$300K ARR, zero SDRs — how Amos Bar-Joseph's AI swarm replaced outbound

Welcome back to another episode of the Bridge the Gap podcast powered by Revenue Reimagine. Today's guest is Amos Bar Joseph, who is the CEO and co-founder of Swan AI and one of, if not the sharpest builders in the AI GTM space right now. He's sold two startups, scaled B2B revenue with automation, and is building a $30 million ARR company with, get this, just three founders in a stack of AI agents that happen to have some pretty cool names. But he isn't just chasing scale. He's actually questioning what gets lost in the process. This isn't going to be your typical hype conversation about replacing sales teams, go use AI, fire everyone. We're not talking about any of that. We're going to have a real look at what breaks when you go too fast, what works better than expected, and what the future of go-to-market might actually look like.

Amos, thanks for joining, man.

Wow. Adam, thank you for that intro. I wish I could just take you with me to, you know, dinners and social events that I'm in.

I will be your hype man. I will just stand up there and do the intro.

Let's do it. This is where you talk. Yeah, well, you messed up the script. I didn't know what you were doing. So hey, thanks for joining. Appreciate it, man. So, 300K in 30 days with no SDRs? Like, let's start where everyone's whispering about. How would you generate 300K?

Yeah. I don't do any operations, by the way. That's not my thing.

No SDRs, no ad spend. Walk us through how you did that.

Yeah. So, you know, I am a one-person GTM powerhouse. I move at startup speed but at enterprise scale. And you know, we did that as a team. It was a team effort actually. We're, as you mentioned, three founders on a mission to get to $30 million ARR with just us and AI agents. That means we can't throw bodies at the problem. We can't hire SDRs. We don't have huge marketing budgets. So we don't have that ad spend war chest. Does it even matter anymore? We're trying to prove it doesn't.

What we did basically is we tried looking at intelligence as leverage. We're trying to scale with intelligence, not with headcount. But the way we approached it is not by trying to automate ourselves away. It's not by hiring off-the-shelf AI SDRs to spam our market. It was more about how can we discover that 100x version of ourself, that 100x seller version of me? Not like what would be the 100x version of everyone, because we don't believe in that.

I want to hit this head on. I just posted about this yesterday and we've been talking about this. There's so many people out there that are just trying to buy like off-the-shelf agents. I'll go to Naden. I'll go to Gum Loop. I'll go to Make, and you want to just download something and get it running because people are lazy. Let's be honest. And like, that's not the easy button. You have to customize and build it out to what you're trying to accomplish. Talk more about that.

Yeah, definitely. So, you know, it's nothing new with AI agents. GTM alpha always comes from being different, not better. It's always about leveraging a playbook that maybe you have a unique insight into it, maybe it's very hard to accomplish. It's always about the fact that you're leveraging a playbook that others don't. If everyone is on the same playbook, then you're all fighting over the same attention and it's hard for everyone. So the quick wins are problems that are built into the go-to-market DNA since forever. And the best teams, they can just overcome it by fighting hard and actually discovering the unique playbooks that could drive the most ROI.

So what we did at Swan is we realized that automating ourselves away with off-the-shelf AI agents, that's the generic playbook. It doesn't work. We don't believe in that because what you'll get basically is like a very low glass ceiling of a cheaper version of yourself with poorer quality. So that's like the glass ceiling—just a worse version of me that I pay less for.

Not that amazing.

Wow. One hundred percent. And I'm going to double down on how important it is. What we did is we built our entire AI agent strategy and go-to-market centered around a human being. Not to replace them, but around a human being, and that human being is me. I have Edo, the CPO, and Neve, the CTO. They are AI agent wizards. They've built an agentic swarm around my strengths and weaknesses. And it all starts with my biggest passion, which is actually storytelling. So I'm not your regular seller. Most sellers are not regular sellers. Every seller has their own unique advantage, that unfair advantage they could double down on. And mine's storytelling. I know I have great ability to write LinkedIn posts and my LinkedIn game is on fire. So we realized that's an amazing opportunity to double down on. Not by building agents that could just help me produce more posts in shitty quality with less resources, but how can we build a funnel of agents around my entire activity on LinkedIn?

So it goes like that. Folks, if you're listening, this is how the funnel goes with our AI agents. I want access to this.

