← The SaaS Podcast - Growing Profitable AI SaaS & AI Agents
The SaaS Podcast - Growing Profitable AI SaaS & AI Agents
463: Escaping the "Consulting Trap": How to Pivot to $1M ARR - with Ibby Syed
00:57:21
Ibby Syed is the co-founder of Cotera, an AI agent builder that helps businesses automate tasks using the tools and data they already have.
Loading summary
Transcript75 lines
- [00:00]
A
Foreign. Welcome to another episode of the SaaS podcast. I'm your host Omar Khan and this is a show where I interview proven founders and industry experts who share their stories, strategies and insights to help you build, launch and grow your SaaS business. In this episode, I talk to Ibi Sayed, the co founder of Katera, an AI agent builder that helps businesses automate tasks using the tools and data they already have. In 2022, Ibbi joined his co founder Tom, who had just gotten into YC with an idea for a customer analytics platform. They did what YC founders do. They talk to customers, found someone willing to pay $200 a month and built a scrappy first version in a few days. Over the next 18 months, they grew the product to over 150,000 in ARR. But something wasn't right. Customers weren't logging into the product. They'd call with a question, get an answer, and disappear until the next question. The founders realized they'd accidentally built a consulting business, not a software company. Then came the turning point. A customer asked them to extract topics from support tickets. EBI built a data science solution, but it was slow and clunky. Tom decided to try the newly released OpenAI API instead, and with just 100 lines of code, he was able to build a better solution. That was the moment everything clicked. The founders made a hard decision. They stopped doing services, fired some customers, and pivoted to building a platform where people could set up AI agents by just writing a prompt. The difference was immediate. Deals became easier to close. Instead of building custom solutions for customers, they taught them how to build their own. Today, Katera has around 15 enterprise customers, a team of 10 people, and generates over a million dollars in ARR. In this episode, you learn how EB used LinkedIn outbound with an 8 to 10% response rate to book over 25 customer interviews per week. Why? Reaching 150k in ARR with their first product made it hard to pivot, not easier. What triggered their pivot when 100 lines of code outperformed a bloated data science solution. We talk about how Katera's outbound strategy evolved from generic pitches to sending prospects actual leads generated by their product, and why IBBI believes most vertical AI startups will grow fast and then die, and how building tools instead of solutions creates a stronger business. So I hope you enjoyed. I talked to a lot of founders stuck in the same spot. They've got a clear vision. They just need the right team to build or scale it. That's where Gearhart comes In they're an AI powered product development studio that handles the entire technical side of building your B2B SaaS platform or AI agent. Built by serial entrepreneurs, they understand the unique challenges of startups and can plug into your team to accelerate growth. They've built over 70 successful products, including SmartSuite, which raised $38 million and and is used by companies like Capital One. Right now they're offering our listeners the first 20 hours of development for free. Just book a call@Gearhart IO. That's Gearheart IO. Your messaging infrastructure shouldn't be your bottleneck. Signalhouse is the next generation communication platform built for modern SaaS companies. Easy to use APIs that connect in minutes, instant A2P approvals, better pricing than legacy providers, and top tier support included. Whether you're sending 10 messages or 10 million, you get speed, transparency and reliability without the complexity. Visit SignalHouse IO. That's SignalHouse IO. Your company's credentials could be on the dark web right now and you wouldn't even know until it's too late. Nordstellar detects threats before they escalate. It monitors the dark web for leaked credentials, tracks your attack surface and protects your brand from impersonation. Built by the team behind NordVPN. Don't wait until your data is for sale on the dark web. Book a demo and mention Black Friday 20 for 20% off at Sasclub IO Nord. That's Sasclub IO Nord Ib welcome to the show.
- [04:38]
B
Thank you so much for having me.
- [04:39]
A
Omer. My pleasure. Do you have a favorite quote? Something that inspires or motivates you?
- [04:44]
B
I do. The quote was said by Willie Guy. He was Mozzie on White collar, like a TV show on USA from like 15 years ago. And the quote, he's got a lot of good ones, but the one that I like the most is if you want a happy ending, it depends on where you stop the story.
- [05:06]
A
Very true.
- [05:07]
B
And that is, I think, how I live my life, where it's always just looking. I don't know, whenever I feel. I mean, I don't know if you feel this way. Startups are very much like a sine wave of emotions. You just kind of hope that the next trough is higher than the previous trough. And that's the kind of thing that I live my life by.
- [05:29]
A
I think that's a good attitude. So tell us about Katero. What does the product do, who is it for, and what's the main problem you're hoping to solve?
- [05:37]
B
We are an AI agent builder. The claim that we make at least is that we are the easiest way to bar non set up AI agents on top of your native enterprise infrastructure. What that basically means is we are a product where you come in, you write a prompt on what you want the AI agents to do. You give it access to tools just like you would a human. So you write, you know the prompt just like you would write an email to maybe a new hire. You give it access to logins so the systems that it needs to use to be able to do its job and then you set it live on top of something. So a great example is, I don't know, maybe you want for every call that finishes to go into every call transcript and generate a case study deck in Google Slides. Right. You basically explain to the AI, hey, I want you to go into this table of call recordings every five minutes that's in Snowflake. I want you to like make a doc, I want you to open up Google Slides. I would like you to go to this template. I'd like you to make a case study. Here are examples of previous case studies. Please write me one and send me a slack message.
- [06:44]
A
And you've.
- [06:44]
B
All you have to do is give it access to where the data sits, write out a natural language instruction guide, give it access to tools and tell it how often you want it to check. And that is our claim to fame.
