#2292 AI Automation that makes cold calls

Summary

Startup Stories - Mixergy

Episode #2292: AI Automation That Makes Cold Calls
Host: Andrew Warner
Guests: Yevgeny Matce & Aiden Richards, Co-Founders of Resora
Date: January 8, 2026

---

Episode Overview

This episode tells the story of how Yevgeny Matce, a former real estate broker, developed AI automation to handle the grueling task of cold calling for real estate leads. Frustrated by the inefficiency and emotional drain of manual cold calling, he leveraged his computer science background to create a calling AI—initially for his own use—then rapidly realized its appeal for other brokers. Alongside co-founder Aiden Richards, Matce scaled the solution into Resora, a growing SaaS business that automates outbound calls for brokers. Their experience offers insights applicable far beyond real estate, signaling opportunities for automated outreach in many other industries.

---

Key Discussion Points and Insights

The Cold Call Problem in Real Estate

• Grueling Work: Yevgeny describes a typical day filled with hundreds of cold calls to expired real estate listings—homeowners who struggled to sell with previous agents ([01:32]).
  • “I would cold call those people because they have a higher chance of relisting.” — Yevgeny [02:05]
• Low Conversion Rates: Despite the targeted approach, actual success rates were dishearteningly low.
  • “If I were to estimate, 1 to 2%. 1%.” — Yevgeny [03:10]
• Inefficiency: Often required hundreds of dials to get through to a single relevant prospect.
  • “You can call 200 people before you actually get connected to someone that is the correct homeowner.” — Yevgeny [03:26]

From Broker to Builder: Birth of the AI Agent

• Technical Pivot: Yevgeny, with a degree in computer science, decided to automate his least favorite task by experimenting with early AI voice agents ([03:56]).
  • “After two years of cold calling, I got very fed up with it… when AI voice agents were becoming a thing. So I’m like, oh, wait, I can implement this.” — Yevgeny [03:56]
• Initial Results: Within a day of deploying his prototype AI agent, he booked his first listing appointment ([05:10]).

Rapid Growth and Early Sales

• Explosive Demand: After launching a video ad showing the AI at work, interest surged instantly—overwhelming what was initially a manual, custom setup ([08:54], [09:07]).
  • “As soon as we launched the Facebook ad campaign, it took us less than four days to get our first customer.” — Yevgeny [07:51]
• Monetization Model: Early pricing included a $1K–$2.5K setup fee, $500/month fee, plus $0.20 per AI conversational minute ([00:48]).
• First Revenue Milestone: "$40,000 in 40 days." ([00:44])
• Manual Onboarding: Each customer workflow was tailored, requiring weeks of manual effort and testing per client.
  • “We had above 80% conversion rate...but it was me actually doing a Google Meet with them, walking them through this, and closing the sale myself.” — Yevgeny [10:45]

Insights on Sales, Scaling, and Agency vs. SaaS

• Bottlenecks: Customizing and tweaking each agent for multiple clients quickly became unmanageable ([24:14], [24:47]):
  • “The worst thing you can possibly do, in my opinion, is give them an okay version and not a perfect version.” — Yevgeny [11:40]
  • Even minor tweaks required direct meetings and manual intervention, making scale impossible for a solo founder.
• Transitioning to SaaS: The team recognized the need for a self-serve platform with customizable, yet robust, AI agents—facilitating true scalability ([19:12], [27:45]).
  • “There’s no solution on the market right now where it allows a non-technical user to launch their AI agent and get it going in less than 10 minutes.” — Yevgeny [19:12]
• Agency Model vs. SaaS: Agencies can offer more bespoke services, but require constant manual work and are vulnerable to SaaS competitors ([26:25], [27:23]).

The Technology: Building AI for Voice Sales

• Early-Stage Workflow: Built with off-the-shelf tools like Vapi (voice AI), Zapier for automation, and CRM integrations ([13:23]).
  • “You’re going to be pretty much set to go with Vapi, Zapier, and maybe a CRM connection.” — Yevgeny [13:23]
• Custom Fine-Tuning: Transitioned from generic LLMs (like GPT) to building proprietary models specifically fine-tuned for phone sales to increase realism and conversion ([15:36], [16:07]).
  • “We take real conversations, we turn that into fine-tuning data...and we run supervised fine-tuning on a large language model… so it has more salesmanship in its responses.” — Yevgeny [16:07]
• Open Opportunity: The process could be replicated for other industries (plumbers, landscapers, etc.), using similar off-the-shelf components or as an agency ([14:25], [14:30]).

