The bundle ticket enters the system with one photo

The problem

The unit of work for the whole sewing line has a paper identity.

the bundle is the unit of work for the whole sewing line, but its identity lives only on a stapled paper slip that can be lost, soaked or misplaced between stations. When it's lost, nobody knows the size-color-style until it turns up at the next station and someone guesses. The line supervisor loses hours every week reconstructing lost bundles' identities.

• Every bundle that leaves the cutting table carries a ticket with size, color, style, piece count and the operation route through sewing. For hours, that slip is the bundle's only traceability.
• It is stapled, pinned or tucked into the bundle, and between the spreader and the last sewing operation it gets soaked, torn, separated or left on the wrong cart. Nobody notices until the next operation needs to know what it is sewing.
• When it goes missing, nobody knows the bundle's size and color until it turns up at the next station and someone guesses from the pieces or the shade.
• The line supervisor loses hours every week rebuilding lost bundles' identities, and a wrong guess is how a size-color mix-up starts its way toward packing.

How it fits the IRIS system

Connect in photo mode — the cutting assistant keeps ticketing exactly as before.

Connect photo mode captures the ticket via the cutting assistant's phone or tablet. An anchored LLM extracts structured fields (bundle number, size, color, style, operation route, piece count) and stores them in central memory. Before it reaches the ERP, the data passes through the early human-validation tablet (case 5) — one glance, two taps.

Nothing changes at the cutting table except one photo at the moment the bundle is born. From then on, the bundle has a second identity that cannot be soaked or lost, and a lost bundle is traced by its last logged photo instead of by a guess. That is also why it is usually the first case a plant approves: it asks nobody on the floor to work differently.

See the full IRIS architecture →

Before and after

The paper bundle ticket versus the captured bundle ticket

Aspect	Today, on paper	With iLEAN Connect
Where the bundle's identity lives	A stapled slip	Central memory, queryable from any station
Lost or soaked ticket	Identity guessed at the next station	Traced by its last logged photo
Size, color, style and route	Readable only next to the bundle	Structured fields per bundle
Supervisor's week	Hours chasing lost bundles	One query by bundle number
Wrong guess on size or shade	Travels on toward packing	Caught against the cutting order
Cutting assistant's routine	—	Unchanged: one photo

bundle identity existing only on paper → bundle identity queryable from any station in central memory. Lost bundle → traceable by its last logged photo.

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 daily bundle volume.
• Supervisor hours recovered from chasing and rebuilding lost bundles, which today stay invisible because nobody logs them. Count them for one week and the number usually surprises the plant manager.
• Bundle identity goes from existing only on paper to being queryable from any station in central memory.
• And a lost bundle stops being a guess: it is traced by its last logged photo.

estimated 4-9 month payback depending on daily bundle volume, recovered supervisor hours spent chasing lost bundles. *Estimate to validate*.

And the fair question from the production manager

“What if it misreads a size or a style code on the ticket?” — reading fields from a ticket with a known layout is an anchored task, where the best models drop below 1.5% error [1]. And there is a second net: before anything reaches the ERP, the summary goes to the line-side tablet and a person confirms or fixes it with two taps.

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