- [06:58]
A
Great. And who's your icp?
- [07:00]
B
We primarily sell to series B, series C and series D businesses that are fast growing and will sell to any team as long as they have sort of like a desire to learn how to use AI products. We generally go after sort of these like growth hacker types.
- [07:18]
A
Okay. And give us a sense of the size of the business. Where are you in terms of revenue, customers, size of team?
- [07:25]
B
Yeah, we've recently switched actually to doing a PLG model that we're testing out. But we have around 15 enterprise customers. Team size is just around 10 and we're just around the 7 figure ARR mark, low end of $1 million.
- [07:43]
A
So tell us about. So the business you founded, the business with, your co founder is Tom.
- [07:49]
B
Yep, Tom Firth.
- [07:50]
A
And you guys got into YC with your original idea and the original product was very different to what you're doing today, right?
- [07:59]
B
Yeah. So the original product, it was actually Tom was the one to get into yc. I joined him, we met around that time. So the original product was. Yeah, customer analytics product. It was something where you know, you, you sign up, you connect it to your data warehouse and it would help you effectively like compare different cohorts together and help you basically create Personas that you could use in your marketing tools to help prevent churn. That was our. That was the original idea. It very quickly morphed from that. But. But yeah, that was. That was. That was the original idea.
- [08:36]
A
Great. So basically you. You started that business 2022, worked on this customer analytics platform for a couple of years, and you got some customers, but not really great traction. And then at some point you pivoted and in a relatively short space of time, we're able to acquire more customers, hit seven figures, ARR, and so on. So I want to talk a little bit about the first half of that story. First, tell me about when you met Tom. What stage was the product at? And how did you guys go about getting those first, let's say first five or ten customers?
- [09:22]
B
The product was basically even right after YC was when we actually started going after this. The way that we realized that we wanted to go after a specific product was we just interviewed customers. We would just go and talk to tons and tons and tons of people. I think YC gives you this advice, but this is kind of common startup advice is if you have an idea, just don't build the hackiest solution you can. Right? Like whatever you can do in a day, a few days. Back then, this was before cloud code and the advent of tools like that. So we basically would go to. We would drum up ideas. We would call to outbound people on LinkedIn. We'd say, hey, we have an idea. Can we please talk to you about it? And we tried a bunch of different things. We finally got one person to agree to pay us. It was like, I think a few hundred dollars a month. And then we were like, okay, cool. So we have somebody that has actually committed to paying us. They need some help figuring out how to keep their customers around and develop Personas and stuff. Let's actually build this out. And we spent. My co founder, actually he's the technical genius, spent a few days building a handful of things out on top of a jupyter notebook that I had. And the issues with that product were very much that we were doing too much consulting. Right. Like every business is set up a little bit differently. You really can't give a company sort of like perfect reasons to churn. It is also just like the by factor of the. Just by fact. Just by the fact that it is sort of like an analytics tool. They'll ask a question, you answer it, and they come back to you with another question. There isn't really Like a product there, right? Like you can give them a dashboard, but once they see the dashboard and they see, oh, okay, like 25% of my customers churn because they buy X product, that's kind of like most of the job that you've done. And the product doesn't really have a ton of stickiness like dashboards for anyone who's like, worked in data analytics. And BI doesn't have a ton of stickiness. So we ran into a ton of problems. One, it was very consulting heavy. The product didn't really do very much. And then it was just as all data analytics products are not very sticky, the sort of ongoing use for it wasn't super high and we couldn't quite figure out the pricing, so we moved away from that. We did that sort of consulting stuff for about 18 months. It didn't really work. And then we just by happenstance, met Rebecca Blunt at Coterie and she wanted something a little bit different and that kind of led into what we do now.
- [12:04]
A
Tell me about the outbound on LinkedIn and getting people to talk to you. How successful was that? What kind of response rates were you getting? A lot of founders I talked to, they're just like, no one's replying. How do I go? And, you know, I know I should go and talk to people, but no one wants to talk to me. So was there something you learned from that experience about what to do or what not to do?