Oh, so yeah. And you can actually build an AI agent that could teach you how to build it yourself step by step. So if you want access to it, you can DM me on LinkedIn and I'll give it to you.

That's a whole separate conversation. Yeah, definitely. So if we look at the funnel, it all starts with actually writing the posts themselves. I have Shakespeare, which really helps me in a collaborative way to write viral posts that generate over 1.5 million impressions on LinkedIn. It takes me less time. Where it takes me four hours, it takes me 30 minutes. I produce much better posts because I have a thought companion that I can collaborate with. Okay, tell me what arcs can we actually approach with this narrative that I want to talk about right now? What amazing hooks can we have? Give me five options. Let's go that direction. Let's try to work on the first intro, etc.

So I have Shakespeare. We have 1.5 million impressions. Then what we have basically is 15,000 reactions to every post that I have. That's a lot of engagement. So we have the Observer, which actually monitors these leads and surfaces up hot ICP leads from the posts that people are engaging with. If you go one step below that, we have the Connector, which monitors my connection requests. I get 300 connection requests each day. So basically I have an agent that monitors these connection requests and surfaces hot opportunities, reaches out to them so I can pick it up only whenever there's a reply basically.

Then if you go one step below that, we have the Hunter on our website. If you rent on our website but you didn't sign up, the Hunter is going to spot you. If you're a hot lead, then expect a message, a personalized one from me. And if you reply, then I get into the loop and we can start B2B on steroids. Mega steroids. If like B2B and Clay and OpenAI had a baby and they had like a LinkedIn automation built into it. So that's the Hunter basically.

Amos, how do you stay out of LinkedIn jail with this? Because when you listen to the gurus, you know, it's like any automation with LinkedIn, you're going to be effed. And I don't believe you got to 300K in ARR by breaking rules and getting shut off. You're clearly thriving. How do you do it the right way?

Great. So first, we obsess over LinkedIn constraints. We have a very opinionated way of approaching it, also within our product itself. Because we sell agents to go-to-market teams as well and they operate on LinkedIn, we don't let you touch the capacity or the scale. We have an algorithm and a queue. What it does basically is it always looks at your queue because you're always at maximum capacity. My state is like always maximum capacity. I always have an algorithm on my queue that surfaces up the relevant leads in terms of date relevancy and how hot they are. So if it's an amazing opportunity, it can bump up other leads that are waiting in that queue. But this is something that we operate in our back end that we build for our customers and I use it for myself as well. I love it. It's super amazing.

Really quick follow-up question on this. Like when you started, like what system did you have on day one? Like when you're like, hey...

Me start this thing, was it just LinkedIn? Was it like LinkedIn and some, are you using like I-passes in the background? So are you referring to like our product, like the LinkedIn automation part, or like the entire agent swarm?

I think not the agent swarm, but like day one you're like I'm going to start this thing. Like people like the agent swarm type of thing, and people going down an agentic path, path orchestration. Like what's good for them? Like their build different than what you built. But like day one you're like people are like how do I get started on this thing?

Yeah, so what we did was nothing new to AI agents. We did, you know, maybe the opposite of 99% of the world. We actually didn't look outwards. We looked inwards, okay? We didn't ask someone which agents should I build? We didn't look at the hottest AI startups out there because we knew that 99% of them are BS. We just looked inward and we had a challenge to solve. We started with a constraint. That's the autonomous business model. We can't scale with headcount. So, okay, we can't hire someone to do it manually. How can we unplug that bottleneck just using AI agents? And it started as an iterative process. So every layer kind of like, you know, helped us discover another one. So, you know, I happened to have it all started actually with the connection requests. I started having so many connection requests and I said, okay, I'm just losing all of this volume. I can't even, I can, I can't even go and accept them because it's too much. I can't actually even go to my LinkedIn inbox and accept them. So we said, how can we solve that? So we went in the unity pile. When there's an API for connection requests, etc., that what people don't understand is that you don't need AI knowledge to become an autonomous business, to become an AI operation expert. You just need business processes, ops, and no-code expertise. That's all that is required to actually start building that muscle. And people are always looking for answers and quick wins outwards, but it all starts with a bottleneck. And then, you know, someone who's like super clunky but and scrappy, but, you know, wants to just fix it with some no-, low-code automation and knows how to map that process. And that's all you need. So you start iterating with prompts, etc. Right? You don't need to be a prompt genius to create an agent that understands who's the hot lead or not.

