How do you build a $30 million ARR business with just three people and a fleet of AI agents doing the heavy lifting? In this episode of Tech Talks Daily, I connected with Amos Joseph, CEO of Swan AI.
From the moment we joked about AI notetakers silently observing our conversation, it was clear this discussion would go beyond surface-level talk about automation. Amos is attempting something bold. He is building what he calls an autonomous business, one designed to scale with intelligence rather than headcount.
Amos has already built and exited two B2B startups using the traditional growth-at-all-costs model. Raise early, hire fast, expand the vision, chase valuation. This time, he is rewriting that script entirely. Swan AI is built around ARR per employee, human-AI collaboration, and what he describes as scaling employees rather than scaling the org chart. With more than 200 customers and only three founders, Swan is already testing whether AI agents can run real go-to-market operations autonomously.
We explored why over 90 percent of AI implementations fail and why grassroots experimentation consistently outperforms executive mandates. Amos argues that companies looking outward for AI solutions before understanding their internal bottlenecks are simply scaling chaos. The organizations that succeed start with process clarity, define what humans should do versus what should be automated, and then allow AI to execute within that structure. It is a powerful reminder that becoming AI-native has less to do with tools and more to do with operational self-awareness.
We also unpacked the difference between automation and agentic AI. Traditional automation follows deterministic steps coded in advance. Agentic AI shifts decision-making power to the model itself. The AI decides what to do next, introducing statistical reasoning rather than predefined logic. That shift in agency changes everything about how workflows operate and how leaders think about control.
Perhaps most fascinating is how Swan generates pipeline entirely through LinkedIn. No paid ads. No outbound. Amos has built an AI-driven engine that creates content, monitors engagement, qualifies prospects, and nurtures relationships at scale. It is an experiment in trust-based distribution powered by agents, not marketing budgets.
This conversation reframes what growth can look like in an AI-native world. If scaling no longer equals hiring, and if every employee becomes a manager of AI agents, what does leadership look like next? How do founders build organizations that amplify human zones of genius rather than bury them under coordination overhead?
If you are questioning long-held assumptions about team size, growth, and AI adoption, this episode will give you plenty to think about.
Useful LInks
[00:00:04] - [Speaker 0]
What would a company look like if it was designed to scale with intelligence rather than just headcount? Well, my guest today is building exactly that, an autonomous business where AI agents run core go to market workflows. Well, humans, they can focus on what they're good at, judgment, creativity, and direction. Well, today's conversation goes far beyond surface level automation, and we'll get into how businesses really change when decision making shifts from rigid systems to adaptive intelligence. And we'll talk today about why most AI transformations fail, why small teams experimenting at the edges, how they often outperform those top down mandates, and what happens when founders rethink how their growth actually works?
[00:00:56] - [Speaker 0]
So what does it mean to build a company that operates at the speed of thought? Well, let me introduce you to my guest now. We'll get into this. So thank you for joining me on the podcast today. Can you tell everyone listening a little about who you are and what you do?
[00:01:14] - [Speaker 1]
Yeah. So, Amos Bar Joseph here, CEO of Swan AI. We're building the first autonomous business. Basically, a business that could scale with intelligence, not with headcount, and can operate at the speed of thought. We're just three founders already supporting more than 200 customers, and we're trying to crack the code on a business that is designed around human AI collaboration rather than human to human coordination.
[00:01:48] - [Speaker 0]
Wow. That is incredible. Three founders, having, well, over 200 customers already. What do you do for people that are hearing about Swan AI for the first time? What kind of problems would you be solving for business leaders right now?
[00:02:00] - [Speaker 1]
So our customers, they call us lovable for GTM. So the folks who you who don't know, Lovable democratized software engineering for everyone, enabled each person to actually build their own software. What we're doing is we're democratizing GTM engineering, go to market engineering. So for businesses, who are trying to improve their go to market processes, then there's a lot of engineering work required. They need to work with their systems and to build custom code and workflows, optimization, things like that, and it requires a lot of effort and time.
[00:02:36] - [Speaker 1]
And it makes them slower and inefficient. And so we're giving them an AI go to market engineer in the pocket of each person on the team that could instantly build whatever they need, basically, in go to market and turn any GTM process into an agentic workflow in seconds, you know, from prompt to pipeline so they could really scale that process with intelligence, not headcount.
