The store took the order. The system ran everything until it shipped.

The problem: success at manual speed

• Every cut file was built by hand — copy the house number out of the store, paste it into a spreadsheet and a template, stamp the order ID and finish, export. The shop's own estimate: ten to fifteen minutes per file, and that understates the nested sheets.
• Orders arrived on two separate store platforms, and every one of them had to walk through that same manual pipeline.
• Customers emailed and called to ask where their order was — and on an eight-to-ten-week lead time, every answer stopped someone on the floor.
• The owner was the integration layer: order status, production status, and customer comms all lived in his head.

What we built

An order-operations system that runs the gap between “order placed” and “order shipped” — without adding headcount and without replacing the tools the shop already uses.

• Cross-platform order capture. A custom app pulls orders live from both storefronts via their APIs and reconciles them into one production-status tracker — the spreadsheet the shop already trusted, upgraded into the source of truth.
• Automated cut-file generation with the shop's judgment codified. Orders become production-ready files automatically — fonts, spacing, and dimensions locked to the cutting standard, optical weighting engineered into the code, each file shipping with a rendered preview.
• Human-in-the-loop by design. The system does the fifteen minutes of production work in milliseconds; a person spends seconds approving it before steel moves.
• A production queue the floor drives — an operator picks a stage from a dropdown and that one update flows everywhere it needs to go.
• Customer self-serve tracking on the storefront and a voice line that answers the phone, plus branded email at every production stage.
• Scheduled, serverless automation keeps every data source in sync around the clock — no infrastructure for the client to babysit.

The results

Engagement started late May. The system was fully operational five weeks later.

• 3,990 orders reconciled across both platforms into one production tracker.
• 2,200+ production-ready cut files generated — covering nearly 2,000 distinct orders — each verified against a rendered preview before cutting.
• Ten to fifteen minutes per file by hand became hundreds of files in seconds. The demo that started the engagement: 500 cut files generated in a tenth of a second.
• 500+ hours of skilled labor returned to the business, at the shop's own per-file estimate — conservatively.
• Customers check their own status on the site or by phone at any hour and get a branded email at every stage. Nobody stops work to answer.
• The owner runs the operation without being the operation — the system carries order state, production state, and customer comms.

Why it worked

This wasn't an AI tool bolted onto a business. The process — order to steel to doorstep — was rebuilt end to end with the intelligence embedded where the work happens, designed by someone who has done CAD engineering and e-commerce for years. The fabricator's own judgment was codified into the system, not replaced by it. The shop floor kept its spreadsheet. The customer got answers. The owner got his time back.