- [12:26]
B
Yeah. So, okay. I actually have a lot of opinions on this. I have a ton of opinions on this. And I've actually seen the world kind of change here back when I started. So back when we started and this was only like really two and a half years ago. It's not like a super, er, I guess three years ago at this point. It wasn't super long ago, but LinkedIn was not sort of like taken over by bots. Email was not sort of like you could definitely do mail merge emails and like do that, but it wasn't as crazy of a. Of a channel cold. Outbound on phone was also not that crazy because, like, not that many people were doing outbound calls, at least compared to the number of pings I get on my phone now. So I will give that asterisk. I think the bar has been raised. I still think that you can outbound people on LinkedIn. And what I'll do is I'll talk about what worked then and what works for us now because we still use this as a channel. What worked then was just like appealing to their sense of wanting to be a part of something bigger, right? We didn't outbound them and say, hey, here's what we do, here's what we sell. Because at that point, we didn't really actually have anything that we could sell them. We basically just were like, hey, we are two guys. We founded a company. We dropped the YC name. We dropped the name of some of the other people that we'd worked with. I was the first data scientist at Peloton, so I put that in there because Peloton was very popular back then. And so I just say, hey, here's a short blurb, two sentences max. And then we're like, hey, we need some help. We want to interview people from time to time. If they were important enough, we would offer them coffee. Like, hey, we'll send you a Starbucks gift card if you want to chat with us. Almost nobody took us up on that offer, but it did increase our open rates quite a bit. I will say I don't remember what our exact response rates were, but they were in sort of like the high single digits. So you can basically estimate like, close to 10%. So for every like 10 we sent out, we would get like one return. And when you think about it, like, if you've got 150 LinkedIn requests that you can make across both founders and you're getting about an 8 to 10 response rate, that's basically 25 meetings booked a week. So you can do about five meetings a day. And that is like a lot of sort of data gathering that you can do at any time now that does not work anymore. I can't, like, I am sure that I can tell people that we have a fun idea, but we actually do sell a product. Now. AI agents are actually like a very, very difficult thing to outbound people with because almost everyone is trying to sell this today. The thing that we have found works is building an AI agent. Because the thing is, the beauty of it is you can actually build products very, very quickly. And our product is super valuable. So we let the product do the work for us. I'll give you an example. If I am trying to outbound a person in sales and I'm trying to sell them a sales agent, one of the things that I can sell them is something that, you know, monitors Reddit for sales opportunities, and I can email them and say, hey, I have built a bot. I have a product. One of the things that you can build on it is this thing that monitors Reddit for sales opportunities and people talking about your product. Or talking about your pain point. And I can tell them that I've built that or I can just send them leads, right? I'm like, hey, like, I built a product. Here is like, here are like five Reddit threads from the last week where people are talking about something that you can help with. Are you interested in, like, taking a call? And that is that people go crazy for. Like, people love that kind of stuff. If you actually, like, show value, like, it all comes down to showing value. Right before, like, when the barrier was a little bit lower, you could show value by, by dropping a name or by, like, saying YC or by explaining something. But now, like, the market is super, super saturated and people are getting inbound all the time. You have to show them something that's truly different. And the best way to do that, in my opinion, is to actually send them value. That is the only way I think that people are going to buy products and by having the product do the work for you. It's very, very simple. So that's what we do. We use our product for our own outbound, and we also use it to test out a bunch of different ideas on what AI agents people are going to like. Because people buy AI, We've started to see that people are not generalist buying it. They start off with like, hey, I'm trying to solve this problem. AI would be a solution for this. And then they just start to ask more and more questions and build more and more stuff.
- [16:58]
A
Unfortunately, most people are too lazy to put that kind of work into doing outbound.
- [17:04]
B
Yeah, I get a lot of really bad. I get a ton of really, really bad, bad messages, right? I think this, the most interesting one here is recruiters. Like, they'll see that we have like a job posting on, on LinkedIn or on the YC on the YC page. And they'll be like, hey, like, I, I send you all here, like, talk to me and pay me like, 20% and I'll send you, like, leads. You can give me like three or four profiles without the name that you think fit. Because if you've done the job, you can actually have AI do that job, right? You can look at my, you can have an AI pull my job description that I have on LinkedIn. You can have it look at all the resumes of all the people that you have. You can remove all of their pii, and then you can send me an email saying, hey, here are three people that I think would be good fits for this job. You've done my job for me. At that point, then I have to basically just click to unlock, right? Nobody does that. And I'm so confused. I'm like, why are you telling me that you can help me hire? Just help me hire. I'll pay you if you find me someone that I haven't been able to find before. If your network is truly as good as you think it is, then have the network do the talking for you. Sometimes people just have a bad product. There are also a lot of scam artists out there. Omer, I'm curious. I've been talking, I've been monologuing and I hate monologuing. I'm curious to hear what you think about this. I think that there's actually just a lot of people that are running scams or they're trying to take shortcuts and they don't actually want to build something great. I don't know. I don't love taking a short termist approach, but I think a lot of people do that too.
- [18:39]
A
Yeah, well, I think I see that a lot with. There are people who are really just don't want to put any. Using AI is now a way to do less work and put in less effort and less thought into reaching a lot more people at scale. And that's never really going to work, right? It's like, hey, I was sending out, doing outbound and reaching out to 100 people a week or a thousand people a week and was getting a crappy response rate. And now I'm going to do the same thing with AI and reach 100,000 people a week and everything is going to start working magically. Right? And I get the same thing with pitches for the podcast where it's this predictable format. Hey, I just listened to Episode whatever. It made me think about what and it made me think about this person. And first of all, it's like that format immediately tells me they didn't actually do any of that. But okay, it's like, all right, so some relevance. Half the time they're recommending somebody who wouldn't even be. It wouldn't make any sense to be on this particular podcast. That's where you lose me. Right? It's like you could have just said, never heard of your podcast before, don't listen to it. But I came across it and I know somebody who would be a good guess because here's why they're relevant. Right? Just based on the type of people you're interviewing, that would be a lot more valuable.
- [20:17]
B
Yeah, I mean, it sounds like same thing, right? We have a podcast and yours is just way more successful than ours is. And so it sounds like you have your pick of the litter, right? You get pitches and you take people that come on. But if you were just starting today, you could probably outbound people. And yeah, you have two approaches, right? You could outbound either every single person on Apollo's database that seems to have started a company in three years, or you can put, you can actually use AI to put more thought into it, right? You can say, okay, look at all of these profiles one by one. Tell me if they've raised a seed round. At least tell me if their website still has like a real, like, because half of the websites are dead, right? Like most, like 90% of startups fail. Like, just make sure that their website is still alive and, and maybe they, and check their LinkedIn to see if they've posted anything. And that will cut your list down by probably 80%. And then of the rest of the 20%, you can lead score, you can have the AI do lead scoring with you. And then what you can do is you can actually reach out to the, the top 1% super manually and put some care and thought into it. And you not only have successful guests that are going to, you know, be a good group to do a show, be a good person to do the show with, but you'll also then, you know, have cultivated real value for your listeners, right? You start off and then the second part is they'll share it with their network because you've done this scoring and you've sort of like checked to see there's just multiple ways of doing things. And I just think that people are lazy and do things very, very thoughtlessly.