Productization and Market Fit

• Customization: The early product provided only basic dashboards for clients; lack of user-level customization demanded a more robust platform ([24:47]).
  • “Our front dashboard—the only thing you could see was audio recordings… you couldn’t tweak it whatsoever.” — Yevgeny [24:47]
• Self-Serve SaaS: The new Resora platform exposes agent parameters (voice, style, prompts, etc.), allowing users to launch agents and campaigns quickly ([27:45]).
• Profitability: Of the first $40K, $34K–$35K was profit, thanks to low running costs apart from Facebook ads ([32:49]).

The Founders' Dynamic

• Meeting Co-Founders: Yevgeny recruited Aiden via Y Combinator’s co-founder matching service. They “clicked” due to shared ambition and complementary skills ([20:27]).
  • “He’s very motivated, very ambitious and has that drive… without that, you’re not going to get anywhere.” — Yevgeny [21:35]
• Role Breakdown: Yevgeny leads everything product and technical; Aiden handles sales, outreach, marketing, and ops ([22:13]).
  • “My responsibilities really lie on the sales and marketing side...outreach and kind of brand creation, brand promotion, and handling everything on the operations side.” — Aiden [22:13]

Reflections on Technology and the Future

• AI Coding Tools’ Limits: Despite tools like Claude and Cursor making technical work faster, deep understanding of systems is still required ([30:19]).
  • “If you don’t know how to swing a hammer, the nails could go in bent… if you don’t know the framework, you wouldn’t even know what to look for.” — Yevgeny [30:24]
• Empowering Small Teams: Modern AI tools allow solo founders and small teams to achieve what used to require entire engineering departments ([31:49]).

Real Estate as an Early Adopter Market

• Willingness to Try New Tech: Some agents were excited to experiment, while others were deeply resistant—creating a mixed landscape ([34:13]).
• Brokerage-Led Training: Brokerages now offer AI classes, pushing adoption among agents ([35:10]).

---

Notable Quotes & Memorable Moments

• “$40,000 in 40 days.” — Aiden ([00:44])
• “I just got fed up with it. There was… no room for, ‘let’s try this new system…’ All the processes are the exact same thing as they were 30 years ago.” — Yevgeny ([07:00])
• “There’s no solution on the market right now where it allows a non-technical user to launch their AI agent and get it going in less than 10 minutes. There’s absolutely nothing.” — Yevgeny ([19:12])
• “If you don’t know how to swing a hammer, the nails could go in bent...” — Yevgeny ([30:24])
• “He’s very motivated, very ambitious and has that drive… without that, you’re not going to get anywhere.” — Yevgeny, on meeting Aiden ([21:35])
• “The worst thing you can possibly do, in my opinion, is give them an okay version and not a perfect version.” — Yevgeny ([11:40])
• “There’s literally no research out there on this topic. Zero.” — Yevgeny, on fine-tuning LLMs for sales ([18:38])

---

Important Segment Timestamps

• [00:44] – First 40 days revenue and business model
• [01:32] – Real estate cold calling realities
• [03:56] – Move from real estate to AI automation
• [05:10] – First success: AI sets listing appointment
• [07:51] – Launching customer acquisition via Facebook ads
• [10:45] – Manual sales closing and onboarding
• [13:23] – Off-the-shelf tools to build the initial workflow
• [15:36] – Evolution from agency to SaaS and need for custom LLMs
• [16:07] – Details on fine-tuning the AI sales model
• [19:12] – Decision to build scalable SaaS for non-technical users
• [20:27] – Co-founder search and meeting
• [22:13] – Division of responsibilities between Yevgeny and Aiden
• [24:47] – Early product’s lack of customization and customer demands
• [27:45] – Turning agency workflow into a flexible SaaS platform
• [30:19] – Reflecting on the role of AI coding tools and current technical challenges
• [32:49] – Startup profitability and ad spending
• [34:13] – Real estate agents’ openness (or resistance) to new technology
• [35:26] – Origin of the name “Resora”

---

Takeaways for Listeners

• AI-driven automation can unlock massive efficiency gains in any field reliant on high-volume outreach—if the tech feels ‘human’ enough.
• Agency models can be lucrative, but are difficult to scale without productization and self-serve customization.
• Technical founders must still possess deep expertise; coding AI tools aren’t “magic” in themselves.
• First-mover advantage is real, but so is the threat posed by the inevitable SaaS-ification of any successful manual process.
• The founding story of Resora demonstrates the power of combining domain frustration, technical skills, and a scrappy approach to validating (and then rapidly scaling) a new business solution.

---

Speaker Key:

• A: Andrew Warner (Host)
• B: Aiden Richards (Co-Founder, Resora)
• C: Yevgeny Matce (Co-Founder, Resora)

---

For founders, marketers, or technologists, this episode is a playbook for identifying automatable industry pain points, hacking together a first version, rapidly selling to eager early users, and then racing to evolve into scalable SaaS before the agency workload crushes the team. The lessons—and opportunities—are endlessly reusable.

