A product they were buying, rebuilt better in 10 hours.

The brief

Optiwiser depended on a third-party SaaS data product and was dabbling with off-the-shelf AI coding tools. The product was a black box they couldn’t shape to their clients, and the team had no real AI workflow of its own. For a shop serving a quality-hungry client base that meant one thing: falling behind on a capability they could own.

The build

We didn’t pitch. We built. In about 10 hours we rebuilt the paid SaaS product in-house on Claude, with proper evals, tests, MCP connections and reusable skills. It shipped to their clients roughly a week later. Then we paired with the team, committed the reusable skills to their repo and handed it over so they could run it themselves.

The result

• Higher quality than the product it replaced, live in production within a week.
• The whole five-engineer team adopted Claude as their default workflow.
• Self-sufficient in weeks, not months: from dabbling to independent operators.
• The skills and the system live in their repo, owned outright.

Debrief

The fastest way to convince a sceptical founder isn’t a deck. It’s replacing a tool they already pay for: visible, falsifiable, immediate return.

With AI, build-versus-buy flips. Things that were obviously purchases become obviously builds, and a capable team can own them outright.