Two hours a day. Given back.

Kingsland Fabrications is a steel fabrication specialist in Warrington. Every job passes through a paint-prep step, but the paint notes that tell the paint shop what to do had to be generated by hand, from PDF cut lists produced upstream by the engineers. That job was landing on the shop-floor supervisor, sometimes eating up to two hours of his day. Two hours he wasn’t fabricating. Two hours he wasn’t quality-checking.

We built a pipeline that takes the PDF cut lists straight to auto-generated paint notes. Mistral’s OCR model extracts the structured data out of each cut list, and a small processing layer turns that into the paint-note format the shop uses. Shipped phased, discovery to production in two to four weeks. First cut needed a round of iteration on the prompt and validation layer (drawings vary, and the cut-list formatting isn’t always consistent) but once dialled in, the supervisor stopped needing to touch it.

The outcome isn’t clever AI for its own sake. It’s a senior member of the team back on the shop floor, doing the work only he can do. Kingsland now fabricates more per day and quality-checks more per day, without adding a head to the team. That’s the shape of applied AI that actually moves the needle for a fabrication business.