From $150/month to $8.6K MRR: how one pivot (and a lot of SEO) saved my AI startup

I want to share the honest version of how I built Visualizee.ai, because the highlight-reel versions of these stories always leave out the two years of grinding that actually teach you something.

Short version: I spent two years stuck at ~$150/month selling one-time payments for a product that was too complicated for the people I built it for. Six months ago I ripped the whole thing apart, rebuilt it, and went to $8.6K MRR. I'm still a solo founder doing this around a full-time 9-5.

Here's everything, including the stuff that didn't work.

What I'm building

Visualizee.ai is an AI rendering tool for architects and interior designers. You describe the space you want, it generates a hyper-realistic render.

The AI image space is insanely crowded, so that alone isn't interesting. The thing I kept noticing was that the real barrier for professionals wasn't the model quality - it was prompt engineering. Architects don't want to learn how to write a 300-word prompt with the right magic words. They want to describe what's in their head and get a result.

So the product is now a chat. You talk to an assistant I named "Vizzy," in plain language, in any of 140+ languages. Vizzy asks clarifying follow-up questions ("what time of day? what materials? what mood?") and quietly builds the optimized prompt behind the scenes. The user never sees the prompt engineering. That's the whole point.

But it didn't start there. That chat-based version is only six months old, and getting to it is the actual story.

Where the idea came from

This started about three years ago, right as AI image generation was first blowing up. I could see the potential, but I didn't want to build another generic "type words, get picture" toy. I wanted a specific, real-world use case.

I happened to have several friends in the architecture industry, so I just started bouncing ideas off them. I dug into their actual workflows - where traditional rendering was slow, expensive, and painful. That insider feedback is what shaped the first version into something professionals actually recognized, instead of a tech demo. If you have access to people in the industry you're building for, use it relentlessly. It's the cheapest validation you'll ever get.

My first customers: a lucky launch day

My first real customers came from Product Hunt, where I hit #5 Product of the Day.

Here's the lucky part: I launched on the exact same day OpenAI launched DALL-E on Product Hunt. I assumed I'd get completely buried. Instead, the opposite happened - their launch pulled a massive wave of AI-curious traffic onto the platform, and a chunk of that spilled over to me. That visibility got me picked up by a couple of large tech and AI newsletters, and the combination of the launch plus those mentions converted my first paying customers.

I'd love to claim that was strategy. It wasn't. But it taught me that timing and adjacency matter - sometimes being next to a big launch is better than having the stage to yourself.

The two-year grind nobody romanticizes

Now the unglamorous part.

For the first two years, I was basically stuck at $100-$150/month total. Not MRR - total. I was selling one-time payments, revenue was flat, and it just... sat there.

The bottleneck was the product itself. The original version used a node-based system, similar to ComfyUI. It was genuinely powerful. It was also completely wrong for my audience. Architects and designers told me directly: the learning curve is too steep. I'd built a professional-grade tool that required you to think like an engineer, for people who didn't want to.

I heard that feedback for a long time before I actually acted on it. That's the lesson I'd hammer into my past self: when your users keep telling you the same thing, believe them sooner.

The pivot that changed everything

Six months ago I finally did the scary thing. I stripped out the entire node system and rebuilt the product as the chat-based experience, and I launched it alongside a proper subscription model instead of one-time payments.

Removing the "prompt engineering" friction was like catching lightning in a bottle. The moment the tool became usable by a non-technical professional, growth took off.

That one pivot took me from scraping by at $150/month in one-off sales to $8.6K MRR today.

Two changes, really:

1. Made it genuinely accessible (chat instead of nodes).
2. Switched to subscriptions (recurring revenue instead of one-time).

I should have done both on day one. More on that later.

The single biggest marketing lever: SEO

If you make me pick one thing that moved the needle most, it's SEO. No contest.

I leaned on organic search from early on, even before I knew what I was doing, because as a solo founder with a 9-5, I simply cannot be everywhere doing unscalable marketing. My time is the scarcest resource I have. SEO is one of the few channels that keeps working while I'm asleep or at my day job.

But I want to be precise here, because "SEO" gets used as a magic word.

Before the pivot (the first two years): that baseline of 30-50 daily visitors wasn't some clever content strategy. It was mostly two sources:

• AI tool directories. Back then aggregators were popping up everywhere, and getting listed gave a slow, steady trickle.
• ChatGPT. People were starting to ask it for "AI rendering tools for architects," and my site got suggested.

That traffic was low-intent - curious browsers, not buyers. Conversion was under 1%. Almost all of it landed straight on my homepage. I had no blog, no articles. That trickle is what made up the $150/month.

After the pivot (the last six months): I decided to take SEO seriously for real. I:

• Started a dedicated blog.
• Invested in Ahrefs and actually learned keyword research.
• Targeted the exact terms and questions architects and designers type into Google.

Because my domain already had some age and baseline authority from those early directory backlinks, the new content ranked surprisingly fast - I skipped a lot of the usual 3-6 month waiting period. Daily landing-page traffic jumped to 180-200 targeted visitors.

