Camera on the weighbridge control panel

The problem

The instrument behind billing and traceability writes on paper.

The scale is the instrument behind billing and traceability for everything entering and leaving the plant. Today its data only survives on a paper ticket or a local log nobody cross-checks against the rest of the process. Replacing it with a connected unit costs downtime and significant investment.

• The weighbridge decides what is billed and what enters the traceability record for everything that comes in and goes out of the plant.
• Many units, and the fixed sorting plant's control panel next to them, are over ten years old and run isolated from the network, by design or by age. Their manufacturers often no longer support an upgrade path.
• The weighing survives on a printed ticket or a local log nobody cross-checks against the intake form, the sorting or the outbound note. When a client disputes a tonnage, someone digs through ticket rolls to find it.
• Replacing the scale with a connected unit means days of downtime, recalibration and a significant investment — so the project never leaves the budget drawer, year after year, while the scale keeps weighing correctly.

How it fits the IRIS system

Connect in panel photo mode — the scale stays exactly as it is.

Connect's panel photo mode: a fixed industrial camera captures the weighbridge screen, OCR plus an anchored LLM reads every weighing and automatically matches it to the same truck's intake record (see the intake form example). The scale stays exactly the same; it just stops being a data island.

Nothing is installed in the weighing terminal and no cable is touched, so the scale's calibration and legal verification are untouched. It simply stops being a data island and starts feeding the same truck's record. The day the plant decides to replace it, the history captured so far stays in the system.

See the full IRIS architecture →

Before and after

The isolated scale versus the scale being read

Aspect	Today	With iLEAN Connect
Where a weighing lives	Ticket or local log	Searchable history by truck, client, date
Match with the intake form	By hand, if ever	Automatic by plate and time
Ticket vs billing note	Discrepancies found at month-end	Cross-checked per weighing
Gross, tare and net	Typed from the ticket	Read off the display
Scale calibration	—	Untouched, no rewiring
Replacement project	In the capex plan	Deferrable

Loose paper or local-log weighings → an instant searchable history by truck, client or date.

Impact estimate

Impact estimate — to be validated against your billing discrepancies.

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 5-10 months, depending on daily weighings and on how often tickets and billing notes disagree today.
• Discrepancies between the scale ticket and the billing note eliminated, which is where margin leaks quietly every month and where client disputes usually start.
• Every weighing becomes an instant, searchable history by truck, client or date, matched to its intake form.
• And replacing a scale that still weighs correctly stops being an urgent investment. The capex can wait until the scale actually needs replacing.

Estimated payback 5-10 months, eliminating discrepancies between the scale ticket and the billing note. *Estimate to validate*.

And the fair question from the production manager

“Can a camera reading a display be trusted for a weighing that gets billed?” — fair question, because that number ends up on an invoice and on a certificate. The task is anchored: a fixed display, digits in a known position and physically bounded values, where the best models drop below 1.5% error [1]. A reading above the scale's capacity or inconsistent with the tare is discarded, and every weighing still goes past the operator's tablet before it enters the record. The camera reads; a person still confirms what is billed.

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