- [21:53]
A
I mean, when I started over 10 years ago, didn't have even one episode out there. And what I decided was I was just going to who are the cool people I wanted to go and interview? And I personally wrote every email explaining why I thought they would be good guest. And even now I look back and I mean, out of my first 10 episodes, like Wade from Zapier, Melanie from Canva, I mean, obviously they weren't as big back then, but I'm sure they had better things to do than come on a podcast that had zero episodes, right? But it's about making an effort, it's about trying to build relationships. But anyway, let's go back to Kotera. So you're on this path of building this customer analytics platform. You start to get some sales. I think eventually you guys got to like over 100k or so. ARR.
- [22:52]
B
Yeah, yeah, we got like to just around 150, 170k of ARR, which is like, you know, it's exciting when you haven't ever done anything before when you're like an engineer somewhere. Obviously, like it's pithy compared to real businesses, but it is really exciting to get that first bit and you can blow yourself into that false sense security where you're like, oh, like this is working now, now. And that is honestly kind of what I did. My co founder is actually very good at, much better at this. Where I was so excited about some early success that I didn't see, I neglected to see, hey, this is not really a software business. It seems to be like a services business. People aren't actually logging into the product. The value that they get is from talking to me and my team. And the thing that we had started to become was an agency and at some point we actually had to like, we actually pared back our business a little bit where we like fired some customers. I can, I'm putting the cart before the horse here. The real thing that happened was we actually had somebody that came in and they asked us to build something hyper specific and you know, eager to please an early stage founder. I tried doing it. It didn't really work. And that was right around the time that OpenAI's API had come out. And my co founder messaged me and goes, hey, what this customer wanted by the way, was something to extract topics from customer support tickets. And my co founder was like, hey, this would actually be a really good fit for this random OpenAI ChatGPT API that's come out. How about I try doing it? And I had written a data science script that was multiple gigabytes of infrastructure and it was really slow and it sucked. And my co founder wrote like maybe 100 lines of API code and was able to solve the problem. And that was when it clicked was we were like, oh man, this is a massive opportunity. We are probably the first people to try doing this very specific thing. Let's go after this market. And we closed a handful, we started to close people that way. And then our current product is basically just like an edit of that initial product.
- [25:05]
A
I mean the problem is in those early days it would almost be easier if you had no sales because then you know, you have to do something different. But once you have some customers, some money, you're like, ah, it's kind of working. We should keep doing more of this.
- [25:17]
B
It's so, so, so difficult because everyone, when you start a business this is the thing that they don't tell you is everyone either says that you're going to die, like, you will, like, blow all your money and die, or you're going to be like, a standout success. But, like, most businesses kind of, like, waver in this. Like, well, not most businesses. Most of the ones that don't die, at least are like, they're very, very wavering. And that is like, a very, very tough place to be in. Right? Because, like, you have customers, you have revenue, you have employees. It is incredibly difficult at that point to either, like, pare back and fire customers and go back to your investors and say, hey, by the way, we're not doing this anymore. Even though you're succeeding at closing deals. Yeah. But what you've done is you've basically found a local maxima and you want to find a global maxima.
- [26:11]
A
So you guys have this wake up moment with the OpenAI API. Tell me about the thought process that went into the pivot. Was it literally like, oh, my God, this is what we should be doing tomorrow? Let's kind of become this AI agent product? Or did it take a little longer than that?
- [26:33]
B
No, it took a while. It took about a year and change. We started basically, like, what was happening is we then became like an AI company, right? It was like this new wave of, like, chatgpt come out. People were, like, starting to look at AI. They were starting to be like, what. What can this thing do? And what would happen is, like, we would get into a business and then that team would then be like, oh, like, we're working with this AI company. And then they would introduce us to other people on the team that would then say, other people at the company that worked in different departments, and they would be like, hey, we hear you're the AI guys. Can I do this with your AI? And it's always like, can I do this with your AI? Where, like, you know, we would enter through, like, a customer support team or like a sales team or something, and they would introduce us to marketing and, like, the thing that we were doing for customer support was like, extracting text. And they would be like, hey, can you, like, generate case studies from, like, calls or. Or can you, like, monitor the Internet? And those are actually, like, quite different use cases. But we were the quote, AI guys. So we obviously, you know, you don't. You don't you want to. You explore some ideas you say yes to, some ideas you don't. You say no to. But enough people had started doing this with us where we were like, oh, like what they want is fundamentally workflows, right? Like we were making an LLM call on top of infrastructure. We need to give the access to tools. We need to basically, like, we actually don't like MCP servers at Kotera because they're incredibly horribly built. And so we had to build some of these integrations to start. It was back into like the customer ticketing tools and the customer support tools to dump the data back in. But then we started just giving them the ability to do action. And the first time, it's like a drug, man. Like, the first time you hook up your LLM to like your Gmail account that gets like 50 spam messages an hour and you basically just tell the AI every five minutes. I want you to look at the emails I got in the last five minutes and send me a slack message with the ones that like, don't suck. Fixer has basically turned this into a $50 million product or $50 million of ARR in the span of like two months or something like that. The thing that, the thing that they've done is they've taken an incredibly simple primitive that is a very easy product to build and they've just made it into. Because it's a problem that real people have. But the second you find that your AI can do this, like the LLM system infrastructure that we've built up could do this, we were like, oh my God, this is crazy. And that was when we basically we changed a lot of things. We stopped being like a sales led organization. We decided we wanted to be product led. We wanted to do, we wanted to build the best product in the world. We didn't want to do services anymore. That was when we were like, oh, this is actually like a real thing that we should build because we have this advantage for now. And that's what we're currently in the middle of doing.