Transcript

[A]

How cool is this? You're about to meet a broker who got sick of making cold calls. So he created an AI automation that made calls for him and it worked. So he started selling it as a service to other brokers and that took off. And so today he's creating software that will automate it for any broker who wants to sign up. This story is so good because it could be replicated in so many other industries, not just brokers. Listen up. Evgeny Matce and Aiden Richards are the co founders of Resora, which does outbound cold calls for real estate brokers. The next new thing presented by Zapier, the AI automation company you guys started selling. Give me the revenue from the first few days.

[B]

40,000 in 40 days.

[A]

What do people pay for at that time?

[C]

So we would charge anywhere from an initial thousand to $2,500 setup fee depending on the complexity and workflow, plus an ongoing 500 month monthly fee, plus 20 cents per conversational minute that the AI did as well.

[A]

How much of it was working in terms?

[C]

Everything was working in general, but it was, I want to say first the demand was too high. So the whole entire solution was not scalable long term. And at the end of the day it was just a GPT with a Texas speech model.

[A]

Let's go back. Understand how you got here. You, you've guinea were doing real estate. What'd you do in real estate that led you to find this problem?

[C]

Most of my day consisted when I was a real estate agent actually consisted of me waking up, getting to the office at around 8am, 8.30am, having my cup of coffee and going into a tiny room and start cold calling homeowners at 9am, have my lunch break at 12 or 1 and then continue cold calling until 4 or 5pm hoping and praying that I can can get a lead and get a listing appointment.

[A]

Wait, were you just calling strangers in a neighborhood and saying can I list.

[C]

Your property so you could do. That's one way to do it. But my specialty was actually expired listing leads. So what that is is homeowners who tried selling with a realtor before and their listing went expired, meaning that it didn't sell. So I would cold call those people because they have a higher chance of relisting.

[A]

Okay, that makes sense. I imagine that you must have gotten a big conversion rate, right? These are people who tried to sell their home. It didn't work. They might not be happy with their current real estate person. Boom, you come in.

[C]

Unfortunately it wasn't that easy. I would say they Have a typical conversion rate as any other cold call. Especially since they have. They had a very negative experience with the whole entire process. Not being able to list, I mean, not being able to sell their home with an agent before. And they're more reserved to that idea of relisting again. Even if a person is desperate, they just don't want to deal. They just. Since that their agent basically screwed them over, they want to deal with another agent again as well.

[A]

What was your conversion rate back then when you were making these calls?

[C]

I want. What do I want to say? I don't have a hard metric, but if I were to estimate 1 to 2%. 1%.

[A]

Okay, that sounds pretty good. No. Hundred calls, you're spending all day making calls. Not good. You were still frustrated by it. Why?

[C]

It's not like I would make a hundred calls and then boom, 1% conversion. First you have to actually get connected to a person that picks up on you. That's where you can call 200 people before you actually get connected to someone that is the correct homeowner.

[A]

Okay.

[C]

And it's not like, it's not like it was a hard set 1, 2%. It's not like I would call, let's say a thousand people and definitely would get 10 listing appointments. It's all very variable.

[A]

How did you decide to start with software?

[C]

I actually started real estate straight out of college and I graduated in computer science and cyber security. And I just wanted to dip my toes in real estate. Always just wanted to give it a try. But honestly, after two years of cold calling. Cold calling, I got very fed up with it. It was very draining.

[A]

Yeah.

[C]

And I couldn't take it anymore. And this is at the time when AI voice agents were becoming a thing. So I'm like, oh, wait, I can definitely implement this, this into my workflow. Because why wouldn't I, Why wouldn't an AI agent, AI agent be able to just call and qualify? So I used a couple of platforms. All of them suck. Because first of all, this is when AI voice agents were first becoming a thing. So all of them were not the best compared to what they are now. But second of all, there was no model that was designed specifically for sales. Conversations, phone conversations, hyper realistic or human like so to speak. And that's when I got the idea of, oh, wait, let me just do this for myself. Let me make it, let me at least prompt it out to be very realistic, a very rigid workflow. And when I did that, I actually got my first listing appointment, went to.

[A]

The first day Wait a minute. Your first AI voice agent got you a sale within a day?

[C]

Not a sale, a listing appointment.

[A]

A listing within a day. And that's the thing that you were getting only 1% response on? Right? That was the 1% conversion rate from the calls. Correct. Okay, all right, that's exciting. So then how far did you get with that?

[C]

So first of all, I wanted to optimize to make it sound as much as best as possible. So what I did is I kept this is when I was just getting comfortable with LLMs and large and machine learning all that stuff. And I realized with prompt that you can only take it so far. But I actually thought to myself, wait, I didn't get that far with it because I thought, so wait, what if other people want something like this? And that's when we launched Facebook ads. And that's when it took off. And I even didn't have time to do listings or run it for myself whatsoever.

