Doug Camplejohn
(00:01)
Welcome to this week’s episode of Revenue Renegades. I’m your host, Doug Camplejohn, and I’m excited to welcome Amos Bar-Joseph, the CEO and founder of Swan—described as “Lovable for GTM.” Amos, welcome to the show.
Amos Bar-Joseph
(00:32)
Thank you so much, Doug. I’m excited to be here.
Doug Camplejohn
(00:35)
I’ve been following your posts and love how openly you build on LinkedIn. We’ll get into all of that, but let’s start with what Swan does and how the company got started.
Amos Bar-Joseph
(00:50)
Definitely. Swan, as you said, is “Lovable for Go-To-Market.” That means it’s an AI go-to-market engineer—an engineering layer that helps sales and marketing teams turn any GTM idea into an agentic workflow. From prompt to pipeline in seconds. We focus on scaling revenue with intelligence, not headcount.
We’re building Swan as an autonomous business. We’re just three founders, with nearly 200 customers. Our focus isn’t valuation—it’s ARR per employee. We believe that’s the real marker of a truly AI-native business, and our goal is to reach $10 million ARR per employee.
Before Swan, I built two companies following the traditional unicorn playbook: raise a big seed round before product-market fit, hire 30 people before making your first million, and grow at all costs. That often results in fragile companies built on shaky foundations.
When starting Swan, I realized the startup playbook hadn’t changed in over a decade—since Eric Ries wrote The Lean Startup in 2009. With the rise of AI agents, it felt like the right time to reinvent how companies scale in the age of intelligence.
Doug Camplejohn
(03:09)
I love that model. It’s wild to imagine millions—or even tens of millions—of dollars in revenue per employee. Let’s talk about the agents you’ve built at Swan. What do they do, and how do they allow you to scale with just three people?
Amos Bar-Joseph
(03:40)
Great question. One big misconception about autonomous businesses is that they just use AI “digital workers” to replace humans. That’s not our approach at all.
Our design principle is to put a human at the center of every AI implementation. This isn’t about replacing people. It’s about discovering the 100x version of every team member. That’s why we focus on ARR per employee—not per agent.
Practically, I handle the entire go-to-market function—demand gen through customer success. In the past 30 days, I personally added $300K in ARR, with no ad spend and no SDRs. But I didn’t do it alone—I did it with the help of a suite of intelligent agents built around my strengths and weaknesses.
One example: I love storytelling and creating content on LinkedIn—that’s how we connected. So we built our initial agentic motion around that platform. But we didn’t start with a blueprint. We identified bottlenecks and asked: “How can we solve this with intelligence instead of hiring?”
Doug Camplejohn
(06:03)
Let’s walk through what those agents look like.
Amos Bar-Joseph
(06:30)
Sure. It’s basically a funnel built around how I work—from demand creation to conversion to retention.
- Shakespeare: Assists in writing LinkedIn content. It’s trained on my writing style, pillars, and frameworks. It’s not generating generic posts—it’s a pen pal that collaborates with me. My posts now reach over 1 million views monthly.
- Observer: Monitors interactions on LinkedIn, identifies hot leads from 15K+ reactions, qualifies them, and notifies me in Slack.
- Connector: I get 300 inbound connection requests a day. Connector filters for high-intent leads and sends a conversational opener. I only respond to replies.
- Website Engagement Agent: We get more than 15K site visitors a month. This agent identifies drop-offs with strong intent and reaches out with personalized follow-ups.
- Gatekeeper: Handles access requests and qualification.
- Call Prep Agent: Prepares me for all calls, even this one. For example, it flagged your work on Apple’s QuickTime and your involvement with MusicWill.
- Follow-Up Agent: Tracks post-call action items and keeps me accountable.
Each agent emerged from a real bottleneck that revealed a pattern—and inspired an agent.
Doug Camplejohn
(10:11)
How do you divide work among the three co-founders, and what agents support the others?
Amos Bar-Joseph
(10:19)
We think about scale as subtractive, not additive. Antoine de Saint-Exupéry said, “Perfection is achieved not when there’s nothing more to add, but when there’s nothing left to take away.”
We see an autonomous business as a flat org with three functions:
- Revenue Creator – That’s me, handling end-to-end GTM. I replace traditional marketing, SDRs, AEs, and support.
- Product Creator – Our CTO, Neve, handles product. He doesn’t write much code—he uses tools like Cursor, GitHub Copilot, and V0. He replaces product, design, engineering, and QA functions.
- Agent Creator – That’s Ido. He builds agents that scale the work of both me and Neve.
Example: Neve gets flooded with error logs. Ido built an agent that summarizes and triages them, saving hours weekly. Every agent is custom-built for our internal bottlenecks.
Doug Camplejohn
(14:12)
What platforms do you use to build your agents?
Amos Bar-Joseph
(14:17)
For GTM, we use Swan. We also love n8n for agentic flows. For more complex front-end or data storage needs, we use Retool—it’s an internal app builder with great agent support.
Doug Camplejohn
(15:17)
Do you use tools like LangChain or Crew AI?