- [29:23]
A
Okay, so you figured out what the pivot is just based on talking to people. You see the pattern and some of these. And obviously you can't go down every one of these rabbit holes and build everything for marketing and customer support and everything. How did you figure out how to stop becoming the AI guys and becoming something that made more sense and didn't send you all around the organization?
- [29:50]
B
Well, one, we actually decided that we wanted to be spent around the organization. A lot of people. And I still get this advice to this day, and maybe this is the reason, this is the thing that will kill us is like, we are building a product that is not vertical, specific it is for a very specific part of the horizontal market is for businesses that are series A, above series, slightly above series above Series A. They've like hit sort of like problems of scale, right. And they have people on the team that are wanting to use AI that is like our subset. But it is very, very different from like we sell a outbound product to sales teams. Right. Like we are a product for tinkerers. And so I will say, like, we are still center on the organization. And honestly, like, I will say we are still somewhat the AI guys. It is just now that our product can do the things that we had to kind of do manually, right? Like instead of setting up a jupyter notebook to try to test an idea and put it into production in a really horrible way for people, we now have a product where somebody can come to us. The other day, like, I won't name names, but a massive retailer came to us and was like, hey, I think people are stealing from our stores and putting the product on Poshmark. Can you please can your AI go and like, look at Poshmark? And the thing that within the we did was we actually were like, yes, but we are not going to build this for you. We are going to show your team how easy it is to build this thing in our product. And then we just did a session with their sort of like their operations team on how to build AI agents and we built it together. And then what happened was their operations people were like, oh my God, this is so easy. Let me try building other stuff. And that was the. We actually are still the AI guys. We just now teach instead of do things for people.
- [31:40]
A
Yeah, but I guess the trouble with just being the AI guys is it's so broad that you get asked about all kinds of things that don't make sense to what you're doing and then also calling yourself an AI agent or agent builder. You and I were talking about this earlier. It's kind of like the wild, wild west out there. Everyone seems to be doing some kind of AI agent thing. And I think it would be a lot easier if people could understand which one is right for me. Where does this thing fit in? How is this different or better than what else is out there? So how have you approached that?
- [32:21]
B
Yeah, and admittedly, I think we'd have more revenue if we had actually cracked this our take. And you know what? Maybe I should post a comment in 4 months to see if this actually worked. I think it is just about providing value. Right? You need to just provide value to People, I think that the biggest complaints we get from customers, one of the biggest things we've been doing recently is, like, companies like Qualtrics are selling AI packages on top of their products for like, a quarter of a million dollars a pop. And, and they're just bad. Like, they're terrible products. What they're basically doing is they're, they're slapping like, a chatgpt, like, webhook on top of their existing, like, data infrastructure. And they're like, we have AI now. Pay us a quarter of a million dollars to, like, give 500 licenses to your employees. And it's just like a bad product. And so that's something that we constantly fight, right? Like, what happens? My most common story is like, a sales leader will. Or someone will book time on my calendar, and they'll be like, hey, I have this problem. And we will show them the product, and then they'll say, oh, like, I'm just going to wait for, like, Gong to build this. And then what started to happen now is that they're using, like, Gong's version of the AI and it costs like, a gajillion dollars, and it's bad and it's not customizable. And so then they come back to us and like, oh, cool. Like, can your product do this? The way that we're, the way that we're trying to do this is we're just trying to show how easy it is to build these products. Like, if you look at my social media accounts, it is just. Most of the conversations that I'm having on there are about how easy it is to. I built. I spent five minutes building an AI this morning. The LinkedIn thing I did this morning was my investor messaged me and he goes, hey, like, how do. Like, what are the biggest questions that people are asking on Reddit about AI? And what I did was I used our Reddit bot on my phone to go scrape Reddit. And then, like, there's an AI agent that we've built that goes and scrapes Reddit and then actually dumps into Gamma, which is like a. A deck maker. And I just showed, I just showed those.
- [34:33]
A
So, yeah, I mean, I think that's super helpful. It goes back to kind of the example of, like, you. You were saying, like, just sending people leads. It's much more helpful to actually see an example. People can connect the dots as opposed to, here's my elevator pitch, here's my blah, blah, blah, AI agent. It doesn't make sense to anybody. I think we're still going through this period where we're waiting for the dust to settle and things to start to make more sense. One of the things was interesting was you said, you know, hey, we're agentic, and that's always an interesting term.
- [35:13]
B
Right.
- [35:13]
A
And the way you sort of told me about it was like, look, we're kind of like totally prompt based. We don't do this workflow and drag and drop. Like Zapier. People just type in what they want and get that set up. And immediately I thought about, well, Zapier has something like that and it's not very good because every time I've tried to use it, I type in the prompt, it generates something and I can see the workflow, and half the time it doesn't work. Another example was where using Claude code and using it inside Cursor or Windsurf or whatever your favorite IDE is, that was pretty cool, right? So you got your code, you got this window, you can start to do interesting stuff. And then when Anthropic released Claude code for the web, I was like, no, thanks. My head for a while couldn't get around. How do you do this when you can't see the code? You just have a prompt to do this all. And now that I've been using it for a while, it's actually pretty cool and it starts to make sense. So I'm just curious, why did you take that route? And you know what, what was the reaction like, did it make it easier to sell the product?
