Hazardous waste traceability in the chemical industry — a broken manifest is a penalty, a bad classification is a premium rate.

The problem

Hazardous waste is managed across three systems that do not talk to each other.

The environmental/HSE manager of a chemical plant lives with the same friction every week:

1. The point of generation. A purge stream, a sludge, a spent solvent, a contaminated container. The composition depends on the batch recipe, on the yield of that shift and sometimes on an incident — and it is almost always assumed, not measured.
2. The licensed contractor. Every contractor has its own portal, its own manifest format, its own collection calendar, its own rates tiered by EWC/EPA code. Every collection is an email, a phone call, a spreadsheet updated by hand.
3. The cradle-to-grave file for the regulator. At inspection you have to demonstrate: how much was generated, which EWC code was assigned, which carrier collected it, what intermediate treatment, what final destination. All of it rebuilt by hand from emails, paper manifests and treatment certificates that arrive weeks later.

The classic system works almost every time. But the day the inspector asks for the cradle-to-grave record of last month's EWC 070104* batch and the treatment certificates cannot be found, the penalty is serious and the responsibility still sits with the generator. It is not a people problem — it is a problem of information living on islands between plant, contractor and authority.

How it fits the IRIS system

iLEAN does not add a fourth system — it seals the cracks between the plant, the contractor and the regulator.

The plant already has a waste management system (sometimes an ERP module, sometimes a vertical tool, sometimes a spreadsheet). The contractors already have their portals. The regulator already has its formats. Each one does its own job well. The problem lives in the joints — between the point of generation and the manifest, between the manifest and the treatment certificate, between the certificate and the file for the inspector. iLEAN acts as the putty that fills those gaps, without asking you to change contractor or ERP.

Tracer records the waste at source, with its real composition. Writer assembles the manifest. Connect closes the loop with the contractor and captures the treatment certificate. The person validates; the cycle never closes on its own.

The three iLEAN pieces applied to hazardous waste traceability:

• Tracer — records the waste at the point of generation: which stream, which process batch it came from, real volume/mass (scale, meter or validated estimate), probable composition cross-referencing recipe + SDS + lab analysis where there is one. It assigns an EWC or EPA code with a reasoned proposal — the person validates or reclassifies. The waste stops being a drum in a yard: it is an object with an identity from second one.
• Writer — generates the manifests (EPA e-Manifest, EU transfer identification document, national equivalents) already filled in with generator, carrier, contractor, code, mass and hazard properties. It asks the person to review and sign; it never issues on its own.
• Connect — integrates with the portals/APIs of the licensed contractors, captures the intermediate and final treatment certificates as they come in (sometimes email, sometimes portal, sometimes a photographed paper), and attaches them to the waste batch file. When the cradle-to-grave cycle is complete, the dossier is signed and ready for the regulator — with no later reconstruction.

See the full IRIS architecture →

Before and after

Waste management in spreadsheets vs. cradle-to-grave with iLEAN

Aspect	Classic management + spreadsheets + contractor portals	With iLEAN Tracer + Writer + Connect
Identifying the waste at source	Label written by hand at the end of the shift	Recorded in the plant with probable composition and EWC/EPA code
EWC/EPA classification	Historical default, sometimes over-classified	Cross-reference of SDS + recipe + lab analysis; the person validates
Transfer manifest	Filled in by hand, paper copy	Draft generated, signed by the manager
Treatment certificate	Arrives by email weeks later, filed by hand	Captured automatically, attached to the file
Cradle-to-grave file	Rebuilt for the inspector	Living dossier, signed step by step
Waste KPI per unit of product	Calculated once a year	Live by shift, by batch, by SKU

Impact estimate

Impact estimate for your plant — to be validated with your numbers.

The block below is an estimate to be validated with the specific data of your plant. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• Chemical plant with 4-10 different hazardous waste streams, 2-5 licensed contractors, a waste module in the ERP or a vertical tool plus a parallel spreadsheet, and periodic regional/national inspections.
• Pilot on one priority stream (the highest-volume one, or the one the contractor reclassifies most often), integrated with SDS + ERP + one contractor. First value expected within a few weeks.
• Expected reduction in administrative management hours of over 30% in the first months.
• Expected reduction in waste over-classification (which triggers the contractor's premium rate) of between 10% and 25% depending on the stream.
• Indicative payback between 4 and 9 months, counting administrative savings + less over-classification + avoiding a non-conformity at inspection.

And the environmental manager's reasonable doubt

"What if the agent classifies a waste stream wrong and we end up with a worse regulatory problem?" — hallucination is a problem of free generation, not of anchored tasks. Assigning an EWC/EPA code is exactly an anchored task: the composition comes from the recipe + SDS + lab analysis; the agent proposes the code from the codified list. In this kind of task, the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: iLEAN proposes; the environmental manager validates and signs before the manifest is issued. The three safety rings exist precisely for this.

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