Non-Technical Builder Rebuilds a Product

Since late 2025, a non-technical builder at ScrapIT, supported and motivated by the Quadshift AI Enablement team, has been doing agentic engineering across ScrapIT's platform, using AI to build the things he understands better than anyone: the product and the customer.

Among the impressive projects in development:

• GizmoAI, a natural-language reporting tool that lets non-technical yard operators ask for a report in plain English. Claude writes the SQL, runs it against a secured, per-customer Metabase instance, and returns the underlying query for audit.
• A photo-classification module that takes images of bulk-purchased scrap and has multiple AI models vote on the match, flagging misclassified metal and contamination before it costs the yard money.
• An automated log monitor and an AI report writer with live SQL calls that round out the operational side.

Then we rebuilt the entire front end. Using Claude Code, hundreds of ScrapIT screens were rewritten in about three months, work that would traditionally take a year or more, modernizing the UI and clearing the way for a major product upgrade. He also wrapped the web app in an iOS layer with an automated build-and-submit pipeline, so a fix ships to the App Store with no developer and no Mac in the loop.

The newest work turns these into customer-facing products, like an embedded ScrapIT copilot that acts as in-app help and a natural-language reporting engine with native MCP support. The person who knows the product and its purpose best is now the person shipping it fastest.