How this digital AI-worker company went from zero to becoming a $350 million company in 4 years
Episode 95
SaaS was revolutionary because it brought the power of software to anyone with a computer delivered via the cloud. But what happens when labor is delivered via the cloud?
We are moving from the code as capital arbitrage to the code as labor arbitrage. Software that can perform key tasks autonomously (albeit with some degree of human oversight).
I wrote about it in my previous essay on Service as a Software where agentic applications will eat into the labor market unlocking trillions of dollars in value and creating new markets altogether.
But in order to unlock this opportunity, you need context-aware models with long inference compute times that can access tools.
Right now, models can only do short tasks in a quick feedback loop - you ask them a question, and they give you an answer. But that’s extremely limiting. Most useful cognitive work humans do is longer horizon—it doesn’t just take 5 minutes, but hours, days, weeks, or months.
The key difference here is that of inference time compute aka the time it takes for the model to think and reason. Deepseek reasoning made all kinds of wave, and OpenAI and Perplexity quickly added reasoning capabilities. The longer the inference time, the better the output.
In a way, it's similar to how self-driving cars work. Self driving cars collect data from their surroundings via sensory inputs like the camera, LIDAR etc, after which, this data is processed using Machine Learning & Deep learning algorithms.
The algorithm is trained across a large dataset to recognize patterns and make predictions. This gives the vehicle the best chance to react appropriately to various driving scenarios without needing specific instructions for every possible situation encountered on the road.
AI agents also take in sensory input from the users’ screens, external databases, APIs etc. Once this data is collected, AI agents use a combination of ML, inference compute & reasoning to make decisions. They can learn from previous interactions, access tools all on their own, and execute multi-step tasks, much like the self-driving cars.
In Kahnemann terms, “fast thinking” AI models will make way for inference-based “slower thinking” models.
AI employees will essentially function like digital remote employees with situational awareness. Here's what I mean by this: I recently made an AI agent that takes in context about MarketCurve and the kind of work the business does. It is trained on case studies, knows about the tech space, and what kind of use-cases we can help with. Armed with this context, this agent sits on the MarketCurve website and answers questions that the customer may have about MarketCurve. But Shounak isn't this just a chatbot? Yes it is.
AI Assistant Demo 🚀 - Watch Video
But this chatbot exhibits agentic capabilities. I have given it access to my Telegram, email and Calendar. So if a customer asks to book a consultation call, the AI agent (who I call Nathan) can go into my calendar, pull up my meetings for the week, asks the customer for his time preference, cross-checks with my calendar to check for clashes, and if everything looks good, it goes to my calendar, books a meeting with the customer, sends both of us an email with a meeting link along with a summary of the meeting agenda.
To make things even more impressive, I can even send a voice text to Nathan and ask him to book a meeting or delete a meeting or show me a summary of the meeting I just had. I can do the same in Slack as well. All of this done on autopilot. For the price of a meal, I get my very own personal customer service agent that works 24*7. How cool is that?
Now imagine companies using multiple such AI agents across inbound leads, setting up outbound campaigns, recruiting the right people armed with all the relevant company information and context. This is exactly the promise of 11x AI - a company that is building digital workers for companies to hire.
Let's say you're running a sales team and you want to update your CRM or launch a email campaign. It would take you 4-6 weeks to get the ball rolling because a lot of the information is fragmented. The sales landscape was dominated by Salesforce after which came a huge unbundling of Salesforce creating multiple point-solutions which focused on very specific use-cases. As such, a lot of the tools at organizations are siloed across prospecting, outreach, CRM, follow ups, demo calls etc. This makes information exchange a nightmare. This complexity is where AI agents like 11x can shine.
They raised $75 million in funding from A16z, Benchmark, 20VC and other investors and are making real waves in this space. 11x now serves over 200+ companies like Otter.AI, Airwallex and Datastax. and are making $250 million in ARR. Let's dive in and read their story.
Brought to you by:
Market Curve is a premium tech storytelling studio that turns tech companies into media companies. We’ve worked with some of the top tech companies in the US and EU backed by some of the top investors, like YCombinator, 20VC and more.
We’ve helped founders grow their personal brand in the v0 stage, while writing essays that go viral on social media for companies closer to the PMF spectrum.
Here are a few things I can help you with —
Longform blogs, essays, and newsletters that drive authority.
Web & landing page copy that converts.
High-value assets like decks, e-books, guides, and case studies.