[A]

Okay, wait, so you were immediately thinking, I've got to create this as software. I would think you ganny that you would think, why don't I become the real estate king of this whole neighborhood? I'm going to do nothing but automate this and I'm going to get myself all kinds of listings and then I'll bring people in under me who are going to run these listings for me and I'm going to make a ton in real estate. Why not think that way?

[C]

That's a very good question. And I'll be completely honest with you. I just got fed up with it. There was no one thing about my personality type. I love implementing new things, working on new things. And real estate is the complete opposite of that. It is the same repetitive thing. And it has been that way for over 20, 30 years.

[A]

I see you're a software guy, you're a problem solver, you're not a real estate, let's make money, let's make deals type of person. You just got into that because you wanted to make a little bit of money for yourself to get started.

[C]

Correct. Don't get me wrong, I was doing, I was doing great when it came to real estate. First two years I learned a lot. I learned, Tommy, a lot. Specific about business negotiations, all these things. I did excellent in input just in the two year mark. I just couldn't take it anymore because there was no room for, oh, let's try this new system, let's try. I saw that everything in real estate, and you could probably ask most realtors, they'll say the same thing. All the processes are the exact same thing as they were 30 years ago.

[A]

Okay. And I could see some people might enjoy that, that this is an area that they could keep improving. Of course, it's something that they could just keep building on because it's not going to change dramatically day to day the way software might. You, on the other hand, said, all right, I'm going into software. You started immediately thinking about how to turn this into software. How long before you got your first customer?

[C]

So I want to say I took at least a good month on it before I even decided to do Facebook ads or anything, just to. Because there's a plethora of agent types people need, so expired listing needs, buyers, etc. But I want to say as soon as we launched the Facebook ad campaign, it took us less than. What was it, four days? I think. Yeah, I think it was less than four days to get our first customer. And then it was. It was the easiest because I've never done something like this before. And it was the. I was so shocked of how easy it was take leads and convert them into paying customers because we did a pay per lead campaign.

[A]

And so that means that you were using their lead capture form. Right.

[C]

On.

[A]

On Meta.

[C]

Correct.

[A]

You work basically, you didn't need to have a landing page. You didn't need to have anything built out. All you needed to do is have an ad and a lead capture form that Meta put together for you. That's impressive. Who created the ad?

[C]

I mean, I did.

[A]

What was the ad, do you remember? Describe.

[C]

Was a video recording with the wave loop of the AI talking to a lead. And that's pretty much it. And with a call to action.

[A]

I see it's basically doing cold calling. There's not even a human being, there's not someone acting like a realtor. It's just a wave of a person getting a phone call from a, From a. Like not even a realtor. It was an AI.

[C]

Yes.

[A]

Damn.

[C]

That Couldn't tell.

[A]

But they could not tell because it sounded so good.

[C]

Yeah.

[A]

All right.

[C]

I mean, of course some people are going to be able to tell, some others won't. You're going to obviously realize it's a little. It's a little monotone, but for the most part that's why people clicked on the ad.

[A]

So then how do you go from them filling in their information on Meta's pages to giving you some cash?

[C]

Yeah, of course. Honestly, I didn't even look at it, at the way for them to give me cash. It would be more so okay, I will call them is because on the capture leave form they would basically also schedule an appointment. So I will call them, confirm the Google me we would have get on the Google Meet with them, explain to them what the system is, how it works and if it, if it is even for them that was the main thing because there were some realtors that were completely new in the game or have never done cold calling and they would want this and I would tell them no, this is not for you. Why? Because they wouldn't be. They wouldn't know how to handle those leads when it came time for the listing appointment. So they'd be wasting money. So it was more so of listen, this is what it can do. If you're already calling or you have someone calling on your on your behalf, this can replace it. That's all it is. It's AI. It's also not perfect. You can make mistakes. Of course that's all I have to offer and shockingly enough we had above 80% conversion rate and.

[A]

But it's you actually doing a Google Meet with them, walking them through this and closing the sale yourself. Correct. All right. Was the software ready to go to actually take care of them once they paid.

[C]

So that's the thing I would tell them listen, since we're having a huge backlog, it'll be anywhere from four to five weeks wait time.

[A]

Okay, and what were you planning on doing? And you can actually set them up within those four to five weeks or was this just a startup mechanism? You could because you were doing personally each one, you were custom coding each person.

[C]

Yes, exactly, I was. It's not even custom coding at that point. It was just custom prompting for each one and a specific workflows. But it was generally I needed the four or five weeks because we already we were having customers one after the other. I do a lot of testing, I can't if someone even give pays for our product. The worst thing you can possibly do in my opinion is give them an okay version and not a perfect version.

[A]

So.

[C]

So a lot of that time went to me testing it a lot because if someone trusts you with their money, it's your responsibility to actually deliver the best product.