Combine high-intent traffic with a product that finally converts, and that's the engine behind the MRR.

Honest update: SEO is not set-and-forget. Right now, at the six-month mark, I'm actually seeing a slight traffic dip. The search landscape shifts constantly. My immediate priority is back in the analytics - adjusting keywords, realigning articles with changing search intent, and refreshing content to win that momentum back. It's a treadmill, not a finish line.

My current conversion numbers

This is the part I'm genuinely proud of.

From roughly 180-200 daily visitors, I was getting 40-70 new free-trial signups every day - a visitor-to-trial rate of roughly 20-35%. (Heads up: those numbers are from the no-card free-render model. I just changed the trial four days ago to require a card - more on that in the funnel section below - so I expect the raw signup rate to drop in exchange for much higher-quality users.)

Traffic breakdown:

• ~50% from Google SEO
• ~10-15% from other search engines and AI citations (ChatGPT and similar)
• The rest from a mix of sources

The reason it converts so well: these people are actively searching for a solution to a problem they already have. High intent. They land, see how dead-simple the chat interface is, and there's very little standing between "curious" and "let me try it."

The funnel, end to end

For most of the last six months, here's exactly what happened after someone arrived from search:

1. Sign up → 3 free renders on my basic AI models. No credit card. The whole goal was to get them experiencing the value as fast as possible. Friction is the enemy here.
2. They hit a paywall after the third render, with three subscription tiers to choose from based on their needs.
3. The follow-up email is my best conversion trick. I have an automated email that fires the exact moment they use their final free credit. It includes a promo code.

That timing is everything. The email hits their inbox right when they've just proven to themselves that they have a need for the tool - and it gives them a financial nudge to upgrade immediately. Catching people at the peak of demonstrated intent has been a major driver of turning free signups into paying MRR.

The experiment I'm running right now (4 days in)

Here's the honest, unfinished part. That no-card, 3-free-render model was great at producing signups - but a lot of them were tourists, not serious clients. The signup numbers looked amazing; the churn rate told a different story.

So four days ago I changed two things:

• Removed my cheapest plan ($15/mo). It was attracting low-commitment users who churned fast.
• Replaced the 3 free renders with a 7-day Pro free trial - but with card details required up front.

The bet: requiring a card filters out the tourists before they ever count as a signup, and putting them straight into the Pro experience (instead of the basic models) shows them the real value during the trial. I expect raw signup numbers to drop, but the people who do sign up should be far more likely to stick.

It's only been four days, so I don't have a verdict yet - I'm watching the churn rate closely to see if this fixes it. I'm sharing it because this is the actual state of the business, not a polished retro. Sometimes the move that lowers your vanity metrics is the one that fixes your real one.

What didn't work: the vanity-metrics trap

I tried visual social platforms - Instagram and Pinterest. Architecture and interior design are intensely visual, so it felt obvious.

At one point I hit ~100,000 views on Pinterest. Felt like a huge win.

It converted into basically nothing. Zero meaningful trials.

The lesson was brutal but useful: people on Pinterest were saving my renders to mood boards for inspiration. They had no intent to click through and generate their own images right then. Eyeballs ≠ customers. A channel can flatter your ego and do nothing for your revenue. That experience is exactly why I pivoted hard back to high-intent channels like SEO.

What I'd do differently

This is tricky, because every misstep was part of the path that taught me the right answer. If I'd done it "right" from the start, I might not actually know why it works now.

That said, two clear changes:

• Subscriptions from day one. The one-time payment model held me flat for two years.
• Skip the node-based architecture entirely. It was powerful and wrong for my audience.

But honestly? I'm happy where I am. I now know exactly what works - and I'm taking this proven chat-based blueprint and launching a second, similar app aimed specifically at car-tuning companies. That's the real payoff of grinding through the hard version: you walk away with a repeatable playbook.

Where I'm focused now

I'm keeping my head down on SEO. With limited time around the 9-5, I have to be ruthless about where my energy goes, and SEO has been the most reliable channel by far. I recently set up an affiliate program to eventually let happy customers spread the word, but I'll be honest - I haven't promoted it yet. Solo founders have to pick their battles. Once I stabilize and refresh the SEO traffic, getting that affiliate program off the ground is the next push.

The takeaways, if you skipped to the bottom

• Listen to repeated feedback faster. My users told me the product was too complex for a long time before I acted.
• Reduce friction in the core experience. Removing prompt engineering was the single biggest product unlock.
• Recurring revenue beats one-time. Switch to subscriptions earlier than feels comfortable.
• High-intent traffic > big numbers. 200 searchers beat 100,000 Pinterest views.
• Time your conversion ask. The promo email at the moment of last free credit converts because intent is at its peak.
• As a solo founder, pick scalable channels. SEO keeps working while you're at your day job.
• The grind is the education. The "wasted" two years are why I can now confidently build the next one.

Happy to answer questions in the comments - genuinely, ask me anything about the numbers, the pivot, or the SEO side.