2 bets in parallel: how a boring website shop killed the AI SaaS

The Case: $100 in ads → 2-3 paying clients

What he does. He builds fixed-price small-business websites delivered in 7 days, with a free homepage preview so the client only pays if they like what they see.

Backstory. He was a developer-designer at a performance marketing firm in India, hated the job, and tested two side bets before quitting: an AI product-photoshoot generator, and a plain website shop. He shipped both, watched the numbers, then quit without knowing which would carry him. The AI tool died with zero traction. The "boring" one started slow, then compounded.

Product. - Fixed-price small-business websites, 7-day turnaround - Free homepage preview built up-front, no card required - Wireframe plus AI-assisted production pipeline - Client pays only after seeing the preview - Distribution shifted to Meta ads once case studies existed

Pricing. Fixed one-time price per site, exact amount not disclosed. Retainer not disclosed. A commenter guessed $400 per site; the operator never confirmed it.

Key numbers. - Ad spend: $500+/month on Meta - Funnel per $100 ad spend: 5-7 inquiries - Inquiry-to-client conversion: 2-3 clients per 5 inquiries - Delivery time: 7 days - First 2-3 clients: from ex-employer contacts - Current run-rate: more per month than his prior annual Indian tech salary (absolute figure not disclosed) - Photoshoot AI bet: zero paying users - Time from quit to profitability: not disclosed

Stack. Wireframes plus unspecified AI production tools, Meta Ads, a lead-capture form, cold outreach for seed clients. Exact tools not disclosed.

Why this works (and what most readers will miss)

1. The free preview isn't a marketing hack, it's a risk transfer. Small businesses have been burned by freelancers who take a deposit and ghost. He removes that fear entirely. What most readers miss: this offer was economically painful for solo designers before AI production. Spec homepages used to be hours of unpaid work per prospect, which is why almost nobody offered them. He says he uses wireframes plus AI tools to speed up the build — that's what makes "free preview, pay if you like" survivable at volume. He isn't selling a website. He's arbitraging the exact moment spec work became affordable again.

2. Two parallel bets, and the one with the faster time-to-revenue won. The AI SaaS demanded belief, integrations, model costs, retention. The website shop demanded one manual delivery and a wire transfer. When you place two bets simultaneously, whichever hits cash first gets your attention, and attention is what compounds. He didn't pick the "smart" bet. He picked the one that shipped a paid invoice.

3. The distribution ladder is warm → referral → paid — never paid first. Ads only started producing after he had case studies from ex-job contacts. Founders who skip straight to Meta with zero social proof burn $500 and quit at week three. He earned the right to pay for reach.

The Distilled Read

Productize the delivery, not the tool. Every reader looks at this case and asks "what's his stack?" Wrong question. The product is a promise: a working homepage in your inbox within a week, and you owe nothing until you see it. The AI in the pipeline is invisible to the buyer. Nobody pays for the model, they pay for the outcome. The operator who names his AI is a vendor. The operator who names his outcome is a business. He didn't build a Cursor-for-websites. He built a delivery contract with an AI-shaped back office, and that's why buyers exist.

Templatize until the preview is one click. Right now he still hunts, drafts, presents. The version of this business that scales isn't hiring humans — it's collapsing the preview stage to a single form submission that returns a finished homepage in ninety seconds. Same offer, ten times the throughput. This is the fork every service operator hits: hire, or template. Templates win because they don't sleep, don't renegotiate, and don't quit at month eight. Weekend homework: find the layouts you've already built more than once, freeze them, and let AI fill the copy on autopilot.

The economics only work because the failed bet was cheap. He ran two bets in parallel while employed, one obviously "big" and one obviously "small," and the small one paid rent. That isn't luck, that's portfolio construction. Every operator reading this should be running two bets right now, not one, and the boring bet should always be the one with the shortest path to first invoice. He doesn't say what the AI photoshoot idea cost him, only that it died quietly with nobody willing to pay. The website shop cost him a form. When the AI died, he didn't grieve — he had runway underneath him. You only need one out of twenty side-bets to hit, but you have to actually place twenty.

Distribution is a ladder, and paid ads sit at the top rung. He seeded from ex-job contacts, converted a handful, banked the case studies, waited for referrals to accumulate, and only then flipped on Meta. Every step earned the next one. Reverse the order — ads first, case studies never — and you'll report $500 lit on fire with nothing to show. The reason his $100 ad spend now produces 2 to 3 clients is not the ad creative. It's the screenshots of finished sites sitting in his portfolio when a prospect asks "is this legit?" Ads don't build trust. Ads deploy trust that already exists.

And the moat nobody in the thread named: the preview is a portfolio flywheel. Every free preview he builds — even for prospects who ghost — is a rankable asset. A screenshot for the next ad, a case for the next inquiry, a template for the next vertical. Losing a client costs him nothing and adds to his inventory. In 18 months, when every teenager with a Bolt subscription is offering the same fixed-price site, the winner will be whichever operator has the deepest library of category-specific previews already indexed on Google, already tagged inside Meta's optimization, already familiar-looking to the next dentist or gym owner scrolling their feed. The commodity is the site. The moat is the archive of unpaid work that becomes paid distribution. He didn't build a service business. He built a compounding proof machine, and he hasn't noticed it yet.

Steal-the-Playbook

1. Pick one vertical, not "small businesses." Dentists. Yoga studios. Auto detailers. One niche means one wireframe, one ad angle, one testimonial format that all reinforce each other.
2. Build the preview engine before the offer. Wireframe → AI copy → hosted homepage in under an hour. If it takes longer, template harder. The offer only defends itself when spec work costs you near-zero.
3. Seed with 3 warm contacts, not 3 cold ads. Old colleagues, past clients, LinkedIn DMs. Free work in exchange for a filmed testimonial and permission to publish the finished site.
4. Publish the case studies before you spend a dollar on ads. Landing page with three logos, three screenshots, three quotes. No logos, no ads. Skip this and you pay to be ignored.
5. Turn on Meta ads only when your reply-to-lead time is under one hour. Speed compounds. A 24-hour reply on a $100 ad spend is a wasted funnel.

Stack: wireframe library + one AI copy/image tool + a Notion form + Meta Ads Manager. Setup: one weekend. Ongoing cost: $500+/month once ads are on.