I said it's kind of like a company in a box in 90 days you have a 5,000.

Howard Zonder: Mm-hmm. Megan Driscoll: potential clients that you are now mining for business. And also we've created automation to your existing clients. but developing, think is a space that some of these kids could graduate with, because I do think there's going to be a lot of people who are going to need, like you, who have the entrepreneurial spirit, have the income to do it, have the interest, and then just need like the backend support to get it like up and going.

So I think that's really cool. So then, okay, so let's just talk about, so so just kind of round this out. So you have the app. It's online, you're hosting it, and then you're going to create kind of a marketing plan, I assume.

OK, super. And then you'll be marketing it from there, basically. I love this. Howard Zonder: And all that's being done with AI too.

Megan Driscoll: Yeah, of course. mean, absolutely. Yeah. Well, let me know if I can help you because I love that's my get in front of a bunch of real estate agents.

That's my jam. Actually, that's my jam. actually have it's interesting. have a few clients who real estate brokers are their primary source too.

So I already have really good database of that if you're interested. ⁓ So yeah. Okay, great. So in the end, if you were to like ballpark it, someone has a great idea.

Howard Zonder: Well, I may very well be calling you for that. Good to know. Megan Driscoll: ⁓ They kind of test it on their, they create it, they sort of test it with friends and family. Let's say they get very positive, like this is really great, I really want this.

So they kind of know, and then they go forward and they kind of do it in the way that you've done it. I think AWS is a great platform to use for people who aren't technical, and I just think it's, like you said, it's solid. It's not gonna disappear. And it's gonna cover your bases with regard to like legality and all the things.

⁓ What do you think start to finish it, what it's gonna cost you before you start marketing it? Howard Zonder: Mm-hmm. Well, it depends how complex the app is. If you're doing a mobile app, I don't know because I haven't built one of those.

To do something like what I'm doing, $1,000, it depends on how complex it is. I probably have spent over $1,000. If I was to do it again, I probably would come in under $1,000 because I'm smarter now than I was then. Megan Driscoll: Well, some of the tips that you gave us actually are ways in which to cost save and again using chat GPT as much as you can or free version and then moving to to cloud for code.

And then also, like you said, having these two separate documents so they're not always constantly having to use tokens. I think those are really great tips. So ⁓ great. Well, this is like then I love this show because I always learn like something new too.

And I hope that my listeners ⁓ also kind of feel that way that they get a little nugget. So anybody who's interested in. and creating an app on their own, this is like the perfect podcast for them. This kind of gives you a how-to, which I love.

It got really granular and some very specific support and I loved it. So really appreciate you coming. Where can people find you if they want to ask you questions or want to know about your app or whatever? Howard Zonder: They can find me on LinkedIn, just LinkedIn slash ian slash Howard Zonder.

⁓ Or they can send me an email at howardz at ear snacks.com that's ear snacks. We're making five minute podcasts. They're little snacks for your ear.

Megan Driscoll: Ear snacks. Like ear. ⁓ my god, that's really cute. Okay, and your last name is spelled Z-O-N-D-E-R in case you're not watching on YouTube or you can actually see your name.

But ⁓ it was really great to have you. Thank you so much for the time. It was awesome. Howard Zonder: It's been my pleasure.

Thank you so much.