Ghostwriting for busy founders & execs who want to stand out.
Want to turn your startup into a media company?
Here's Hasan, the founder of 11x talking about 11x:
“You can think of them as virtual employees that you can hire like you can hire humans. They work around the clock, 24x7. Instead of being productivity enhancers, they’re capable of filling a role end-to-end.”
Armed with this vision, Hasan and his team did what I would have done in his shoes - he went after the freelancer market.
Legacy labor marketplaces like Upwork and Fiverr are ripe for disruption by AI workers and having a marketplace for people to hire agents would be the logical evolution of the market.
A variant of this is being executed by Dharmesh Shah of Hubspot who is now building agent.ai - a marketplace to hire AI agents who perform specific tasks. It's no surprise that Hubspot Ventures has also invested in 11x.
But when you're going into a marketplace, you run into the very familiar chicken and egg problem. Do you optimize for demand first or supply?
Historically, the answer lies in making sure you have the demand side ready before streamlining the supply side. You need a constant demand supply to support the costs of the supply side. Airbnb for example, captured demand first by getting booking orders from people and only when demand was validated, did they scale up the supply side.
When demand is picking up and is gaining momentum, the best bet you can place is to niche your supply side down to one specific use case. For Airbnb, it was allowing people to rent out their room. For 11x, it was allowing their AI workers to do one task in one specific vertical and do it really well. So they designed Alice, a virtual SDR (Sales development representative).
Hasan had the idea for 11x in Q3 of 2022. He refocused on the SDR vertical in Q1 of 2023.
Back at his first job, Hasan was responsible for moving data between spreadsheets - the kind of job that makes you pull your hair out. I've been there and let me tell you it's not pretty. This is exactly the kind of job that AI agents are perfect to do, and so Hasan got to work.
He and his team spent the next 6 months shadowing the SDRs. They would sit next to SDRs to observe their workflows. They would look for workflows that could be automated across prospecting, setting up campaigns, following up, and key milestones. This data would form the basis for the AI digital worker they were building.
Goes to show that even if you're building an AI-first company, the principles of doing things that don't scale still hold true. The highest leverage activity you can do in the early days is to talk to your customers, understand their workflows, their problems and build for them.
Hasan and his team built the first versions of the product based on this data, gathered feedback, kept iterating and improving the product.
Based on these intial conversations, he zeroed down on his ICP - SMB customers who had a self-serve motion. They learned that their best and most successful customers were sophisticated sales organisations with clear PMF and ability to benefit from scale, automation, and complex workflows/use cases.
He also discovered that the AI agent's technical capabilities needed to be robust from day one. They needed to be able to browse the web and find information autonomously even though this data was available through third parties.
Brex did something similar too when they launched their corporate card - they made the entire thing from scratch even though they could have utilized APIs to get to market fast. Sometimes its better to wait and do things right so you give yourself the best chance to set yourself up for scale.
"We like to operate at five or 10 times the speed most people think is normal.”
But Hasan didn't have it easy - people were skeptical of the idea of hiring a digital worker. So Hasan had to humanize it. So he gave her a name Alice, a face and a specific job. They also made sure Alice was constrained, verticalized, and trained on best practices. They invested a lot of time in building guardrails that govern how Alice would operate autonomously at scale.
They also innovated with their pricing and changed the pricing from the traditional usage-based pricing sales tools usually charged that of task-based.
They charged based on the different tasks that Alice completed like (1) identifying accounts, (2) researching those accounts, (3) preparing outreach across email and LinkedIn, (4) scheduling meetings when the prospects respond.
90% of their customers are charged based on this task-based model but Hasan tested outcome-based pricing for select customers.
This is the next frontier of pricing for AI-agents. A16z made this thesis a while back where the pricing would move from usage-based to outcome based. This follows the value-based pricing framework which charges based on the value provided. Historically, labor services have charged on either an hourly basis or based on outcome so this new pricing change to AI workers feels much more natural from a labor perspective but completely alien from a softwate pricing angle.
11x raised a $2M round in August of 2023 - around the same time they publicly launched 11x on Techcrunch. But the pricing was a hit as by the end of 2023, 11x had reached $700k in ARR on the back of Hasan's founder-led-sales motion.
At the early stage, founder led sales especially in b2b are an underrated distribution channel. Retool reached $1 million in 9 months by doing founder led sales and Amplitude reached $500k in ARR by doing this too. By early 2024, they crossed $2 million ARR with one primary growth channel - Alice herself.