[A]

Okay, so tell me what's the workflow? What did you use to create this for them?

[C]

At the time I was using VAPI to just prompt it out, have the LLM connected to the text to speech model and the transcriber. Then I also had a front front end connected to VAPI API where we just show the user the Recordings, appointments and that's pretty much it.

[A]

Let's, yeah, let's talk about what you, how you set this up, what was the workflow that you set up for them?

[C]

So I would use a low code automation tool such as Zapier and it was a very janky. The whole entire ecosystem overall, not the automation tool, but the whole entire ecosystem was not a full backend like you would expect, like a full stack web app. It was just something, have a workflow running for these customers and basically see if this is what people wanted. But it was not my intention to fully deploy and have this as the main, as the main code, so to speak.

[A]

I get it. You know what's interesting though, essentially what you're telling me is I could probably do something like this for people right now using let's say Zapier where I create an individual workflow for a customer that makes outbound calls for them. That is using nothing but Zapier and Vapi and what other tools would I be able to use to build this for people individually?

[C]

I think that you're going to be pretty much set to go with Vapi, Zapier and maybe a CRM connection because Zapier can connect to people's CRM directly. Okay.

[A]

So, so they use clothes or HubSpot or whatever. You pull their contacts out of there. There's a zap that then connects with Vapi. Vapi uses is the voice agent. Zapier would make the call out on behalf of the real estate broker and then say do you want to book this meeting? If they do, how does the meeting get booked? How does it get on the calendar of the real estate person?

[C]

So on Vapor you have a tool call or you can make your own tool call of course of connecting it to Cal.com or another calendar app.

[A]

Okay.

[C]

And that will automatically once you configure it. I can't, I won't go into the details because it is very mundane of how to fully set it up. But it directly books into the person's calendar schedules it.

[A]

Yep.

[C]

And they will get a Google notification. Hey, you have a meeting booked.

[A]

I see. So that's what Aidan was saying earlier. Aidan was saying earlier. Look, a lot of people are doing this as agencies. You're saying this is essentially what you did in the beginning before you created SaaS. That was self serve. But the big takeaway before we continue with your story is it seems like there's an opportunity for somebody to go and do this right now to say, you know what, I'm going To do this for plumbers. These guys Evgeny and Aidan are doing this. Realtors going to do it for plumbers. Everybody's got some kind of plumbing need that they need, but they don't have time to call. I'm going to do outbound calls on behalf of plumbers and I'll do it for them as a custom setup for every single company, maybe a thousand bucks a pop. And if it's not plumbers, maybe it's electrician. Probably plumbing. Actually, as I think about it, there's more plumbing needs than there is anything else. That's the kind of stuff that we're talking. Oh, I know. What else lawn and garden. Do you need? Some landscaping work done? I happen to be in the neighborhood. Got it. That's essentially what you're saying can be done. Aiden, that's what you said. Most people do this as agencies. You guys started as agencies, but at some point you gain. You said, I know, I need this to be SaaS. Am I understanding it right?

[C]

It's SaaS, but it's also development of an actual large language model that has a high enough EQ for phone and sales conversations that was the biggest thing, the agency model. Getting rid of the agency model was the icing on the cake. But the main thing is there's no LLM, there's no ChatGPT, there's no Claude that is specifically meant to respond for sales or human like conversations. Too verbose. How did you do that?

[A]

You personally would fine tune?

[C]

Not back then. So that's when, after six weeks of launch, that's when I stopped and I worked on the development of actually making the LM sound ultra realistic. And also condensing down from an agency model to something people can just launch right away, which makes it cheaper for the end consumer as well, but at the same time more scalable for us. But currently what we're doing is we take real conversations, we turn that into fine tuning data, to be specific, supervised fine tuning data. And we run supervised fine tuning on a large language model, which in end results in the language model actually sounding more human and more has more quote, unquote salesmanship in its responses.

[A]

I got it. Do you think that the agency model can ever, or at this point, with technology as it is today, use any of that fine tuning or you think that's a limitation of the automated workflows that we're talking about?

[C]

Yeah. So agencies, right now, they're not, and I don't mean this in a bad way, but they're not the most tech Savvy. These are just people doing automatic automate just through Zapier, connecting it to an API. Whatever, whatever.

[A]

If I were going to do this for plumbers, if I were going to do this for landscapers, would I be able to fine tune the model? Fine tune the way that the AI voice agent talks using recordings of other calls? Or do you think that's a limitation of that model? I could.

