Launched in 14 days, first paying customer in week three

The client was a first-time founder, previously a product manager at a mid-size SaaS company, with an idea she'd already validated through a dozen customer calls: AI-assisted first-draft generation for marketing agencies producing client reports. She had a Figma file, a waitlist of around 200 signups, and a talk slot at an industry conference five weeks out where she planned to demo the product live.

There was no code. The constraint was the date on the conference program, which wasn't moving, and a budget that assumed a single focused build rather than a drawn-out engagement.

She needed a real, working product — not a clickable prototype — because the conference audience was exactly the buyer she needed as her first paying customers, and a fake demo in front of them would cost more credibility than it bought.

We scoped the minimum real product: account creation and login, Stripe Checkout for subscription billing, a report-generation flow backed by an LLM call, and a dashboard to view and export past reports. We used a hosted auth provider rather than building session and password-reset flows from scratch — on a two-week timeline that's a solved problem not worth re-solving, and it's a call we'd defend on almost any MVP with a deadline like this one.

The generation step routes through the same multi-provider setup we use on our own products rather than a single hardcoded model integration, so the founder wasn't locked to one vendor's pricing or uptime from day one. Data lives in Postgres on Neon, and the app is deployed on Cloudflare — a default pattern for us, because it means the founder pays close to nothing at zero users and the same setup holds without a rebuild if usage spikes after the conference talk.

We shipped a working, deployed link by the end of week one, and treated every day after that as refinement against real usage rather than one big-bang launch at the end. That gave the founder something she could put in front of her own advisors and early waitlist users a full week before the conference, rather than seeing the product for the first time on stage.

In week two, the founder changed the pricing model — from a flat monthly fee to a seat-based structure — after a customer call convinced her the flat fee was leaving money on the table. That's exactly the kind of requirement change every real project has, and it landed after billing was already wired up. We rebuilt the Stripe integration around seats with three days left before the conference, which meant cutting a planned polish pass on the dashboard to protect the launch date instead.

The trade-off we made explicitly was to protect the billing change and the launch date over visual polish, and to say so to the founder rather than quietly deciding it ourselves. She agreed immediately — a working seat-based product with a plainer dashboard was worth more on stage than a beautiful dashboard billed the wrong way.

The product was live and demoable on stage as planned. The founder's own tracking showed her first paying customer converting in the third week after the conference, from a conversation that started at her demo table.

She's stayed on a monthly partnership retainer since launch, and we've shipped two feature requests directly from customer feedback in the months since — the kind of steady iteration a fixed-scope project doesn't leave room for.