Right, Amos, man, that you're, you're, you're repeating everything Dale tells me every day. But that fireman, like we have to do the same thing in our business, right? Like a lot of the things that we do is go to market where it's all scalable. Like, it's like, how do you, what we started telling our clients is like, and it's a scary thing. Do you go traditional go-to-market or do you do this hybrid AI agentic? Because they've already gone down the traditional route. Like they didn't build from the ground up like you guys did. And think about, I can't go hire 85 people. So that now they're now they're in this weird hybrid state where they're like, I I need more top of funnel. I need more awareness of my product and service. But I could do, like, so I I have this example. You can either go hire three content editors and build a bunch of derivative content off of an ebook, for example, or you can start playing around with the side, let it generate content for you, but generate an AI system. Like, don't just use ChatGPT. Don't go use Claude. Like, you need to have a system in place to actually do that execution process for you. And then you can start weaning yourself off of potentially high-cost hires that may or may not stay around through that process.

Yeah, I I think that, you know, most of the companies are playing the wrong game right now, which is like looking at roles, functions, processes, how can we automate them with AI? That is not the right type of thinking actually. And I think that type of thinking leads you to just trying to figure out outside who could help you with that without really developing that, you know, AI competency inside of the company. And second of all, it has again that low glass ceiling. All you can do is just, you know, have something a little bit worse but cheaper. That's what you're looking for.

Not what you try, what you know, what you need to do is to reimagine fundamentally. How can that process look like with human-AI collaboration at the core? And so, you know, we believe that SMBs are actually much well-positioned for the LLM revolution. That's the autonomous business concept. We believe that they have the ability to restructure their entire DNA around human-AI collaboration. But enterprises as well, they have, as I see it, like two routes. They have the painless route, which leads to very mediocre performance, and they have the painful route, which leads to exceptional performance. The painless is, you know, taking their same processes, not trying to really innovate from within, but taking the same system that they have and try to put AI on top of it so they can have fewer employees doing the same amount of output. That's the painless mediocre outcome. But the painful exceptional outcome comes from reimagining these processes from the bottom up. Maybe they need to open a new sub-org within their go-to-market team. So a new SDR team that could actually build that, you know, culture from the ground up. Maybe they can't even take their own team and change it. They need to start it from the bottom. That it's a painful route, but that would actually lead to that 100x improvement to the fact that they won't be left behind when their competitors become AI-native.

People too scared. It's a scare, like the mindset shift is not, like people are scared.

Sorry, I no, you're, you're fine. This is all great. I I I want to I want to pivot slightly. Um, so I agree with everything that you're saying, but there's this kind of wave right now of teams that are jumping on this AI bandwagon specifically for like AI outreach, right? Like AI SDRs and even like whether it be AI LinkedIn responders, but like really taking it to the extreme of like AI cold callers, which by the way is illegal. Um, AI SDRs, AI AEs, AI sales enablement, like you name it, it's AI. Um, where in your opinion, Amos, where's the disconnect between like expectation and like reality of what should be versus like what people think they should be doing?

People buy from people. That's why companies who invest in meaningful connections win. The best part, gifting doesn't have to be expensive to drive results. Just thoughtful. Sendoso's intelligent gifting platform is designed to boost personalized engagement throughout the entire sales process. Trust me, I led sales for a Sendoso competitor and I could tell you no one does gifting better than Sendoso. If you're looking for a proven way to win and retain more customers, visit sendoso.com.

Yeah, so, um, first of all, this spam cannons um approach, it's nothing new. So um, you know, in the GTM space, since I don't know, like, maybe 2010, like 15 years ago, we had the sales automation golden era, right? Um, so there was a specific point in time where spamming prospects were valuable. Right, it was maybe 15 years ago, okay? And since then, it's not that valuable to spam prospects. And the value of spamming prospects just decreases, you know, very steadily over time. And AI just takes that, you know, momentum and amplifies it. So, um, okay, spamming, we had, you know, personalization, we had actually, you know, SDRs at a lower cost. So you can always outsource your SDR function and have, you know, cheaper output for lower quality. That something that existed. So it's nothing new. It's just, um, you know, a continuation of long-lasting. AI is the new offshore.