[00:03:00] - [Speaker 0]
Incredibly cool. And before you join me on the podcast today, I was reading how you've argued that real AI transformation doesn't actually come from executive mandates, but actually from small teams experimenting at the edges. So what convinced you that grassroots experimentation beats that top down strategy when companies want those real efficiency gains? And I think it's such an important question now when, what, a year, eighteen months ago, so many businesses were struggling to find ROI in their tech projects. But tell me more about them.
[00:03:33] - [Speaker 1]
Yeah. I think the the reason why over 90% of AI implementations fail is because, you know, to become an AI native company really has nothing to do with learning about AI itself. Basically, if you don't have a good understanding of your own processes and bottlenecks, then AI would just scale, you know, your chaos. But if you do have a good understanding of your business and you can identify exactly where are the bottlenecks, map a great, you know, new process and how this could be looked like when it's improved, understand what humans should do, and then what do you want to automate. Then if you have all that figured out, the AI becomes the easy part.
[00:04:15] - [Speaker 1]
And I think what people are trying to do is they're trying they're looking for shortcuts. They're looking for that top down approach that says, we should become AI native. We should do x y zed. And if we'll do if I'll give chat GPT subscriptions to anyone, we'll see 100 x improvement in efficiency. And that's not the real story.
[00:04:32] - [Speaker 1]
Though what happens in the 10% of these successful AI implementations, I know that because I've seen hundreds of those with my company, with Swan, across different customers and different sizes, the what's the common ground for these 10% is that they had a good understanding of what's not working right now, and they had a hypothesis of what that process could look like if it would work in a right way. And so they did all that work before they came to a solution out there. They started by looking inwards, not outwards. They were able to specify, and then they were able to review the output of the AI and improve it over time. So eventually, Neil, it comes down to your ability to specify and review the output of the AI.
[00:05:24] - [Speaker 1]
And without having a good processes, systems, best practices understanding, then you won't be able to specify and review, and you won't be able to work with AI at all.
[00:05:34] - [Speaker 0]
And on a personal note, looking at your backstory, your origin story there, you're someone that's built and sold companies in traditional growth at all cost kind of models. So what was it that pushed you to throw that playbook away, the one that that you built your success on, an attempt to reach $30,000,000 ARR with just three founders and AI agents? It feels like a big change.
[00:05:57] - [Speaker 1]
Yeah. So Swan is not my first rodeo. I've built and scaled two companies before that. Both of them were, like, in the b to b customer communication space, and both of them were based on the old growth at all cost unicorn playbook where, you raise a lot of funds before you even know who you're selling to and you really have anything around product market fit. And then you scale to 40 people before you've reached your first million dollars in revenue.
[00:06:24] - [Speaker 1]
And then each round, you kinda, like, roll the bluff. You expand the vision and total addressable market, but not the real metrics that are important to the business. And what I realized is that it creates this dynamic where the business is really optimized towards valuation inflation and not towards value creation. And I felt like it's not for me. And when I wanted to open my third company, I've seen that, okay.
[00:06:50] - [Speaker 1]
There's AI agents that really changed the game. If you look at the playbook of building a company, it hasn't been changed for the last fifteen years. People are still talking about MVP, the minimum viable product, and product market fit. These are terminologies that were coined in 02/2009 by Eric Rice in his book, The Lean Startup. So nothing has significantly changed since 02/2009, and I felt like this is the time to reinvent how do you scale a business in the modern AI native world, a business that is optimized for human AI collaboration, not human to human coordination, and is focused on ARR per employee, not valuation inflation.
[00:07:33] - [Speaker 1]
It's focused about ARR per employee because it's just the best metric to measure leverage, AI leverage. If you can expand the amount of revenue contribution of each employee at the team, it does two things. One, it positions the human at the center of that business, not the AI. And then what you're trying to do is you're to scale your employees. And second, it's a great way to measure if you're using AI to create more leverage and to enhance your employees.
[00:08:04] - [Speaker 1]
So it really centers the business around the right direction.
[00:08:08] - [Speaker 0]
Incredibly cool what you're doing here. And Swan is running real go to market operations through autonomous agents rather than just adding more human headcount. So what was it that surprised you most once you started letting intelligence, not people, run those GTM workflows? Any big big moments there or any big lessons learned?