[C]

No, no, it's. Anyone can find it, Anybody can find it, but you have to know what you're doing. That's the thing. It's not like by fine tune, I don't mean prompt, I mean data orchestration of current audio recordings. Transcribe. So you take the audio recording, you transcribe it, then you run it through a judge LLM that grade every single conversation, make sure there's no artifacts. Then you also assign a scalar score to each conversation. So you rank on different parameters, those different parameters are added up, there's a scalar score, and then you go through each conversation and you choose which ones get ended up in the supervised fine tuning data set. And that's when you run your fine tuning cycles. But there's a lot of nuances to it. You have to know, like the learning factor of what to run it at the epochs, all these things. It comes through a lot of trial and error. You're learning these things. And interestingly enough, I was telling Aiden this the other day. There's literally no research out there on this topic. Zero.

[A]

I could understand that. And I'm getting what you're saying. Somehow we need to analyze real conversations and see what makes them magical, what makes them great, so that we can recreate them through these voice agents. And that's what you're working on. And you're saying it's hard even at your stage. It's possible even at the stage of somebody who's doing some kind of nation. All right, you then are getting sales. How do you transition from doing this like an agency owner who's customizing for each person to a SaaS builder. What's the next step?

[C]

Well, the transition wasn't that hard for me because that's. I knew going in that's what I wanted to do. Okay, this was just meant to confirm the market for me. Okay. This is meant to see. Should I even allocate all my time and energy into this? And once we did our six week run, I got that confirmation. I told my customers at the time, listen, we're pausing during the time and I just got to work on Coding a whole entire front and back end, testing different flows and doing a lot of research of how large language models work because all this is very neutral. There's still some studies on machine learning algorithms, but it just took up a lot of research and it was a lot of brainstorming, of thinking, okay, there's no solution on the market right now where it allows a non technical user to launch their AI agent and get it going in less than 10 minutes. There's absolutely nothing. So it was just a lot of trial and error of testing and how to do it. And that's all I basically did. So I went from an agency owner agency, you know, manicuring AI agents to right away I was excited to start thinking, okay, how can I streamline this whole process? It got me already ramped up and going, okay, the idea of it.

[A]

Tell me how you and Aiden connected then.

[C]

Absolutely. So at the time I was coding, I still am over 12 hours a day. And I was thinking to myself, oh, how am I gonna even be able to onboard customers? How am I be able to sort out business nuances, just handle everything else because just alone coding takes up forever. So I thought to him, maybe I do, I do need a co founder. I went on the Y Combinator co founder matching platform. I just want to see like, you know, maybe there was like minded individuals out there. I just put my profile profile out there. I got message from a few people, did interviews with them, and I just think to myself, yeah, definitely not. Then Aiden reached out to me, we did an interview, we instantly hit it off. We talked for our first zoom meeting or Google Meet meeting. We talked for hours. Not hours, but more than expected. We met in person and we ended up talking for hours and it was an instant click.

[A]

Why? Why did you click with him?

[B]

What?

[C]

His hair. His hair was just so amazing. He. I clicked on him because he was very. Not very, I don't want to say very direct, but he knew what he was talking about. He is very motivated, very ambitious and has that drive. That's one of the most important thing things in a person is drive and ambition and motivation. Without that, I don't care how smart you are, what you know how to do, you're not going to get any worse. And Aiden had 200 of all of that.

[A]

What was the role breakdown going to be? Aiden, let's bring you in here.

[B]

Yevgeny is pretty much responsible for everything on the product and engineering side, as you've just heard. And then of course, I kind of, you know, give my feedback and we, we collaborate on where we want the product to head and what we think is next. But obviously he has a better understanding of, you know, what's possible and what should be done and what comes next. So my responsibilities really lie on the sales and marketing side. So I'm doing all of the outreach and kind of brand creation, brand promotion, and then handling everything on the operations side, things around like, you know, legal and setting up new software and making sure that we're kind of maintaining relationships with people that we need both on the client side and just regular colleagues like that.

[A]

And people like me. I mean, you basically messaged me.

[B]

Exactly.

[A]

Endlessly. And when I said I think it's a good fit, you kept sticking on me until we found a way to make it work. That's, that's.

[B]

Yeah, exactly.

[C]

That's why I love them.

[A]

I do respect that. I do like that a lot. What about the ads? Are you taking that on now?

[B]

Aiden, we're not running any ads right now. We're launching Access to our wait list tomorrow and that's going to be a little bit of a smaller campaign just in terms of kind of activating everybody that we've already spoken to and then kind of getting the ball rolling on our initial go to market strategy. We do want to run some ads in the next few weeks, but we're still kind of trying to figure out exactly where they're going to fit in terms of resource allocation and time, such as SEO.

[C]

Everything else piled in on top of the ads.

[A]

The original people who are paying you a thousand dollars for the service, are they still paying you? Is that part of the revenue that you're bringing in?

[C]

No, no. So at all? Oh yeah, I stopped at all because it was a lot of work just to, I'll put it like this. Let's say I have a AI voice agent for you and you needed something tweaked. That means you come to me, we have to have a meeting, and then I have to go over everything exactly which you need specifically fix or change all these things. Now imagine with 40 plus people, you don't have time to do anything else.

