We Had $2M ARR, Yet We Had to Jump the Tiger.

A few weeks ago, I sat down with Omer Khan on the SaaS Club Podcast. We talked for about an hour - about how our team built Product Fruits to 1,300+ customers competing against companies that had raised hundreds of millions, about the moment AI nearly made everything we'd built feel obsolete, and about the decision that probably looked suicidal from the outside but turned out to be the best thing we ever did.

You can watch the full conversation here. But if you'd rather read, below is the short but honest version.

You can find the SaaS Club Podcast on Spotify or Apple Podcasts as well.

I wear a T-shirt that says "Riding the Tiger"

People always ask what it means.

In Asia, there's a saying: You can ride a horse. A horse is slow, and sometimes may even be dying, and you don't see it. I think that's where Product Fruits was two or three years ago. We were riding the horse. Everything seemed fine. Competitors here and there, everybody was fine, you know.

And then you start to notice - there are new guys behind you, on motorbikes, AI-powered, and they will take everything from you.

So you have to get off the horse.

When you're riding a tiger, the tiger will eat and destroy anything in front of you. But you can't stop. If you stop, the tiger will eat you. The only way is to aggressively continue.

That's where we are now. This is the story of how we got here.

We grew to $2M ARR with one guy spending one hour a day on marketing

I want to tell you about Dan.

When we hired Dan, we called him our Chief Marketing Officer, which was, honestly, a complete lie. He was a plumber. He fixed everything - leaks here and there, without even telling us. "Oh, there was a problem - I fixed it." He wore about twelve different hats in a four-person company, and marketing got maybe 20% of his time. One hour a day, max.

So we did PPC. Mostly Google Ads. Not because it was fashionable – in fact, at that time, everybody was obsessed with content. We heard it constantly: "We'll help you with content, but it'll take nine months to see results." And they were right about the nine months part. They just always seemed to find another client before the nine months were up.

We needed to know if we could compete in the US market fast. PPC told us quickly. Our early free-trial conversion was 24–25%, and with an average ticket of $100, we recovered our acquisition costs in about eight or nine months. That's good enough. We scaled it to over $1M a year in spend.

Eventually, we hit a ceiling - there are only so many people searching for a digital adoption platform at any given moment. But it worked long enough to get us to around $2M ARR, which is more than most people said we'd reach, competing against Pendo and WalkMe with $3.5M in total funding.

We had an Israeli company tell us recently: "There must be something you're not telling us. It's amazing. Why is it so cheap? There has to be a hook." There's no hook. We just started as a company for SMBs, priced accordingly, and apparently, some people find that suspicious.

PLG is great - until the product gets too good

We used Product Fruits to sell Product Fruits. That still makes me smile. A digital adoption platform that uses its own adoption features to onboard its own customers. It made every kind of sense.

And it worked, until it didn't.

The problem is that PLG works best for simple products. Ours stopped being simple. Five years ago, we had product tours and hints. Now our AI can run a discovery call with your new users, figure out what they're trying to achieve, tailor the onboarding to them, handle support questions, nudge them toward upsells, run churn surveys, and push product announcements - and I can ask it every morning what's preventing better adoption of a specific feature, and it will tell me.

You can't self-serve your way into buying that. You need a human to sit with you and show you what's possible. Almost every time our account executives do that, the customer says: "Wait - I didn't know any of this was possible."

So we built a sales team. Five people, including implementation engineers. Small, but focused on bigger contracts. And that's what carried us from $2M toward where we are now.

Two weeks of dark times

I want to be honest about what it felt like when AI competitors started showing up.

We were scared. Really, genuinely scared. We'd done everything right - good product, smart growth, real customers - and suddenly the world was changing faster than we could react. We weren't an AI platform, and that felt like a death sentence.

I remember every single day of those two weeks.

And then Láďa, our CEO, and I had a conversation that changed everything. We stopped asking "how do we defend what we have" and started asking: “what did we always want to build, but couldn't because the technology wasn't ready?”

The answer was obvious. Not onboarding tours. Not checklists. An invisible buddy. Something sitting next to you while you use software - not giving you a tutorial, but actually understanding what you're trying to do and helping you do it. If you're using Photoshop and ask, "How do I paint clouds?", it doesn't just point to a tool. It says: "For clouds, you could try this, or that - or should I just try painting the clouds for you?"

That's what we wanted to build. And for the first time, the technology made it possible.

So I wrote an email to our investors - Lighthouse, Reflex Capital, Leverage - and I told them: we're stopping work on Product Fruits. We're rebuilding the platform from scratch. I expect you will see a decline in MRR.

I hit send and said to myself: "Okay. Now let's wait for the shit storm."

Twenty minutes later, Ondřej from Reflex Capital called me. He said: "How much money do you need? We're betting on you. We want winners, not survivors."

That was not the reaction I expected.

What we actually built (and what we decided not to build)

One of the first AI ideas we got excited about was churn prevention analytics. Everybody was excited about it. We tried it. It didn't work. So we went back to what we know: communication between software and the people using it.

That's our domain. That's what we've been doing for years. AI made us dramatically better at it, but the foundation was already there.

Our AI assistant - we call him Elvin - now handles discovery calls inside the product, answers support questions, runs surveys, pushes contextual upsells, and even tells me when my UI decisions are questionable. I once asked him to make something red, and he told me: "Karel, I don't recommend it. It's a scary color." We went with green.

Today, 80% of our customer support tickets are resolved without any human intervention. That's not a marketing number - that's what we actually see.

One honest problem, I don't have a perfect answer to yet

Pricing.

We pay per conversation to LLM providers. Our costs are genuinely usage-based. But our customers - especially the bigger ones - need a flat monthly number they can take to procurement. Nobody can get a purchase approved for "somewhere between $50 and $50,000, depending on usage." That's not how companies buy software.

We're working through this. Right now, we give generous free usage during trials - I call it "zillions" of free interactions - because we know that if someone actually uses Elvin, they'll become a customer. We've validated this. But the right pricing model at scale is still a question we're answering in real time.

If you've solved this problem in your own AI product, I'd genuinely like to hear from you.

What I've learned from building this

We are lucky. I know that. We happen to operate in exactly the space where AI makes the most sense - communication between software and users. Not every company gets that alignment.

But I also think there's a principle here that applies beyond our situation: you need a real foundation before AI can help you. Experience talking to thousands of customers, years of understanding why users get stuck, genuine domain expertise - those things aren't replicable quickly. A new AI-native startup can't shortcut that.

The companies that are going to win with AI aren't the ones that added it fastest. They're the ones who added it to something real.

We're Czech people. We're not easily impressed - by ourselves or by others. That skepticism has served us well. Whenever we considered an AI feature, the question was always: why does this need to exist? What real problem does it solve? If we couldn't answer that clearly, we didn't build it.

It's not a complicated filter. But most companies aren't using it.

Want to see what this looks like in practice?

If you're building or running a software product and you want to see what modern user onboarding and adoption actually looks like - not the horse, the tiger - come take a look.

👉 Start a free trial - no credit card required. Elvin will help you get started. He's good at that.

👉 Or book a demo if you'd rather talk to a human first. We have those too.

I spoke with Omer Khan on the SaaS Club Podcast. If you prefer listening to reading, the full YouTube episode is at the top of this page. You can also find it on Spotify or Apple Podcasts.

And you can reach me on LinkedIn.

‍