- [36:33]
B
Yeah. So I will say I think the only real reason we have this advantage is not due to any, like, intelligence. It's literally just because, like, Zapier, Zapier and nad, what they have to do is they have all of this infrastructure that they've already set up. Right? Like, they have to. They have two solutions when it comes to implementing AI. And clay is actually this exact same thing is their AI can either be an AI that has tool calls to set up a workflow in Zapier, which is something that they do well and something that their customers expect from them, or they can take a risk and go the other direction. And because we're newer, we took a risk and went the other direction. Right, because, like, the way the cloud code works, you explain this incredibly well. The reason that Claude code works so well is, and it's such a beautiful product to use, is that it is at its very core a system prompt, but it has access to tools that are very, very good at reading code and searching for code and understanding code structures and Accessing documents that a person has hopefully set up that explain the way the code looks. And you can watch it, right, like it'll, it'll, it'll have a prompt, it'll chuck some stuff into context. It'll use its like, tool to go and grab something or search for something or, or, or, or read something. It'll make a couple requests and it has these, like, it has these tool primitives that the AI can actually access. The only thing that Kotera has done is made it so you can take your system prompt and attach tools for the rest of your, the rest of the parts of your life, right? A tool that can access Gmail, a tool that can access Google Docs, a tool that can access Reddit, a tool that can go look at someone's, go read someone's like LinkedIn posts. We just have all of these tools and we just give people the ability to hook them up to a system prompt. And then the benefit of it is that you can attach it to a snowflake or a redshift or a bigquery to be able to run them at scale. And just because of the fact that we're newer, we were able to go that route. And I think that every company from here on forward that is going to be launching one of these. It's not entirely true because OpenAI release agent kit, which is in my opinion a lazy, horrible product that is also just a slapped on like Agent Builder. Watch, like I put this out the day this episode drops, the day two, they're going to announce Agent Builder 2, which is just going to like, put us out of business. But like, what this is fundamentally, that's fundamentally the only reason we had that advantage. It's not any sort of intelligence. It was just that this is a path that we were kind of forced to go down from our customers. And they kept asking us. And this was just the way that made, this is just the way that made sense to have our AI integrate with external products.
- [39:25]
A
So tell me how different the traction was. A lot of the times I've spoken to founders who say, you know, when we got traction or started to get traction, it was different. You could tell before we were like, we were kind of getting customers, but it was like, you know, we weren't there. But when things start to click, it just feels different. And so for you, you guys, first couple of years you're building this customer analytics product, you get to six figures in ARR, and then you did this pivot and in a relatively short space of time, you Hit the first million in ARR. So just tell me about that. Like what was different before and after with this pivot?
- [40:06]
B
It was just easier to close deals. Like we were closing. Like we were just, we. It was just like much easier. Like the product started to talk for itself and it was just easier to. It was just easier to start selling it. More people were interested for more reasons. And you. They just got visibly excited because they could see it, see it working. Like I could, I could jump onto a sales call and they could explain their problem to me and I can just, I could in the span of 3 minutes write not the best, but an agent that did the thing that they were asking for and that was like, oh wow, this is very cool. And that. And then they buy. We also figured out pricing a little bit. Like we didn't try to sell an enterprise deal. We gave people like more, more reasons to, to, to jump in. Is, is.
- [41:00]
A
Yeah, let's, let's talk about pricing because I think today when I looked at your pricing page, you have a $20 a month plan and then the next tier up is like $500 a month. Which, which tells me you're not really going after the core Zapier or N8N sort of crowd who would go from 20 bucks to 50 bucks to 100 bucks a month. It's almost like that 20 bucks is there just to let people try it enough. And the ones who would happily pay 500 bucks a month can then upgrade and move to the next tier. And you don't have a free plan. And you told me earlier free was probably a mistake. Just talk about there's a journey you've gone through with that.
- [41:48]
B
You get a seven day free trial. When you sign up for the 20, you actually get a seven day free trial. I think when you sign up for any plan, we're meant for teams and we're meant for workflows. Like the $20 a month plan is basically just gives you the credits that you need to be able to chat with a bot. And it's actually quite a nice product when you think about it. Because for $20 a month, you get access to Anthropic's models. You get access to ChatGPT's models, all of them. Right. Because you can just select in our product which model you want to use and then you can get it to access your tools. So if you just want a chatbot to talk to you that connects to Gmail and like can slack you every few hours, that's something that's really really easy to set up on our product for $20 a month. And you can just get rid of your ChatGPT subscription if you want, because what we're using underneath the hood is still chatgpt. But you have the benefit of, like, being able to use Claude for, like, it's better at coding and writing and a few other things. It sounds, it's less sycophantic, so it's better for those things. And. Yeah, that's exactly right. Like, you can come in for $20 a month, you can, you can mess with it and then for $500 a month, right. Like, you just need enough credits to be able to run this on top of a data warehouse. You need to be able to. You need to be able to run this very deep for $20 a month. You can come in and you can do deep research from a chat perspective, but you want to do deep research on top of everyone that's coming to your site. And you probably have at least like a few dozen people coming to your side a day that you want to do deep research on. And, like, you're just going to run out of credits really fast. Right. Like, it's just we. You can do a lot of damage for $20, but if you want to do anything at scale, like, it will cost you. So that's our. And again, like, I will asterisk this with the fact that, like, I'm sure that we will. We will put like an intermediary thing in there if we need to at some point. We're still early at this.