[A]

I see you. So you didn't even set them up. How many people did you actually set up?

[C]

No, no, I set them all up.

[A]

If you're setting up, I said, once you set up, isn't it ready to just go on its own?

[C]

Not necessarily, because like I said, people, a lot of people need tweaks, a lot of people need changes. And there was no customization to it. Our front, our Front dashboard. The only thing you could see was audio recordings and. Yeah, just audio recordings. And you couldn't tweak it whatsoever. There was no customizability at all for the end user.

[A]

What type of tweaks were people asking for?

[C]

I wanted to say this instead of that. Can, can it go through. Can it go, go through tasks 1, 2, 3, but skip 5 instead of things like this. It was just everyone wanted such a different style and people, once they hear it, everyone, even if it got that point, people want improvement. Oh, but can I do this? Can I do that? Can you do this? And I just. And that's when I talk to myself, oh, okay. I need a platform where people can do this, everything themselves.

[A]

But they can.

[C]

They don't have to have the technical knowledge to do. My mind frame at the time was why isn't there anything like this right now? That's what got me the most excited I see.

[A]

And if you were running an agency, you'd be able to keep servicing them. We're not talking about big complicated tweaks, it's just little tweaks. But it does take time to interact with them. It does take time to go and do the work yourself.

[C]

No, it's not even, it's not even little tweaks. Because some people would want to a whole redo of. Oh, okay, I don't want to. It's not even a whole redo. Let's say, for example, I want a mortgage guy wants an agent. That means I have to go, I have to research the scripts. I have to tell him to send me his scripts. But the thing is, I would never, even when I was doing the agency model, I would never script my AI agents. I would never say step one, say parentheses, 1, 2, 3, 4, 5, 6. I would never script that. I would actually instruct it step by step. So that's what takes up a lot of time. And then back testing that or test to get nonsense until it sounds good.

[A]

I got it. All right. I do still think though, this would work as an agency and someone who did it as an agency. No, it would, right. If you're, if you're looking at $40,000 within 40 days, my hunch is that you'd probably be able to do 50,000amonth in agency business with this. Right. Recurring revenue. Just keep taking your customers, deal with churn by bringing in some new ones. It would just mean constantly tweaking people's automations.

[C]

Yes, absolutely. And that's how a lot of companies do. And they're very successful with Us why we could, we weren't able to is because usually these AI agencies, they have a team or it's two or three people and they're constantly tweaking it. Their customers, they tweak it, whatever, whatever. So it's ongoing work, but it's go it think about almost like a marketing agency, right?

[A]

Yeah.

[C]

A marketing agency is running ads. They're going to be constantly tweaking ads for the person. Like the person say no, I don't like this or the type of customer I'm bringing is not good. So they're going to be tweaking as the exact same.

[A]

And the challenge also if, if you were running an agency is somebody is going to sassify this. It's just a, it's just a natural step. And so an agency doing this would have to keep running fast because there would be some, someone in the background who's coming in. Tell me about how you turn this into, into software. Oh.

[C]

Sure. So I broke it down to the bear. I didn't even break it down to the bare bone. I want to say I thought to myself of how I can let users deploy their own AI voice agent that is good that they don't have to prompt out because prompting is a full time chain in itself. So that's the first thing I thought and the first thought that came to my mind is of course you need a dashboard. So what I did, I started coding and just testing a lot and one of the first things we did was build out individual AI agents that we already know people wanted. So specifically an expired listing voice agent, a whole real estate wholesale agent, all these things. But these AI voice agents are connected to our fine tune LLM which is our biggest MO because anyone can go ahead and prompt out a few agents and whatnot and satisfied as well. So what we did, we fine tuned LLM, we built out a bunch of agents that we already know people want and we put those agents on the platform, we then expose those agents, those agent parameters of, you know, the prompt, the voice styles, the type of voice, all these things on the platform itself. So the user has the option to choose, oh, I don't like this voice. Okay, cool. You can change the voice, change the speed of the voice, all of these things and also at the same time input parameters into the AI agent. For example, you can name your AI voice agent, you can give a background about yourself and mind you, not through prompting, but just through text input that we expose on our website. You can hit deploy, then you can launch your own campaign. And it'll start calling for you. So everything is done and this is all connected via backend, of course. So that's how we turn from agency model to a SaaS model. Basically of thinking what's the best way to let it be customizable, but at the same time sound exceptionally good. And that's where the fine tuning also comes in.

[A]

Talk to me about that, about the work involved in it. I think when I talked to you it seemed like, well, there's cloud code, there's cursor, everything's so easy. And you said, yeah, it's easier, but I'm still spending 12 hours a day working. I've never worked so hard in my life. What's making it easy? And then where's the difficulty?

