Photo of the weighbridge intake form

The problem

The first link of the traceability chain sits in a booth tray for hours.

The waste identification form (or intake delivery note) is the first link in all downstream traceability: where the truck comes from, what waste type is declared, what plate number. Today that paper sits in the scale booth until someone transcribes it at shift's end, 4-8 hours later. Any same-day issue (misdeclared waste, suspicious mix) has already been processed before anyone reviews it, because the data only lives on paper.

• The waste identification form is the first link of everything downstream: which site the truck comes from, which waste code is declared, which plate and which carrier.
• It is filled in by hand by the driver or the operator, often in a hurry, and stays clipped in the scale booth until someone transcribes it at shift's end — four to eight hours later.
• During that gap the load has already been tipped, sorted and sometimes fed to the crusher, so a misdeclared code or a suspicious mix is found after it has been processed.
• When a contractor calls asking what was delivered from their site on Tuesday, the answer is a box of paper forms sorted by date, not a lookup. And the contractor needs that answer for their own waste plan.

How it fits the IRIS system

Connect in photo mode — the operator takes the same photo they already took.

Connect's photo mode: the weighbridge operator photographs the form with a phone or the booth's tablet. An anchored LLM extracts structured fields (origin, waste type, plate number, gross weight, carrier) and stores them centrally. Zero habit change: the operator takes the same photo they already took for internal control, only now the data flows straight into the system.

It asks the weighbridge for nothing new: the photo already existed for internal control, and now it carries the data. That is why it is usually the first case a waste manager approves, and the one the rest of the chain hangs from. Without a clean intake record, no later control can say where a ton came from.

See the full IRIS architecture →

Before and after

The intake form in a tray versus the intake form captured

Aspect	In the booth today	With iLEAN Connect
When the load exists in the system	At shift's end, 4-8 h later	On arrival at the scale
Declared waste code	Read when someone types it	Extracted and tied to the truck
Misdeclared load	Found after tipping	Flagged before sorting
Lookup by plate or site	Leafing through paper	Web search by truck and date
Driver's handwriting	Retyped by guesswork	Read, low confidence flagged
Weighbridge routine	—	Unchanged: one photo

4-8 hours of latency until transcription → zero latency. Scattered paper records → instant web lookup by truck and date.

Impact estimate

Impact estimate — to be validated with your daily truck count.

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 truck volume at the weighbridge.
• Admin transcription time recovered: the forms stop being typed at shift's end by someone who never saw the truck.
• From 4-8 hours of latency down to zero: the record exists while the truck is still on the scale.
• And scattered paper becomes an instant web lookup by truck, plate, date or contractor site, from the office or from the yard.

Estimated payback 4-9 months depending on daily truck volume, recovering admin transcription time. *Estimate to validate*.

And the fair question from the production manager

“What if it misreads the waste code a driver scribbled?” — reading fields from a form with a known layout is an anchored task, where the best models drop below 1.5% error [1]. The code is also checked against the list of codes your environmental authorization accepts, and the summary reaches the booth tablet so the operator confirms or corrects it before it enters the record. A wrong code caught at the booth costs a tap; caught at inspection, it costs a finding.

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