Photographing the work order at every lot startup

The problem

The lot is defined on paper that nobody reads until evening.

the work order prints from ERP data and any change (bar-stock lot, tool used, quality note) gets handwritten. The first-off sheet is handwritten by IATF requirement. Both can sit 4-8 hours undigitized, and if a complaint arrives that same afternoon, someone has to manually reconstruct which lathe and which tool produced the part.

• The work order prints from the ERP, but the moment reality deviates it gets corrected in pen: a different bar-stock lot, a swapped insert grade, a quality note on a thread or a chamfer.
• The first-off sheet is handwritten because IATF 16949 asks for it: critical dimensions, concentricity, chamfers, thread gauge result and the inspector's signature.
• Both documents can sit 4 to 8 hours in a folder next to the lathe before anyone in the technical office types them into the ERP.
• If a complaint lands that same afternoon, someone has to walk the floor asking which lathe and which tool produced that shaft or that brake piston, and the answer depends on who remembers.

How it fits the IRIS system

Connect in photo mode — the paper stays, the lot goes live.

Connect photo mode. A grounded model extracts structured fields (part number, bar-stock lot, lathe, first-off dimensions, sign-off) from the two photos into central memory. Zero habit change for the operator; early human verification (in-line tablet) confirms before the data crosses into the ERP.

Nothing is redesigned and nobody types more: the operator adds two photos to a routine that already exists. In a plant where every lathe starts several lots per shift, that is the cheapest traceability gain on the table, and usually the first case a plant manager approves because it asks nothing of the floor.

See the full IRIS architecture →

Before and after

Lot startup on a clipboard versus lot startup in memory

Aspect	Today, on paper	With iLEAN Connect
When the lot appears in the system	4-8 h later	At the moment of startup
Pen corrections on the work order	Lost unless retyped	Extracted and tied to the lot
First-off critical dimensions	In a binder by the lathe	Queryable by part number
Which lathe and which insert	Asked around the floor	One web lookup by lot
Technical-office transcription	Every order, every day	Recovered
What the operator does differently in bar turning	—	Same paper, plus two photos

4-8h of startup-digitization latency → instant. Manual reconstruction of which lathe/tool produced a lot → instant web lookup by lot.

Impact estimate

Impact estimate — to be validated with your numbers.

The block below is an estimate to be validated against your plant's actual data. We put it forward so the committee has an order of magnitude; we refine it during the assessment.

• Estimated payback 4-9 months, depending on how many work orders start per day across the lathe fleet.
• The saving comes from transcription time in the technical office, which today goes into copying work orders and first-off sheets lot after lot.
• Startup digitization goes from 4-8 hours to instant, so the lot is queryable while it is still running.
• And the question of which lathe and which tool produced a lot becomes a web lookup instead of a floor walk.

estimated payback 4-9 months depending on daily order volume, recovered technical-office transcription time. *Estimate to be validated*.

And the fair question from the production manager

“What if it misreads a diameter on the first-off sheet?” — reading fields from a sheet with a known layout is an anchored task, where the best models drop below 1.5% error [1]. And there is a second net: the summary goes to the tablet at the lathe, any value outside the drawing tolerance is highlighted, and the operator confirms or corrects it before anything crosses into the ERP.

^[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.