Yeah, exactly. It's just, I blame marketing for all of it. It's always started like marketing automation started like this big trend, and like marketing will completely destroy anything that is getting any little bit of like traction.

Yeah, so I I actually I actually think daily that every software wave promised you know sales teams to give them more human interactions, but actually, um, you know, left them with more system interactions than human interactions at the end. And um, it's all because these companies have great marketing and go-to-market teams that eventually try to educate the market about, you know, a way to approach it. But it's only a moment in time where you have that alpha, and it just disappears. And so I just want to get back to the AI SDR analogy here. Um, what we're seeing is the first wave of AI applications, okay, that is the first wave. So Steve Jobs calls the first wave of a technology revolution as skeuomorphic design. What does it mean? Skeuomorphic is a bad name, but um, it means that the first wave tries to mimic what we have, um, just like putting the new tech on top of it basically. So it lacks the creativity and understanding of this new technology. When we look at websites, for example, when websites came, so the first wave, you know, we had Barnes & Noble, a bookstore, said yeah, let's put all our books on a book catalog on the internet. It was like a read-only website. And they were pretty early to that revolution, but they just, you know, they had a read-only website with look at our books, this is what we have. But then came the second wave of Amazon who imagined a digital-native bookstore. And we know where these two companies are. So what we're seeing here, it's the same unfolding of, you know, AI. So the first wave, we have SDRs. Let's create AI SDRs, right? We have support. Let's create AI support agents, right? It's like, let's do the same thing just, you know, in a shittier way, right? But the second wave, which is what Swan...

# Transcript

Promised to bring to the world is a different use of the technology when we try to reimagine that process with human AI collaboration. And so what we bring to the table at Swan that is extremely different than all these AI SDRs is actually we look at ourselves as an AI go-to-market engineer. Okay, so we are actually a platform that is an agent that can help you build AI motions for your business. Specifically, what we believe is that GTM alpha comes from differentiation, not from being better or more activities. It comes from being different. And so we're moving from an era where there was an app for this, where SaaS solved use cases and built features for use cases, to an era where AI agents build solutions for you for your business. And the AI go-to-market engineer helps you build agentic GTM motions that fit your go-to-market DNA.

If you're a cold calling organization and you have a unique understanding of your buyers and messaging and positioning, then you need that AI GTM engineer resource to build an agentic motion that supports your cold calling. You don't need an AI SDR that will replace your callers because you're doing a good job. You just need to amplify that. So how can you find something that will double down your cold calling efforts? You need an AI GTM engineer.

And so what we're promising is finally, from SMBs and go-to-market teams that bent around their tech stack, finally your tech stack bends around your business. It's a value equation. This has always been a value equation thing. And the differentiation piece—I think what is happening today, people are just trying to differentiate any way they can. Like they don't know how to differentiate and they forget about the fundamentals. Like, why did your product or service come to market? What's that value proposition and why should your buyer care about it? Now, how you build it and they will come. Swan did that, right? Build it and they will come. Right? One of how many tens of thousands that actually have an idea and a product that is so good that people do see it and they're like, "I will actually disagree guys. I will disagree. We didn't build it and they came."

It's just because it such a core aspect when I talk to young founders, the number one tip that I actually give them is product last. Okay, so in the last 20 years, it was all about building an MVP, right? That's the old startup playbook. And so the old startup playbook was build an MVP, was built around it. And so what happened? It became more and more expensive to actually build that MVP. And so startups raised more and more money to actually do it, et cetera. And we ended up at that growth at all cost playbook.

But what's happening in the last five years, which is interesting, is building products became super easy, but going to market became super hard because the fight over attention is the hardest thing to do right now. And when you're entering an extreme mode of that situation when you have these AI developers like Lovable and Base4, and you can build an app—everyone can build your app in no time—the hardest thing is actually to take it to market.