[00:08:28] - [Speaker 1]
Yeah. I think that what we're seeing is that it's called, like, the collapse of the middle, basically. And the founder of Linear coined that, where we've we started seeing that in coding and software development where, you know, engineers and developers stopped doing the coding itself, the execution part. And what they were focusing is on specification and review, like the back ends of the process. So execution in the middle collapsed.
[00:08:58] - [Speaker 1]
AI is doing that. And what developers are doing, they're specifying and reviewing the work of the AI, specifying and reviewing. And what I've realized after working with dozens of different AI agents in go to market is that, since AI is really collapsing that execution, we're just specifying and reviewing. What we're transitioning is from, you know, systems engineering. When we were used to working with systems and moving data from places and trying to build these workflows and understand how to work with these systems was a skill, now we're transitioning into context engineering.
[00:09:32] - [Speaker 1]
So from system engineering to context engineering. I'm just chatting with with the AI. I'm talking to Swan. I don't need to understand the technical literacy here on how to move things. I can just talk to Swan and tell it what I want, but I need to understand how to structure the context of each task.
[00:09:49] - [Speaker 1]
How should I explain it? How should I specify it? How should I communicate it? It really becomes about communicating context and engineering the context so that the AI knows what to do, when to do.
[00:10:01] - [Speaker 0]
And there will be many people listening and founders may be listening as well. When they hear the phrase agentic AI, they might automatically assume automation or tooling upgrades. So how do you explain the difference between automating tasks and building an autonomous business model? It's a very different things.
[00:10:20] - [Speaker 1]
Yeah. Agree. And so it's an important distinction. So we love making a distinction between AI automation and agentic AI. It's not the same thing, basically.
[00:10:31] - [Speaker 1]
And the difference is basically in the power balance between code and the AI itself. It's kinda like there's a power balance between them. And what does that mean is that when we're automating something, then the code, the software decides what to do at each step of the process. So we're saying we're we can go to, like, a no code automation tool, and we can say, you know, step one, check my email. Step two, if the email has a PDF file, you know, proceed to step three.
[00:11:04] - [Speaker 1]
Step three, put that PDF file in a Dropbox folder or something. Okay? And so each time, 100% of the times that this workflow will run, these are the the three steps that will be performed because the code deterministically says that this is what you should do. And when you have an AI agent, what's happening is that that power balance shifts, and now the AI, the the large language model, is the one that is deciding what steps should be taken, not the code. So we're just pushing tokens into the large language model that says, this is what's happening.
[00:11:41] - [Speaker 1]
And the large language model comes back and says, this is what I think you should do. You should go to continue to step two. You should jump to step three maybe. This is statistical model. It's undeterministic.
[00:11:51] - [Speaker 1]
You can't foresee 100% of the times what would happen. And so a Genetic AI is that where the decision making of the process is really being controlled by the AI, not by the code. And if we have just a workflow automation where we just each step, we're calling to the AI to answer something, but it's happening the same thing over and over and over because the code says this what should happen. This is it means that it's automation. So to wrap it up, it's about agency.
[00:12:23] - [Speaker 1]
Age agentic. Agency. It's the a we're giving the large language model the agency to decide what to do. We're giving control. And before we had the control because we codified that control into the software.
[00:12:39] - [Speaker 1]
Now we're not codifying it. We're giving the agent instructions on when to do what, but it's the agent agency to decide.
[00:12:50] - [Speaker 0]
And something else I love about what you're doing here is you're building Swan very publicly, sharing what works and what breaks. So why was was building in public so important to this experiment? And what have you learned in in being this transparent? Feels incredibly brave and and forward thinking too.
[00:13:09] - [Speaker 1]
Yeah. So I think, you know, people, especially in entrepreneurship, has this tendency to keep everything close to the chest. And it's actually the wrong intuition because billing in public has so many advantages. It's even hard to count. But if I'll try to sum it up to the top most important things that you gain from pursuing that route is a, you know, to build with the you're not building in public.
[00:13:36] - [Speaker 1]
You're building with the public. You're basically, you can get their feedback. You can get, you know, their insights. You have this hive mind coordination where you have a problem. You can share it.
[00:13:46] - [Speaker 1]
You can see what they're saying. You have an idea for a feature. You can share it, etcetera. You're messaging. So you can do a lot of messaging, message market fit, message validation.
[00:13:55] - [Speaker 1]
You're talking about something. You can see how the market reacts to that message. So it's basically building with the public. So that's one thing that is super powerful. The second thing is that it allows for introspect, rethinking in the most extreme way because you have to sit down and think, okay.