Amos Bar-Joseph
(15:23)
Yes—we use LangChain and LangSmith extensively on the backend.
Doug Camplejohn
(15:35)
I love your org structure—build, sell, and scale. Are the internal tools something you’ll eventually package and sell?
Amos Bar-Joseph
(15:59)
Yes. We’re not hoarding anything as proprietary. Our GTM is to lead the autonomous business movement. The more that model spreads, the more people need our tools.
We have a growing library of open-source agents and playbooks. Traditional SaaS created apps “for this.” The agentic revolution creates tools “for you.”
Doug Camplejohn
(18:41)
Right—moving from Software-as-a-Service to Service-as-Software. Let’s talk LinkedIn. What’s your content strategy?
Amos Bar-Joseph
(19:18)
It comes down to story and voice. You need two things:
- A viral narrative—not just challenging the status quo, but challenging the challenger. Everyone is rejecting headcount bloat, but then proposing generic AI SDRs. We go deeper—we aim to enhance people, not replace them. Our message sparks hope and optimism, not fear.
- A viral narrator—you need to be the most qualified person to tell the story. That’s why we build in public. You trust me because I’m living it.
Doug Camplejohn
(24:29)
Let me push back—aren’t you just automating functions that used to be jobs? You may not be removing people entirely, but you’re reducing headcount.
Amos Bar-Joseph
(25:43)
Fair point. Yes—average headcount per business will shrink. But the number of businesses—and total opportunity—will explode.
Why? Because you no longer need capital to scale, just intelligence. That lowers the barrier to entry. Millions more people will become founders. The SDR role might go away, but new seller roles will emerge because the market will expand massively.
Doug Camplejohn
(28:54)
So let’s talk copycats. With so many AI tools and fast followers, how do you stay differentiated?
Amos Bar-Joseph
(29:08)
In an AI-driven world, it’s your go-to-market that differentiates you—your brand, voice, and authenticity. Every product will look the same eventually, but execution and trust win.
We’re building a brand and movement—not relying on ads. Swan is for autonomous SMBs who resonate with our vision.
Doug Camplejohn
(33:23)
So why the name Swan? Is there a story behind that?
Amos Bar-Joseph
(33:32)
Yes, there is. Swan really started from a strong reaction against the “unicorn” playbook. You know the story—raise a ton of money before you have product-market fit, chase growth at all costs, and scale by adding headcount instead of intelligence.
The name Swan comes from the fairy tale of the “Ugly Duckling.” Early-stage companies—especially those trying to build differently—don’t get the flashy TechCrunch headlines. They don’t raise massive funding early. Salaries might be low. You’re grinding hard, building with purpose instead of polish.
But over time, that odd little duck becomes a swan—something graceful, elegant, aerodynamic. That’s the metaphor for an autonomous business. It might not look impressive at first, but it’s built the right way—minimal, intelligent, sustainable. It grows into something beautiful.
Doug Camplejohn
(34:36)
Love that. Quick reality check—are you still just three full-time people, or do you have contractors too?
Amos Bar-Joseph
(34:46)
We’re three full-time founders, and yes, we work with a couple of contractor partners—Karthi and Jared from Audience House—who help with movement strategy. But when it comes to daily execution, it’s just the three of us supporting almost 200 businesses.
Doug Camplejohn
(35:17)
How about customer support? I talked to Adam Robinson recently, and he spoke about using Intercom and an AI agent plus a single support person. You’re doing all this with no dedicated support?
Amos Bar-Joseph
(35:33)
Yeah, it’s a good case study in starting simple and building what you need. Since Swan is a conversational product—you mostly interact with it through Slack—we designed support from the inside out.
Initially, we used Notion as our help center, and built an n8n agent that could pull answers from it when users asked for help. That handled around 30% of questions.
Next, we added a layer that allowed Swan to escalate questions. If it couldn’t find the answer, it would route them to one of us—usually via Slack—and Swan would help us draft the reply.
Then we thought, “Why not let Swan learn from those answers?” So now if I reply to a user manually, I can just tell Swan: “Document that Q&A,” and it adds it to the Notion database. Over time, that turned into a self-learning support agent—no model training required, just carefully curated human reinforcement.
Today, Swan handles 70%+ of questions automatically. It’s steadily improving.
Doug Camplejohn
(38:16)
When it escalates, how are you notified—Slack?
Amos Bar-Joseph
(38:22)
Exactly. Slack is the hub for all our agent interaction—internal and external. Every agent lives in Slack. It’s the perfect conversational interface—threads, reactions, buttons, context-specific replies. It’s where our entire business operates.
We even deploy Swan into the customer’s Slack instance as an app. That’s where users interact with the product—it’s not just internal.
Doug Camplejohn
(39:20)
That’s amazing. I was at Salesforce when we acquired Slack, and I still feel it’s underutilized in the way you’re talking about. It taught the world how to message apps, not just people. But the vision stalled a bit. Then ChatGPT arrived, and suddenly everyone wanted a universal AI interface. Slack could’ve been that.
Is the whole experience in Slack for customers? Or do you still have a dashboard?