[C]

Okay, so this is like if we're just talking specifically about full stack development, right?

[A]

Sure.

[C]

The thing that people have to understand is that with cloud code or any of these coding tools, they're just that they're tools. If you don't know how to swing a hammer, the nails could go in a bent. If you know how to swing a hammer, the nails to go in straight. Right. And it's the same exact thing. Why I say that? Sure, you can tell, you can vibe code with clock or you could tell, oh please implement this and build you some things up. The thing is, if you don't understand what you're coding, if you don't know the framework, if you don't know the architecture or the workflows behind it, or they need to be in place, you're going to have a lot of errors, you're going to have a lot of bugs and you're going to have a lot of issues with it. So cloud code is exceptionally excellent if you tell it specifically out of what you need implemented, how you need it implemented, and you have a whole plan in place, cloud code is exceptional for that. But if a person doesn't know the framework, doesn't know how a connection from A to B happens, or how it should look like, then they wouldn't even know what to look for. When Claude, if cloud generate good code or bad code as well, talk to.

[A]

Me about what's possible. Today I remember talking to Gary Tan, president of Y Combinator, and he basically said, look, smaller teams are able to create so much more than they've ever done before. It feels like that's where you are right now. You're a one person operation creating the software.

[C]

Yeah, no, absolutely. You can definitely create a. I can't imagine doing this all manually. I really can't. So think about it. Think about like this. Either you can write the essay yourself. Yeah. Or you can get GPT to write that essay for you, and then you make tweaks to it to make sure it sounds good. Right?

[A]

Yeah.

[C]

So I can't imagine without them that I would be.

[A]

To do.

[C]

Be able to do it at the scale that I have done it at. Yes, absolutely. People can do a lot more than they kind of even a year ago.

[A]

I see. So it's just if you know what to ask for, you can get it, and it's faster than doing it yourself. But you still need to know what to ask for, is what you're saying.

[C]

Yes. You have to know how the architecture should be of that said framework or code, whatever you're working on.

[A]

All right, so right now, how much of the $40,000 did you guys get to keep? You did 40,000 in sales. How much of it did you get?

[C]

I want to say we kept straight profit. 34. Around $34,000.

[A]

Okay.

[C]

$35,000.

[A]

What kept you from keeping the rest and what allowed you to keep that part?

[C]

Facebook ads. That was basically our only.

[A]

Oh, that's it. So you did get paid. You built the automations. You just couldn't take on more automations for more customers. Yeah, I see.

[C]

And I just. I. And I was itching. I was itching to finally. Okay, let me start. Fine. Too. Let me start this whole entire. I was literally itching to already start that.

[A]

And what about the people who paid you and needed tweaks? Were you still tweaking it or did you finish? You say, look, I'm not.

[C]

No.

[A]

Anymore.

[C]

I was still tweaking it for them, and I still took care of them until I told them, hey, listen, I'm gonna build this out, rebuild, and make it a lot better. You're obviously gonna.

[A]

You're.

[C]

You're going to obviously get compensated. You're going to get a free. Free few months. All of these things when we launch.

[A]

Got it.

[C]

And it's going to sound a lot better for you as well.

[A]

It does seem like real estate brokers are willing to try new technology more than others. I don't know about plumbers, but I do know that when there's a new tablet, I would see it on a real estate broker's desk. There's. When there's some new app, they're likely to play with it and even talk to you about it. I don't know what it is. I don't know if it's. They're all entrepreneurs, I don't know if they have a lot of time on their hands. What do you think it is?

[C]

You know what? I'll be honest.

[A]

You.

[C]

It's a. It's a very. It's a 50, 50 mix. Some real estate brokers, agents, especially the older ones, they're very against anything new.

[A]

Okay.

[C]

It's. It's like if you bring it up, you personally offend them, essentially. While others are very excited to try and currently, what's happening brokerages across the United States, real estate brokerages, and this was prominent in my brokerage too, was now they have classes on AI. They have classes of how to utilize business. So. So I guess that's also what helped us out a lot too, was since a lot of these realtors were their own brokers were telling them, listen, you have to utilize AI. You have to utilize AI. AI's next books, investing. They're being constantly set this. I guess that also helped with onboarding. The first. First users that we ever onboarded.

[A]

I see. I do get that too. They have always been really good at training. They've always been good at doing motivational speakers. And so now they're teaching people how to do AI. All right, why'd you come up with the name. Let's close it out with that. Why'd you come up. How'd you come up with a name?

[C]

RE is of course, the real estate. Zara is in a lot of Slavic languages. New beginnings. And that's when I was finishing real estate and going into this new field, essentially.

[A]

I see. All right, fantastic. Thank you, gentlemen. Good luck. Bye.

[C]

Thank you so much, Andrew. Thank you for having us.