[00:14:13] - [Speaker 1]
I wanna talk to the public. I wanna tell them something about what I'm doing today. What should I talk about? What should I say? How should I frame it?
[00:14:20] - [Speaker 1]
Then you can capture a moment in your business that you wouldn't have the time to reflect on, and you can just start breaking it down. What did I do there? How did I solve that problem and realize, oh, there's actually a framework. I I was actually thinking in that framework where that was laid down in that specific challenge, and I used that framework to solve it. So that introspecting thinking really creates methodologies and frameworks and allows you to codify a lot of the problems that you're solving and how you're solving them in a way that other people in the business could enjoy and even you could articulate better to yourself and repeat that.
[00:15:00] - [Speaker 1]
And so you're improving yourself by this introspect process that you're creating by building in public.
[00:15:08] - [Speaker 0]
And another thing that really stands out is your entire growth engine runs on LinkedIn. But what makes it even more impressive is it's without outbound or any paid ads. So what principles guide how you turn content, trust, and distribution into a consistent pipeline rather than just another page of vanity metrics?
[00:15:30] - [Speaker 1]
Yeah. So my intuition around marketing was always that there's a rule of thumb that says the more popular a marketing playbook is, the less efficient it gets. So, it really guides me to look for all these, hidden opportunities, these untapped opportunities to grow the business. And, for a lot of folks, just creating content on LinkedIn is really hard to measure, and it's hard to justify, and so they don't pursue that playbook. And it's not really popular in the sense that every business is doing that.
[00:16:06] - [Speaker 1]
Moreover, it actually builds a personal brand, takes a lot of investment, a lot of time, and the return on investment is not really clear most of the time. So most of the fund founders are not really doing so. And I realized there's an opportunity here. And because I love storytelling, then I realized, wow. This is perfect for me.
[00:16:25] - [Speaker 1]
And I started posting on LinkedIn, sharing the autonomous business journey with the public, how we're building this unique business model, the wins and losses, and I started getting some traction. And then over time, what we did is we created an AI engine around that content creation process. And what happened is that I started working with agents to create these posts and then working with agents to monitor the engagement on these posts and then working with agents to qualify and reach out to the folks that are engaging with these posts, etcetera. And we built an entire pipeline generation machine around my LinkedIn game. And that really changed how we grow Swan.
[00:17:06] - [Speaker 1]
We're not paying for ads. We're actually investing in creating the best content out there, the most authentic content out there so that people earn we can earn their trust, and they follow us to learn more about their own business. And, eventually, it leads them to actually trusting us enough to say, I wanna work with Swan, and I wanna buy their product.
[00:17:27] - [Speaker 0]
And you're someone who does not shy away from challenging legacy thinking, legacy assumptions around everything from team size, AR per employee, and scale, etcetera. So I I'd love to give everyone listening a valuable takeaway here. What what belief about company growth do you think founders need to unlearn most urgently right now and and change that mindset? You you must see a lot of the same old mistakes, but, what do you see here?
[00:17:56] - [Speaker 1]
Yeah. I I would say the first thing that you need to unlearn is that scaling equals growing headcount.
[00:18:02] - [Speaker 0]
Yeah.
[00:18:03] - [Speaker 1]
And I think that's the the biggest shift that we're seeing right now because of AI. And what you need to to start changing is the entire mentality of the business. Like, the old mindset was I call it, like, cog culture is where, you look at employees, they're just cogs in a bigger system where their entire purpose is to scale the business. And and the entire system is designed around how each employee could scale the business. It's given a small scope.
[00:18:35] - [Speaker 1]
It should operate within that scope, and that's how, they will operate within the company. And what the autonomous business is doing to that model, which I think every founder should implement in their own business, is it flips its on its head, and it says the the business exists to scale its employees, not vice versa. Yeah. And what we're doing basically at Swan is we're trying to ask, okay. How can we 100 x each person on the team?
[00:19:01] - [Speaker 1]
How can we create the 100 x engineer, the 100 x product, the 100 x seller? And we're thriving as a business to unlock the people that are working within the the the business. And so we have this framework where we look at their zone of genius, their intersection between their skills and passion that could create disproportionate value for the company. And we try to build a system around them, which includes a lot of AI agents that automates the mundane, the what's outside of that zone of genius, what's deflating them, and amplifies what's within the zone of genius, inflating them, and enabling them to move faster within their zone of genius. And when we're building that system, what it does, it scales its employees, and that's how you can create disproportionate value and get to a small company that operates at enterprise scale.