And so we built Swan, our product. We built product last. First, we built a movement around the autonomous business concept. And we started by that. And I grew my LinkedIn followership not because people love our product, not because people use our product, but because they believe in this new concept of scaling with intelligence, not with headcount. And the autonomous business movement—when I talk about it, I don't mention Swan and how we leverage intent to generate pipelining. That's not what I talk about. That's boring. What I talk about is the future of business operations and the operating system of a business with human AI collaboration at its core. So we actually build a movement first and product last. It's a story arc that people get enamored with, and you need to build that story arc so that they understand where they are and where they should be going. And if they don't get there, their customers will get there and they'll eat their lunch.

So yeah, I agree. I want to double down maybe as a last thing on what Adam said, that you know, one out of tens of thousands of businesses can generate sustainable growth from a good product. Okay, to generate sustainable growth from a very good product—that's like, it's not even a unicorn. It's much, much more rare than that. And so what people need to invest more in, and I feel like it's getting a lot of really overlooked in GTM specifically, is in your story. People today fall in love with the story. They fall in love with the people behind that story. They don't care about brands anymore. And people should start embracing that because if you're not doing a really good job on your story and on the people behind that story, the faces of your story, then you're leaving very hard work for your sellers.

Totally agree. We talk about this all the time. It's the origin story. It's the value prop. It's tying it together. What is the problem you solve for your prospects in 30 seconds or less? And what made you want to do this? I couldn't agree with you more. And we talk in sales—forget about founding a company and building a product—but we often say the best sellers are the best storytellers. They're able to take a complex or simple problem, build a story around it that resonates and makes you feel deeply. Sales is the transference of feelings. So like exactly what you're saying, the story. Yes. Don't get me wrong, although I will say I was going to say you don't have to have a great product. But you do—you can't have a bad product. There are some companies that have shipped products that can be successful for a certain amount of time, but eventually that comes out. Put a great product with a fantastic story, a charismatic founder, a great sales team, use AI to leverage that. And I love what you said about not having AI for the use case, but having AI for you. So the difference of using AI versus building AI. This is where I think go-to-market is going. And I think very few people are getting it right.

Awesome. One more question, then we'll go do a couple of fun things. I'm trying to decide what question I want to ask you, and I think this one makes more sense first. So rebuilding GTM from first principles—we talked about this a little bit. My understanding is you rebuilt GTM many several times. How's it changed and how you approach it? And I'm going to follow it up with a couple of questions. What beliefs about GTM have you completely dropped? Like, what is that complete nonsense? What does your team not do that most startups do? And I think I know a lot of that answer. And then, what would you never outsource again?

Okay, so let's break them down one by one. These are big questions. One at a time. Big, big, big questions. I feel the pressure. I feel the heat.

So first of all, one thing that I've learned—maybe the most fundamental one—is that playbooks can take you a very short distance, right? So if you're always looking at other people's playbooks, then you're always a step behind everyone basically, or maybe you're following your competitors. It depends where you want to be. And if you feel like being the tenth or hundredth player in your space, then that's okay, and then learning playbooks is important and you need to keep up. And so depending on what your goal is as a company, you need to actually align to that.

When you're building a startup, then usually you're trying to be the best, right? And if you're not—maybe if you're building like a regular business, a lifestyle business—then learning the playbook is okay. But what I realized is that when you try to become the number one player in your space, everything becomes easier. Okay, so that being number one mentality unlocks everything for everyone basically. And if you try to settle down from being number five, then everything becomes harder basically. And just becoming number five is much harder.

So the playbook will get you to the tenth number at maximum, right? So you need to actually reinvent the playbook by understanding your unfair advantage as a founder, as a GTM leader, and as a company. And so if you don't understand what is your unfair advantage and you don't build your playbook according to that, then you'll end up number ten, number hundred, maybe you'll just close the business.

I like that. Yeah, I think the playbook can be super overrated.

What will you never outsource again?

So I will never outsource pipeline generation as kind of like a service, right? If someone promised me leads, right? So like, "I will generate pipeline for you," right? It's like outsourcing your product manager. Like, yeah, maybe I'll have a product manager that just tells me what to build for my company. And yeah, you should build an AI SDR. And I would say, "Okay, you're right, I will build an AI SDR. That's good advice, thank you." It's not like that basically.

You need to understand the top of the funnel. It's like building—so Brian Halligan, you know, CEO founder of HubSpot, he has this notion of looking at go-to-market as a product. I love it, okay.

