The Brief
The client came to us with a clear problem: their industry (logistics) had no modern tool for real-time shipment visibility that small operators could afford. Enterprise solutions existed but cost $50K+/year. They wanted a web app that could launch in under two months, targeting owner-operators running 5-20 trucks.
We scoped it to five core features: live GPS tracking dashboard, automated status notifications, driver mobile check-in, basic invoicing, and a customer-facing tracking page. Everything else was post-MVP.
The 6-Week Timeline
Week 1-2: Discovery, design system setup, component library, database schema. We used AI to generate initial wireframes from the brief, which we then refined with the client.
Week 3: Core dashboard and tracking engine. This is where we hit the scope creep wall - the client wanted real-time chat added. We pushed back hard and won.
Week 4-5: Notifications, driver app, customer portal. Parallel workstreams with daily integration.
Week 6: QA, performance optimization, deployment, and launch prep.
How AI Accelerated Everything
AI wasn't a feature of the product - it was a feature of our process. Claude generated component specs from Figma exports. Copilot scaffolded React components from those specs. We used AI to write initial test suites, generate seed data, and even draft the API documentation.
Conservative estimate: AI saved us 120 hours across the six weeks. That's three full developer-weeks - nearly the difference between shipping on time and missing the deadline.
The Decision That Nearly Killed It
In week 3, we agreed to a "small" scope addition: custom branding per customer account. It sounded trivial. It wasn't. It touched the auth system, the CSS pipeline, the email templates, and the tracking page. What we estimated at 8 hours became 28. We caught it early enough to recover, but it was a reminder: in a 6-week timeline, there's zero margin for unplanned work.
Launch Results
The MVP launched on day 42. Within the first month: 340 trucks onboarded, 12,000 tracking events processed, 94% uptime, and a Net Promoter Score of 72 from the beta cohort. Four of our five core assumptions validated. The one that didn't - that operators would self-onboard - led to adding a guided setup flow in sprint 1 post-launch.