[00:20:01] - [Speaker 0]
And if we were to look further ahead, if AI agents really do become those standard operators inside businesses and everything's pointing towards that this year, how do you see the the role of the founders and the leaders? How do you see that changing over the next few years? It feels like there's there could be a lot of big changes there too, presumably.
[00:20:21] - [Speaker 1]
Yeah. So first of all, one major shift that is coming for everyone is that when you're working in an AI native environment, every person becomes a manager. What does that mean is that you have a lot of parallel work that is going because you're working with all these AI agents. You need to specify and review, specify and review. Instead of doing the work, you're sending these AI agents to perform the work on your behalf.
[00:20:50] - [Speaker 1]
So you're always multitasking and have this parallel work between specifying a work of an agent, sending them to do something, and then going to another one and reviewing their work, etcetera. So what happens, first of all, is that everyone becomes, can become a manager in that sense and need to start mastering managerial skills. What I think happens to leaders in this world is that an organization becomes more flat, and each person is in charge of their own kinda, like, department of AI agents that they can run for themselves. And you have less hierarchy in these types of businesses. And so changes are not driven by your ability to fire someone.
[00:21:32] - [Speaker 1]
They are driven by your ability to communicate and convey and align people around goals by giving them more autonomy and agency to perform their work however they think. So leading becomes much more challenging because you cannot use your, can I quote, unquote, corporate force, corporate power to enforce your way of thinking, you need to unleash your employees but still make sure that they're running in the right course? And so that becomes a very difficult dynamics to master.
[00:22:05] - [Speaker 0]
Wow. So many big takeaways from listening to you today. And for people listening, maybe they wanna connect with you or, find out more information about Swan and some of the announcements coming out, including some of the big work you're doing on LinkedIn as well. Where would you like to point everyone listening? Where should they go?
[00:22:23] - [Speaker 1]
Yeah. So first of all, follow me on LinkedIn. Amos Bar Joseph. I write regularly every week about, you know, how we're building Swan. Then I have my newsletter.
[00:22:33] - [Speaker 1]
It's called the big shift in Beehive. And, basically, I share the wins and losses of our journey at Swan. So it's if you want, like, a backseat into, you know, behind the say the the scenes, so you can subscribe to that newsletter. And finally, I have a digital clone. So if you you came up with some question you wanna ask me, so you can just ask autonomous.
[00:22:58] - [Speaker 1]
It's in chat GPT, in the GPT store, and you can just chat with it and ask it questions. It's trained on all the posts and playbooks and frameworks that I've built, so it should know some of the things that I know.
[00:23:11] - [Speaker 0]
Oh, that's incredibly cool. I'm gonna be having a play with that, and I'll also add links to everything that you mentioned. And just I love this idea that genuine AI transformation doesn't come from those grand directives, but empowering small teams to experiment and test right at the grassroots. That that seems so right on point right now, and I think, I love what you're doing here, reframing experimentation, digital playgrounds. Oh, we could talk about this stuff for hours, but just thank you for shining a light on this.
[00:23:43] - [Speaker 0]
And, hopefully, people will come and check you out, but thanks again.
[00:23:46] - [Speaker 1]
Thank you for having me, Neil.
[00:23:49] - [Speaker 0]
We covered so much ground there from agentic AI and autonomous workflows to rethinking leadership, scale, and what productivity really means in an AI native world. And this idea that businesses should exist to scale people rather than the other way around challenges some deeply embedded assumptions. And I think the conversation today raised some uncomfortable but important questions about how founders lead, how teams operate, and what skills really matter as AI takes on more execution. Well, if this episode made you pause and question how your own organization approaches growth, experimentation, or leverage, it's gotta be a good sign, hasn't it? Means we're doing our job here.
[00:24:39] - [Speaker 0]
So as AI agents become standard inside businesses so as AI agents become standard inside businesses, what part of your work would you choose to hand over, and what would you choose to protect at all costs? As always, let me know. Techtalksnetwork.com. You'll find a myriad of ways you can get hold of me there. But I'm out of time for today, So a big thank you for listening as always, and I'll return again tomorrow with another guest.
[00:25:11] - [Speaker 0]
Bye for now.