# Transcript

Feel like you have this notion of go-to-market fit, right? Like product-market fit, and you need to iterate on that. You always need to understand how do you find the perfect motion, right? And that starts from the top of the funnel. And so you can't outsource that top of the funnel to someone who just brings you leads. You can work with experts, with consultants to help you develop that in-house. Maybe you want to use specific capabilities outside of the company. So like if you want to reach massive scale and you have an agency that can send millions of emails a month, so you can outsource that and use that capability if you'd like. But don't ever think about just letting someone generate pipeline for you. I would venture to say even doing what you do and your company doing what you do, you still get those pitches all day long of, "Hey, Amos, I could generate 10 leads for you per day and you only have to pay $22 per lead." And it shows like if your targeting is this poor, I'm certainly not going to trust you to do it for me.

100%. All right, we're coming up on time. Let's jump into some fun rapid fire. I feel like we could go for another hour. Unfortunately, the show's only supposed to be 30 minutes, but we're likely going to need to change that.

Emma, what's one part of your tech stack other than Swan that you would never trade or give up?

So that's easy. I would say Claude from Anthropic. It's a ChatGPT alternative. Without it, I feel like I would lose my arm basically. I just recently started doing a lot more in Claude than ChatGPT. I used to be the opposite. I think they each have their uses. ChatGPT can't design anything, can't format anything for its life, whereas Claude gets it right almost every single time the first time. Claude is really good at content writing, by the way.

Yeah. Okay, good to know.

Finish the sentence: In three years, sales development will not be called sales development. What will it be called?

So I believe that the future is like full lifecycle growth operators that care about what I call sticky revenue. So sticky revenue is the end goal of your business. And when you have sales development and you have MQL people like marketing are in charge of MQLs and that's it, and sales development are in charge of meetings booked, and AEs are in charge of meeting quota, and CS are in charge of retention. What you get is this misalignment of incentives that really breaks the funnel down into a zero-sum game. And when you look at a persona that can orchestrate context around different types of places, but all they care about is sticky revenue, like I do as a single person operator. I care about if I look at a bad lead, I will never close that meeting. I can look at it as a demo and I see that's a bad lead. I won't close it as a customer actually. And also if I'm at CS and I see a customer that reaches out for a question, I can understand this is an opportunity for an upsell. So the fact that I have that context all across the board because I use AI agents, not because I'm Superman, brings the ability to increase sticky revenue. And that's what you really want as a business. Sticky revenue.

Love it. What's one common startup best practice that you absolutely actively avoid and stay away from?

AB testing. AB testing is a terrible idea for startups. So basically, the only advantage of a startup—the only one, there's only one advantage—that a startup has against an enterprise, against the incumbents, and that's decision-making velocity. A startup can make much more decisions in a given week than an enterprise. All the rest, an enterprise has an advantage over a startup. And so the fact that your only unfair advantage is having better decision-making velocity, AB testing hurts that ability because it slows down your ability to make decisions. So instead of doing an experiment, measuring it over a period of time, and then deciding what to do, go with your intuition, try to measure it. If it failed miserably, do something different. But decide, decide, decide. You could do that with AI much easier than you could do that with hiring three people. So I think that is a really key part of moving fast and going forward.

What scares you more—going too fast or going too slow?

Going too slow. Going too slow definitely. From even from a CEO perspective, from a personal perspective, that's my biggest fear. You know, just drowning in stagnation. It's where I feel like I'm at the least of my potential. And when it's hypergrowth, that's where I shine. I'm good. I'm very resilient, so I don't really get afraid with hypergrowth. I love it.

You have two little swans behind you.

Yes. Yeah. They just flew in seconds. I noticed them about halfway through. Are there more swans in the house?

Yeah, they are everywhere basically. And you know, every day passes, you'll see more and more swans. We believe that the unicorn playbook is dead. No one really wants to build a unicorn. It's about valuation inflation. It doesn't really relate to value. Building a swan is actually about value creation. It's about ARR per employee. It's about, you know, starting as the ugly duckling where you don't get these massive funding rounds and massive TechCrunch headlines, but you grind, you build real value, real ARR per employee, and then you become this elegant, aerodynamic beast.

And that is a great place to end it. And that is a take I agree with. Amos, thank you so much for joining. Folks can find you of course on LinkedIn, getswan.com.