- [43:42]
A
Yeah. So before we started recording, I asked you, who do you see as your competitors? And your answer surprised me because I thought you were going to talk about other sort of AI agent builders. And you said, well, actually we're looking at Zapier, N8N and these types of guys, which is huge market, hugely horizontal. Why is that your focus?
- [44:10]
B
Those folks have been very, very successful. And foundationally, we are doing a lot of the same stuff that they are, but just with a modern twist. Right. Like N8N is still like a drag and drop workflow builder. Zapier is still a drag and drop workflow builder. We are the same thing. We are the same problem solution, we are a solution to the same problem, but it is just a much, much easier thing to set up. And it's like the agentic thing really does give you a lot of superpowers.
- [44:43]
A
Right.
- [44:43]
B
Like a great example of something that you can do with our product that is very difficult to do with N8N is someone just has a, has a. There's a major restaurant chain that uses us to go and look at their Google reviews and their internal Zendesk reviews. And the only thing the AI's job is is to send a pager duty notification if someone has said that they've gotten sick inside of a restaurant or like falling down inside of a restaurant. And if you use like a keyword based model, it's just gonna either miss stuff or it's gonna like ping constantly. It's very difficult to tune those things. Having an AI and just explaining to it, hey, like, this is the only thing that you need to like warn me about. It is quite a lot like easier to set up and safer. And so we are just a much better, in my opinion at least way of doing solving the same problem. And we're focused specifically on enterprises, which is the data warehouse integration piece. A lot of businesses do not want to send their data to a third party and have it process it and then have it do everything they want it to be sourced from their internal infrastructure, go to an LLM provider and land back in their internal infrastructure.
- [45:57]
A
It's interesting, when you told me about that, I was like, I'm not sure I ever thought about it that way. But when I look back and companies or CEOs that I've talked to who are kind of strategically looking at something like this, they're like, the conversation always seems to be about, we have this AI solution, how do we get all our data into there as opposed to how can it just sit on top of the data that we already have? It seems pretty obvious, but is it really like that unique like you guys are doing that?
- [46:31]
B
We haven't found anyone else that's doing the agents on top of the data. We have a lot of people that are okay. And when I say agents on top of the data warehouse, it is not like an analytics tool. It's not like, hey, tell me how much user growth I had. It is, I have all of this data and I need to run it. I need to run like a lead generation model effectively, right? Like I have this list, I have a list in mixpanel of everyone who's like signed up to my website. I need you to go down this list every five minutes and forever. Anyone who's new, I need you to like write me a debrief on their business using like Google Article Search right in like LinkedIn. And I need you to like enrich all these tools. Super easy to do that. It's always like stuff like that, look, I don't want to say anything here that is particularly mean. I think that a lot of the businesses that we see today that are hyper specific use cases using AI are going to get to millions of ARR very fast and then I think they will die. I think the reason for it is it's incredibly easy to build those things today. Right. Like the very vertical use cases. If you don't have knowledge or a moat that's specific to that industry, it's actually quite easy to like use AI to build those things. And so we think that in the next few years that just giving people access to tools where they can explain what they want to do is going to be like a very easy, a much just better way of approaching the same problem. Obviously those companies will win out in the short term, but they're also the easiest problems to go after and they make the most sense to investors, which is why they get funded out the wazoo. But we are trying to build an easy way for people to solve hard problems. And again, we might fail, like it might not work. We think it's working and it seems to be working well. But that is the thing that we've spent the last two years building is basically making it easy for these AI agents to sit on top of native infrastructure.
- [48:49]
A
It reminded me of a conversation I had with David Shim, the founder of Read AI, and he shared a similar story where they, I think it was the OpenAI API and suddenly his team built a solution in like over a weekend. And he was the one who was like, wait a minute, if we can build this over a weekend, that means everyone else can do the same thing. So that doesn't sound like a particularly smart way to go forward here. We got to figure out how we're going to do this differently, better, whatever, right. So, and I think we're going through that right now where it's like it's really easy to be a wrapper on ChatGPT or OpenAI, but when everyone's doing it or everyone can do it easily and Claude Code comes along and all this stuff and keeps getting better and better. Yeah, I think the bar is going to keep getting raised and it's going to be harder and harder to stand out.
- [49:49]
B
Yeah, and I agree with this. I think that one of the other things is that you need to be able to make these things work at scale. Right. Like the lovable V0 sort of problem where they've sort of like they very quickly got very, they very quickly got to a few, a lot of many millions of dollars of ARR. But then also their retention curve is terrible is because, like, engineering is difficult, right? You will like Vibe. Coding an entire app is easier today than it was yesterday. But we still have a lot to figure out. And the thing that. The challenge that we think that we solve pretty effectively is making all the things that you have to think about, like API rate, all the boring stuff, right? Like when you talk to, like when you're trying to, like, access information from Google 50 times in a second, you have to deal with rate limit issues and you have to deal with token costs, and you have to deal with budgets, and you have to deal with sort of like networking errors. We try to make it really, really easy to. For people to just. They just never have to think about the boring stuff, right? They explain their problem, they connect their tools and they save time and they launch it at scale.
- [51:03]
A
Cool. All right, we should wrap up. Let's get onto the lightning round. I've got seven quick fire questions for you.
- [51:10]
B
Let's do it.
- [51:11]
A
Okay. What's one of the best pieces of business advice you've received?
- [51:14]
B
Never give anyone anything for free. Always make sure that you get something for what you're giving. Your time is incredibly valuable.
