I’m a Product Manager Who Built a Travel-Tech Startup Solo — Using AI as My Engineering Team

The €300 roaming bill — and why I was uniquely placed to fix it

A few years ago, I landed in Istanbul and turned on my phone. Two days later, I had a €300 roaming bill from my carrier. I hadn’t made a single call.

What made it sting more: I’d spent years working inside the travel industry and knew exactly how unnecessary it was. I’d optimized booking funnels, run A/B tests on checkout flows, worked on airline booking engines and GDS connections. I understood distribution, conversion, and travel tech infrastructure from the inside.

The eSIM product existed. The technology was there. What was broken was the experience — and building better experiences is exactly what product managers (PM) do.

The Missing Piece: A PM Brain, Not a Dev Team

I’m not a developer by background. My career is in product management — 10+ years across travel, aviation, and e-commerce. At some point I did a full-stack web and mobile bootcamp so I could speak more fluently with engineering teams and prototype ideas myself. I came out of it able to build, but slowly.

Then generative AI (GenAI) changed the equation.

In early 2026, I started building Simsima. What would have taken a team of 3–5 engineers and 6+ months, I shipped solo in a matter of weeks, using AI coding tools as my engineering team. I was the product manager, designer, QA, and architect — the AI handled the bulk of the code generation.

But here’s the non-obvious part: being a PM made me better at using AI to build, not worse. Every good AI-assisted development session starts with a clear spec, a precise user story, and an explicit definition of done. That’s literally what PMs do. The discipline of thinking in flows before writing code — something developers often skip — turns out to be the critical skill in an AI-assisted development workflow.

What We Actually Built — And The Stack Behind It

Simsima is a travel eSIM marketplace: 190+ countries, plans starting from a few dollars, instant delivery by email. The product has two surfaces: a web app and a React Native/Expo mobile app, both built end-to-end by one person.

The full stack:

• Frontend/App: Next.js 16 (web), React Native + Expo (mobile)
• Backend: Node.js, Express, PostgreSQL — self-hosted on Hetzner via Coolify
• Infrastructure: Cloudflare for CDN and WAF
• Analytics: PostHog + GTM for funnel tracking
• Lifecycle: Brevo for transactional email, Crisp + PostHog for push notifications
• Payments: Stripe (card, Apple Pay, Google Pay)
• eSIM supply: API integration + direct carrier relationships

I also built a custom backoffice dashboard with real-time margin tracking, unit economics, EU VAT rules, and pricing management. On a team, that’s a 2-week sprint for an engineer. Solo with AI, it was a weekend.

The signature product feature — the Smart Link — is what differentiates Simsima from every other eSIM provider. Instead of forcing users to scan a QR code or enter a 40-character manual activation code, our Smart Link is a single tap in the confirmation email that automatically installs the eSIM. It’s the UX detail I fought hardest for, and it shows in the data: we’re seeing a +22% eSIM activation rate versus QR-only flows.

The Growth Bets: 25 Languages, Zero paid ads

We launched in 25 languages simultaneously. This was a deliberate SEO-first bet. Most eSIM competitors are English-only or English-first. We indexed 200+ country destination pages × 25 language locales from day one — that’s thousands of URLs with fully localized content and hreflang signals, not machine-translated filler.

The logic: a Japanese traveler searching for a France eSIM in Japanese deserves a Japanese-language page, not an auto-translated one. A Brazilian searching for a Japan eSIM in Portuguese should find us before they find a competitor with a single-language site. This architecture is invisible at launch and compounds over time.

We don’t run paid ads. Not yet. Our acquisition strategy is SEO + referral (our SimCoins loyalty program rewards both referrers and new buyers) + cashback (4% flat on qualifying purchases, credited as SimCoins). The goal is to build a durable organic engine before pouring paid traffic into the funnel.

What The First 90 Days Showed Us

We shipped our first paying customers in April 2026. The PostHog data since then:

• Weekly active users grew 25x in 10 weeks — from a handful of visitors in April to a consistent and growing base by mid-June. Early numbers, but the slope is what matters.
• First purchases from 8 countries, 4 continents — Spain (home base), then Hong Kong, Italy, Japan, United States, Germany, Netherlands, United Kingdom. We didn’t market in most of these. They found us organically.
• 3% funnel conversion — among users who engage with a plan page, 17% view details, 9% add to cart, 5% start checkout, 3% complete. Solid for an early-stage checkout; top-of-funnel drop-off is where the work is.
• Average time from first view to purchase: ~2 days — eSIM purchases are deliberate, not impulse. People plan a trip and think it through. Email nurturing matters more than we initially expected.
• First AI-search attribution: Perplexity — not Google, not a paid ad. A traveler asked an AI assistant, What’s the best eSIM for Hong Kong? and found us. LLMs are becoming real acquisition channels for travel-intent queries.

Six Lessons — For Marketers and For Builders Using AI

1. PM thinking is the killer skill in AI-assisted development. The bottleneck isn’t code anymore — it’s clarity. Clear user stories, explicit acceptance criteria, and precise flow definitions translate directly into better AI output. If you’re a PM thinking about building your own product, you’re more ready than you think.
2. Solve the UX problem before scaling traffic. The Smart Link is what actually makes Simsima different from dozens of competitors. We almost shipped without it because the QR code technically works. The right call is always to eliminate friction first, then grow.
3. Multilingual SEO is underrated and underused. Building a 25-language content architecture from day one is a moat that’s very hard to copy quickly. In a category where most competitors are English-first, being globally indexed is a structural advantage.
4. Optimize your AI content surface, not just your web content. We publish an llms-full.txt file specifically to help AI models understand our product accurately. Our first Perplexity-attributed purchase wasn’t luck — it came from being well-described where AI search queries land. Treat LLMs like you treat Google: give them clean, structured, accurate information.
5. Loyalty programs matter most for infrequent purchases. People travel 3–5 times a year and could easily forget you exist between trips. SimCoins create a pull-back mechanic: every purchase earns credits toward the next one. In a category with zero switching costs, this is the retention engine.
6. Watch your time-to-purchase (TTP) metric. Our median gap between checkout start and completed payment is about 24 hours. That tells us buyers deliberate. Trust signals — reviews, a visible refund policy, security indicators — at the final step can meaningfully improve conversion.

What’s next

We’re in the build phase: optimizing conversion, expanding our eSIM partner network, and letting the SEO flywheel compound. The first 90 days validated the model — travelers will buy eSIMs online when the experience is frictionless.

For anyone building a product solo and wondering whether AI tools can really replace a dev team: they can’t replace one. But for a PM who knows exactly what they want to build, they’re close enough. SimSima is proof.