800万美元ARR产品一夜关停后，他又打造出年入200万美元ARR的电商AI工具

Killing an $8M ARR business

I'm Ruslan, Bulgarian, based in Sofia. I started as a UI/UX designer and taught myself to code because I couldn't afford to hire anyone. In 2017, I built Checkout X, an alternative checkout for Shopify stores. I bootstrapped it to $8M ARR. Then, Shopify changed its platform, locked down the checkout, and the product died. Years of work vanished because someone else changed their mind. I then built and micro-exited another Shopify app, Vanga AI. And in 2024, I cofounded Zipchat. Zipchat is an AI sales agent for e-commerce stores. It handles any conversation between a brand and a customer — on any channel. We combine all your conversations into one place, and ONE agent handles them, an agent you train only ONCE. We currently handle 700k conversations a month, growing about 10% week over week, and we are about to cross $2M ARR.

Experiments over epiphanies

This is my fourth business, so it wasn't some epiphany that started it. I wanted to start something new, I had experience in ecom, ChatGPT had become all the rage, and I thought: "Let's build a ChatGPT wrapper, learn about AI, and hopefully make some money." The building process was straightforward. We built the most minimal product version, launched it, started running ads on the Shopify App Store, gathered feedback, and gained momentum. We run on Ruby on Rails / Hotwire / Postgres, and we leverage RubyLLM for our AI implementation. Regarding AI, we started with OpenAI, then moved to Anthropic, and now we're working with all the major labs. The stack didn't change that much, but our development process did. Nobody writes any code anymore, so we're changing our development philosophy to reflect this change. We no longer do peer reviews, and we don't enforce readability/maintainability as we did before. And instead of having a Head of Engineering who supervises our developers, each developer now acts as the "Head of Engineering" for their own agents.

Bootstrapping at a loss

At the time, I didn't realize "building a wrapper" would be much harder than it sounds. Nor did I realize that AI would completely disrupt the shopping experience. Financially, I was fortunate that I didn't have to worry about money for a while, so I could focus on building something interesting. But our costs were significant. My partners required small salaries, we paid for GPT-4 (which was super expensive at the time), and we ran ads. The result was that we ran at a loss of about $20k per month for almost a year. Thankfully, we ground through and began breaking even. Eventually, we secured revenue-based financing (loans based on your MRR), which helped us significantly. We stayed super lean until we hit 50k MRR. I built the product, Carlo did sales and support, and Luca ran ads. We were reluctant to hire ANYONE until we could break even and afford it.

Usage-based plans

As far as our business model, we sell plans based on AI replies. The more usage a merchant has, the more our revenue grows. If they go over their limit, we also charge them additional usage fees. To expand customer revenue, we need merchants to grow their usage. So, we employ a land-and-expand model. Merchants start using us for their website chat and, over time, they add us to their search, product page, social media, emails, and so on. We haven't iterated much on our business model until now — we're starting to run A/B tests to see what works best.

There are no growth hacks

As far as growth, there's no story here. No growth hack, no viral moment, no single channel that cracked it open. It's a slow grind, and we keep grinding. Here are our marketing assets — long-term plays where every action compounds over time: YouTube- We don't push the brand; we just make stuff that's genuinely interesting for ecom folks. Social media- All the founders are building personal brands on X and LinkedIn. SEO / GEO- We optimize specifically to get cited by LLMs. App partnerships- We're building an ecosystem of Zipchat "skills" inside other apps. Shopify app store- We're working with an ASO consultant and pushing for reviews. Cold outreach via emails, calls, and conferences has been good for awareness. We buy data from Store Leads, Similarweb, etc., and send 1M+ emails per month. If it might work, we test it. Most of it doesn't. Some of it does, and we double down on that until it stops working; then we go find the next thing. The only real "strategy" is to never stop shipping and selling. Work on your product and talk to your customers.

Too much competition

The biggest challenge we still face is having the least creative business idea of our time! Over 100 new competitors have launched in the Shopify App Store alone, probably thousands worldwide. Even if we're better than everyone else, each competitor only needs to acquire two customers, and that will still compound to hurt us. That being said, I picked this business because I wanted to learn, and it's a great space to be in. All of e-commerce is experiencing disruption, and we get to be among those figuring out how it should happen.

Cofounders for the win

My greatest advantage is having great cofounders who are as obsessed with winning as I am. In all of my previous businesses, I was a solo founder. I had to figure out everything by myself. I didn't want to do that again, so I found people who are 1000x better than me in their fields. Besides that, my advantage has been talking to our customers and solving their problems. Ignore startup advice, books, and the competition — all that matters is your team and your customers.

Pick an idea space and get started

Here's my advice: Pick an "idea-space," not an idea — do something you are interested in and want to learn about. Even if it fails, you'll get inspiration for your next attempt. Don't think — just do! Launch your product, try to sell it to someone, and repeat. If you can't sell it, why not? If you sold it to someone, great! How do you sell it to ten people? 100? It is that simple. And the best way to learn is to do. All it takes is a bit of luck. My previous business was more successful than this one, even though my execution was much "worse." You might get lucky on the first try, even though you're not experienced. Or you might not get lucky for a long time. But the only way to get "lucky" is to keep playing.

What's next?

I want to build something that makes a "real impact". Something people across the world use at scale. Something that makes the world tick. I don't have clear criteria for what this means, but I guess a billion AI conversations would suffice. You can follow along on X and LinkedIn. And check out Zipchat!