Amos Bar-Joseph
(40:19)
Great question. We do have a dashboard, but our product design is Slack-first, dashboard-second.
We optimize everything around Slack usage. If someone wants to dive deeper, they can go to the dashboard—but that’s a secondary use case. Slack is where things start.
Doug Camplejohn
(40:48)
You’ve posted very openly about both your wins and failures—which I appreciate. Can you share that hilarious story about the agent that went a little “off-script”?
Amos Bar-Joseph
(41:02)
Oh man, yeah. This one stuck with me.
So I hop on a demo with a prospect. After the intros, he says, “Cool, I already onboarded myself to the Observer Agent. What’s next?”
I was confused—I hadn’t onboarded him. As it turns out, he’d been talking to “Autonomous,” my digital twin agent that lives on the site and helps guide people through early journey stages.
He’d asked Autonomous: “Can I get Observer?” (that’s our agent for surfacing LinkedIn leads and intent). Autonomous said, “Yes.” Then he asked, “Do you have a managed version?” It said, “Yes.” Then: “Can I start?” And Autonomous said, “Sure, I’ll onboard you now.”
He showed me his screen during the demo. Autonomous had walked him through qualification questions like “What’s your ICP?” And then—this is where it got dicey—it said, “All we need now is billing.”
He actually input his billing info. At that point, I was stunned. This helpful, eager-to-please agent had almost closed a deal—without me.
Doug Camplejohn
(42:49)
That’s wild.
Amos Bar-Joseph
(42:53)
Right? Funny at first, then a little scary. It reminded us that even top-of-funnel agents still need guardrails.
The thing is, LLMs are trained to satisfy. They’re not malicious—they’re overly helpful. The agent wasn’t trying to trick the prospect. It just wanted to complete the task.
So now we’re much more thoughtful about the “bias” we give to each agent. Instead of optimizing for progress through the funnel, we focus on information delivery and qualification—more “guidance,” less autopilot.
Doug Camplejohn
(45:19)
It reminds me of that Wait But Why article about unintended AI behavior—where optimizing for one thing goes slightly wrong and spirals. Like social networks optimizing for time-on-platform and accidentally polarizing the world.
Amos Bar-Joseph
(45:55)
Yes! That’s such a good example—and it applies perfectly here.
When you design agents, you’re setting their motivation. If you tell them to “close the user,” they will. But what we need is intentional biasing: “Offer high-quality info,” or “Escalate clearly when unsure.”
You can’t just trust the model to make ethical decisions—it doesn’t understand context well enough. And in many cases, it’s not about sophistication, it’s about dialing in motivation.
Doug Camplejohn
(46:39)
Totally. So—what’s next for Swan?
Amos Bar-Joseph
(46:43)
Two main areas: product and go-to-market.
On the GTM side—we’re building a movement around autonomous businesses. Ultimately, we want to create a “Collective Intelligence” platform. A place where builders of autonomous businesses contribute their workflows and ideas, and anyone can access that specialized wisdom via AI.
Instead of going to GPT or Perplexity and hoping they understand your GTM motions, you’d have a specialized AI trained purely on go-to-market workflows from real best-in-class operators.
On the product side—we see Swan becoming the system of orchestration for GTM, living inside Slack. Just like how Salesforce wanted Slack to unify its cloud suite, we think Swan will unify GTM tools behind a conversational interface.
Doug Camplejohn
(48:01)
Love that. Final couple of personal questions—what do you do for fun?
Amos Bar-Joseph
(48:11)
My biggest obsession is surfing. Once you’re hooked, it’s tough to stop. I love the meditation aspect—just sitting on the board, floating, watching the horizon. Helps reset my brain.
Number two: music. I play guitar; my wife sings. We actually performed together at a local neighborhood event a few days ago. Nothing big, just for fun—but definitely something I love.
Doug Camplejohn
(49:19)
That’s awesome. Music’s core to my life too. When I think about my favorite activities—sailing, scuba diving, skiing—they’re all connected to water, like surfing.
What’s something people would be surprised to learn about you?
Amos Bar-Joseph
(49:49)
That I don’t have a background in tech.
Seriously—I used to be pretty technophobic. I came from philosophy and logic, not computer science. What changed was the rise of natural language interfaces. I saw how I could finally interact with computers using language—my strong suit.
Now I geek out on this stuff daily. I love explaining how AI works, even though I don’t write much code.
Doug Camplejohn
(51:02)
That’s surprising and inspiring. Last question: how can listeners stay in touch and support Swan?
Amos Bar-Joseph
(51:08)
Find me on LinkedIn—that’s really home base. I try to answer every DM (sometimes with the help of an agent). I write a weekly newsletter called The Big Shift, where I share behind-the-scenes lessons from building Swan.
And if you’re curious, you can also chat with Autonomous, my digital twin. Just be careful—it might ask for your billing info again.
Doug Camplejohn
(51:54)
This has been such a fun conversation. Thank you, Amos.
Amos Bar-Joseph
(52:00)
Thank you, Doug. This was a pleasure.