- [51:22]
A
What book would you recommend to our audience and why?
- [51:25]
B
How to Get Filthy Rich in Rising Asia. It's a kind of weird book that's written in second person, and it's not what you think it's about. It makes you think about why you're doing the thing that you're doing. And maybe not in a good way, maybe not in a way that's fuzzy and happy.
- [51:46]
A
What's one attribute or characteristic in your mind of a successful founder?
- [51:51]
B
The ability to focus. Focus is very, very difficult. It's something that I find trouble with myself. I think somebody who can focus well and only focus on the things that actually matter are people that tend to, in my opinion, do pretty well.
- [52:06]
A
What's your favorite personal productivity tool or habit?
- [52:09]
B
I mean, I'm supposed to say us, but in reality it's definitely Claude Code. Like, Claude code is, I think, the tool that I has actually genuinely impacted the way that my team works and how I' work is because I've actually got it sitting in the corner right now trying to build an integration for me. We just have a part of our code base that's very, very well specked out, and I just run constant experiments with it all day. I'd say only 20% of what I build actually gets shipped. But as somebody who used to be an engineer and is now no longer really an engineer in terms of my job. It has gotten me back to engineering and actually building impactful things in an easy way.
- [52:49]
A
What's a new or crazy business idea you'd love to pursue if you had the time?
- [52:52]
B
The one that I find the most interesting is it would be really, really interesting to build something that is hardware related. I think hardware is going to be like a pretty unique point in the next few years. Like, I think software might become commoditized and hardware might not. And I think making it easy for these agents to be able to interact with a physical product, right? Like just creating a hardware device that has a very, very easy software interface for AI agents to be able to talk to so you can modularly build robots is interesting. I say that with absolutely no knowledge outside of like the LEGO robotics that I played with as a kid, but that is genuinely something that is very, very interesting to me.
- [53:31]
A
What's an interesting or fun fact about you that most people don't know?
- [53:34]
B
I'm from Nebraska. People always, like, find that super strange. They're like, wait, you're from Nebraska? Like, how did you. How did your, how did your. For people who are listening to this podcast and not watching it, like, I'm a brown skinned person. My parents are Pakistani immigrants. And the biggest question I get is like, how the hell did your parents move to Nebraska? It's a fun story for a different time.
- [53:55]
A
All right, and finally, what's one of your most important passions outside of your work?
- [53:58]
B
I live in New York City. It's my friends. Like, I love my friends more than anything else in the world. My friends and my wife. I think that, like, as I grow older, the opportunity cost does not become money, it becomes time. Like, the thing that I'm most afraid to lose is I'm in my late 20s. I live in one of the, what I think is the best city in the world, surrounded by the best people in the world. And the opportunity cost of doing this founder job is that I don't get to spend nearly as much time with the people that I love. And that is actually like a really, really, very, very real cost. I like, it's. I don't want to be a billionaire who's alone.
- [54:35]
A
Love it. Okay, great. Um, so if people want to find out more about Kotera, they can go to Kotera Co. That's C O T E R A co. And if folks want to get in touch with you, what's the best way for them to do that?
- [54:48]
B
You know, What? There's a direct line. Text me if you want to talk about AI agents and you want to ideate on building AI agents. My cell number is 402-468-8492. If you've gotten to this point and you haven't turned off the podcast and you want to talk about AI agents, shoot me a text. That's my number.
- [55:05]
A
You know, it's like, I don't know whether I'm just lucky and I have a great audience, but I've never had anybody come on the show and then say, I got spammed by these annoying people or whatever. But they have come back and said, you know, I had some really interesting conversations with people who are listening and followed up. And so, you know, thank you for being great listeners.
- [55:28]
B
Open your phone, you're probably listening to this on Spotify or Apple Podcasts or YouTube. Pop up in the phone, send me a text message. I promise I'll respond.
- [55:36]
A
Sweet. Well, thanks, man. It's been a pleasure. Thanks for unpacking your story and sort of the journey you guys have gone on since yc, doing the pivot and then what you're doing today with building the business. And I wish you and the team the best of success.
- [55:50]
B
Thank you so much.
- [55:52]
A
My pleasure. Cheers. If you're building an AI agent, a SaaS product, or stuck trying to scale, check out Gearhart. They can act as your fractional CTO and technical team, bringing AI expertise from projects for Meta and Google plus, strong Silicon Valley connections with founders and VCs. And since they're a Ukrainian born company, you get senior engineers with an offshore pricing model with offices in San Francisco and London, and a distributed team of 40 experts. They've helped build over 70 successful products. Right now they're offering our listeners the first 20 hours of development for free. Just book a call at Gearhart IO. That's Gearheart IO. If you've dealt with Twilio, you know the pain, clunky setup, confusing pricing, support that costs extra. SignalHouse is different. Modern APIs, fast ATP approvals and pricing that actually makes sense. No hidden fees, no run around when you need help. Whether you're building a SaaS product or a marketing platform that needs messaging, Signalhaus just works. Check out Signalhouse IO. That's Signalhouse IO. Most SaaS companies react to security breaches after they happen. Nord Stellar gives you real time visibility into what attackers see. Leaked credentials, exposed assets, session cookies from malware, brand impersonation attempts. Built by the team behind NordVPN. With access to one of the largest dark web data pools in the industry. Stop reacting to breaches only after the damage is done. Be prepared with Nord Stellar visit Sasclub I.O. nord to book a demo and mention code BLACKFRIDAY20 for 20% off. That's Sasclub I.O. nord.