July 14, 2026 · 5 min read

Building Presmo: What Shipping a Solo AI SaaS Taught Me

Everyone who has ever turned a dense document into a presentation knows the feeling: the content already exists, and yet the deck still takes hours. That gap — between having the material and presenting it — is the problem I set out to solve with Presmo, a micro-SaaS that turns a PDF, pasted text, or URL into a themed, fully editable PowerPoint deck.

It's live, it has paying customers, and I built and run it alone. This post isn't a technical walkthrough — it's the story of what shipping a real AI product solo actually taught me.

The hard part isn't the AI — it's trusting it

The demo version of any AI product takes a weekend. The version you can charge money for takes months, and almost all of that time goes into one thing: making the AI dependable enough to put your name on.

Large language models are like brilliant colleagues who occasionally have a very bad day. Left unchecked, they invent facts, drift from instructions, and produce output that almost works. The bulk of Presmo's engineering went into making sure a customer never sees that bad day: the model's output is constrained, checked, and corrected at every step, and a deck either comes out right or the user is told clearly that something went wrong.

That last part became a principle I'd now apply to any AI product: never let the product silently degrade. If the AI is unavailable, Presmo says so. It never quietly hands over something worse and hopes you won't notice — because users absolutely notice, and trust is the only real moat a small product has.

Small on purpose

Presmo runs on a single small server. That's not a limitation I'm embarrassed about — it's the strategy.

A solo founder pays for every moving part twice: once in hosting bills, once in maintenance attention. Every piece of infrastructure you add is something that can wake you up at night. So every decision went through the same filter: does this add enough value to justify what it costs to run and to babysit? Most of the time the answer was no, and the product is better for it. Boring technology, aggressively few moving parts, and discipline about what actually matters to the customer.

What does matter: Presmo produces real, editable PowerPoint files — genuine slides, charts you can restyle, speaker notes in the notes pane — not images of slides. That single product decision, more than any feature, is what makes people convert. The output isn't a demo artifact; it's the actual file they need for Monday's meeting.

Quality is the sales pitch

Presmo's free tier gives you your first deck at full quality — no watermark tricks on the thing that matters, no crippled output. That was deliberate: for a product like this, the output is the marketing. If the first deck impresses, the upgrade sells itself. If it doesn't, no pricing page will save you.

It's a nerve-wracking choice for a bootstrapped product, because free usage costs real money on every conversion. The answer was guardrails rather than stinginess — usage limits that keep the free tier sustainable without making the first impression worse.

Building a SaaS from Sri Lanka

Some of the most educational obstacles had nothing to do with AI.

Stripe doesn't support Sri Lankan business accounts. For most indie-hacker advice on the internet, "just use Stripe" is step one — and it simply wasn't available to me. The answer was merchant-of-record billing platforms, which handle payments, tax, and compliance on your behalf, plus a local payment option for LKR customers. It cost me a week of research and integration work, and taught me a lesson every founder outside the usual markets learns eventually: the standard playbook assumes a geography you might not live in. There is always another route; it's just not in the tutorial.

The same pattern repeated in smaller ways — hosting platforms that block the ports you expected to use, services that assume a US business address. None of it was fatal. All of it was invisible in every "how I built my SaaS" post I'd ever read.

What actually got it shipped

Looking back, the things that got Presmo from idea to paying customers weren't clever. They were:

1. A painfully small scope. One input, one output, done extremely well. Every feature idea that didn't serve "document in, great deck out" went on a list I mostly ignored.
2. Testing like a team, alone. Presmo has a few hundred automated tests and they run on every change. As a solo founder, the test suite is the second pair of eyes you don't have.
3. Treating quality as the growth channel. No ads, no growth hacks — the first free deck is the pitch.
4. Accepting constraints early. The payment situation, the small server, the solo pace — fighting those would have burned months. Designing around them shaped a leaner product.
5. Shipping before it felt ready. The version that got the first paying customer was missing things I considered essential. Customers disagreed about what was essential.

Presmo is live at presmo.app — the first deck is free and full-quality, so you can judge the output yourself. And if you're building an AI product and want an engineer who has taken one from idea to production solo, let's talk.