Scratch your own itch is a powerful framework to validate startup ideas to pursue. Typeform started this way. So did Mailchimp. But 11x took it a step further. It used its own product not only to validate but also to grow their own business. They were their own case study.
The team trained Alice to automate outbound prospecting across emails and LinkedIn messages. Alice monitored hiring alerts for anyone hiring SDRs around the world. Alice would then send out a message saying I’m AI and I do this job.
The messages were highly personalized based on hiring event, fundraising events or how long the job postings have been live for.
This approach created a growth flywheel. Relying on Alice for growth forced them to be incredibly product focused, as incentives were aligned. So as Alice got smarter, 11x grew faster.
Use Alice to generate leads and meetings for 11x.
This usage provides with real-world data and insights to improve Alice.
As Alice improves, she generates better results for 11x and customers.
Better results lead to faster growth and more customers.
More customers mean more data and use cases to further refine Alice.
The team realised that one of the biggest challenges of automation is maintaining creativity – a crucial element in sales. So they started baking creative playbooks into Alice's repertoire. Their most successful experiment? The meme strategy.
They tested multiple follow-up formats: case studies, additional research, confirming interest, and offering freebies. These worked well, but they wanted to push the envelope. So, they tasked Alice with being more creative than traditional SDRs, leading to experiments with images and GIFs in follow-up emails.
"The results were staggering. Not only did this strategy deliver higher response rates in our outbound sequences, but it also led to a 35% response rate when used to re-engage cold pipeline prospects".
"Reaching out with intent with the right triggers at the right time is extremely effective,” Hasan said.
Alice generated 70-80 qualified meetings per week for 11x. Hasan did most of the sales call himself in true "doing things that don't scale" mode, until 11x crossed $700,000 in ARR. 11x had finally found some degree of PMF.
PMF is a vague term and it exists on a spectrum but 11x could feel the pull of the market at that stage. For B2B SaaS companies, getting to that first $1 million in ARR in 1-2 years is a telltale sign that it's on the way to being a big bet.
But PMF in the AI world is a different beast. Retention is a differentiator with switching costs next to zero.
Amjad Massad of Replit said that companies like Bolt or Lovable seem like breakthrough companies because of their insane ARR in a short time but he says its not so simple since users can switch to competitors in 5 mins.
Retention is how companies will win in this AI-first age so 11x needs to optimize for retention if it's to maintain its growth sustainably.
For 11x, retention meant doing the basics right: focusing more closely on who their ideal customer is, continuously improving the product and then really focusing on the service side of SaaS.
Going back to Alice, Hasan discovered that the key to get success in oubound was to figure out a way to leverage personalization at scale.
"You need to personalize and have relevance in a scalable way, which is becoming increasingly easy. Instead of having someone spend 20 minutes researching the person and the content they wrote, AI can do that very effectively,” Hasan says.
11x also saw more than 1,000 inbound demo requests after they publically launched via a TechCrunch article in August of 2023. The positioning of the product as hiring digital workers was an appealing one. The benefits itself were hard to say no to.
11x was 5x cheaper compared to hiring an SDR
It performed 2x better than the average SDR
Faster to deploy with less overhead and more scalability
They are now doing around $25M in ARR, a 150% increase in ARR.
You’re going from automating small tasks to automating whole workflows. The value is much higher and the willingness-to-pay for solving a holistic problem is higher - Hasan
This is an interesting point because when AI-first companies are facing resistance to automate complete workflows, which makes sense. The solution then here, if you're building in this space, is to create a wedge and automate one specific JTBD.
Clay did it when they were starting off too and automated just one specific task. This makes adoption easier and reduces signup friction, which can kill momentum, plus it reduces feature bloat right at the start.
But now that they are more mature (they raised a Series B recently), they are spreading their wings just that bit more, and extending their capabilities to voice, which seems like an obvious next step. They acquired voice AI company Opkit to help them build out Alice 2.0 and render her with voice capabilities.
Opkit started as a voice AI application that launched a healthcare-focused AI phone solution. Since then, both horizontal and vertical voice agents like Vapi, Hana, Sanas have entered the market. It's early days but we are beginning to see 11x pivoting itself into a compound product integrating multi-modal functionalities and targeting both horizontal and vertical markets, much like Rippling did.
3 weeks back, 11x launched Julian, an AI Inbound Sales Rep equipped with voice capabilities calling, qualifying, and routing leads at record speed, 24/7, on autopilot. To do this, 11x partnered with Cartesia to give their AI digital workers reps the speed, reliability, and natural expressiveness required to engage customers at scale.
And now in true compound product style, 11x vision is to launch a variety of adjacent Digital Workers before the end of the year, enabling autonomous GTM functions - across channels, modalities, and use cases. Their end goal? Create the AI GTM team.
Here’s the playbook to win in this market:
LLMs are commoditized. The real game Is in the application layer.
Here’s the reality: Large Language Models are no longer a moat. Anyone can access GPT-4, Claude, or open-source alternatives. So where’s the edge? It’s in the application layer. And for 11x, that means one thing: they’re in a knife fight. They’re competing in a crowded market—AISDR, Artisan AI.
Everyone’s racing to build the same thing: a team of autonomous workers powered by LLMs.
The question isn’t who builds it first. It’s who builds it best.
In this space, the company that wins isn’t the one with the best prompt.
It’s the one that:
Delivers real outcomes.
Executes tasks autonomously.
Optimizes for delightful user experience.
Handles “human in the loop” moments without friction.
Integrates seamlessly with tools people already use.
Clear & frendly UX for the ICP
Text-based agents are step one. But we all know what’s coming next:
Image.
Video.
Voice.
Code.
API.
An AI agent that can see, hear, speak, click, and act? That’s the bar now.
If you’re trying to win this game now, or especially if you're entering late, focus is your best friend.
Pick a vertical:
Legal.
Healthcare.
Real estate.
Finance.
Then wedge your way in. Own it. Go deeper than anyone else. Horizontal AI agents look good in pitch decks. But vertical AI agents win customers. Even competitors like AI Artisan or AI SDR could dominate their category by focusing deeply on a niche. That’s how you own mindshare—and eventually, market share.
Build multi-modal agents.
Deliver outputs that matter—not just actions, but outcomes.
Set up workflows where agents can integrate with tools and with humans.
Focus UX on trust, transparency, and speed.
Create reliable integrations that don’t break under pressure.
At the end of the day, users don’t care how smart your agent is. They care if it gets the job done.
Winning the product game is step one. Step two? Distribution.
1. Blend AI with Human Outbound
AI agents might handle the grunt work, but outbound still needs humans in the loop.
Think:
AI drafts the message.
Human tweaks, validates, and sends.
Together, they scale personalized outreach with precision.
2. Build a Content-First Brand
Content isn’t fluff. It’s strategic distribution.
Publish founder thoughts and product updates.
Share behind-the-scenes on how the agents are evolving.
Break down industry shifts and where things are headed.
Showcase customer case studies that highlight real outcomes.
Podcast episodes
This builds credibility, thought leadership, and differentiates you.
3. Use Winning Messages to Fuel Ads
Use winning messages to run SMB-targeted ads at scale.
Arcads.
AI voice actors.
UGC-style videos.
Show, don’t tell—how real companies used your tool to generate real results.
4. SEO, AI, and LLM Optimization
Old-school SEO still works. But now, you also need to rank in LLMs.
Create:
Bottom-of-funnel content that answers “best AI SDR tools.”
Pages that rank in Google and show up in AI-powered search.
If someone asks ChatGPT or Perplexity, “what’s the best AI SDR platform?” — your tool needs to be the answer.
5. Go Native on LinkedIn & Reddit
Post where the tech crowd lives.
Share wins, experiments, and use cases on LinkedIn.
Go deep on Reddit—AMA-style threads, helpful comments, early access invites.
LLMs pick up signal from these forums. Treat them like SEO v2.
Labor is moving to the cloud and is here to stay - in the short term, I suspect agents will interact with humans to create polished outputs before becoming autonomous.
It will free up time for people to pursue their passions and expand the creator economy.
People can now quickly start side projects that weren't possible before. The creator economy, previously confined to music, content, and videos, will now embrace tech.
The pie is clearly expanding. Building is being democratized.
Distribution, speed, and a community-first approach are the only advantages.
But success in this space won’t come easy.
Whether you’re targeting verticals, in-house roles, or outsourced services, the playbook remains the same: start small, build trust, and scale with precision.
If you liked this essay, consider sharing it on your socials or inviting a friend to read this essay. Feel free to say hi to me on Twitter or on LinkedIn or email.
That’s all for today! See you soon!
Amor Fati Amor.