Zero weeks reconstructing before the audit

The problem

Nothing went wrong, and the nonconformity comes from the paperwork.

The day the organic certifier's auditor arrives, or the sanitary-registration inspector, someone today spends weeks reconstructing the cleaning and receiving traceability for the period. The auditor asks for proof, the plant shows scattered paperwork, and the auditor raises a nonconformity simply because of documentation gaps — not because anything actually went wrong.

• The day the organic certifier's auditor arrives, or the sanitary-registration inspector, someone spends weeks reconstructing cleaning and receiving traceability for the period.
• Cleaning sign-offs are in one binder, strain and extract receipts in another, expiry checks in a spreadsheet, supplier certificates in someone's email, signatures wherever they were taken. Nobody has the full picture until someone builds it by hand.
• The auditor asks for proof, the plant shows scattered paperwork from four different places, and a nonconformity is raised because of documentation gaps — not because anything actually went wrong.
• And the people who rebuild the file are the same ones who should be running quality during those weeks, so the plant is least supervised exactly when it is being audited.

How it fits the IRIS system

Edge and Connect capture at the moment of the event; Agents files it by standard.

With iLEAN Edge + Agents, every reactor cleaning or critical receiving event automatically generates before/after photos, a decision log and timestamped signatures.

The evidence is not prepared for the audit: it is produced by the event itself, at the moment it happens, by the people who were there, with nothing left to remember or rebuild weeks later. Each cleaning and each critical receiving generates its photos, its log and its signatures, and Agents classifies them by standard and by period as they happen. The audit stops being a project and becomes a download.

See the full IRIS architecture →

Before and after

Preparing the organic audit today versus the evidence pack always ready

Aspect	Today	With iLEAN Edge + Agents
Time to prepare an audit	Weeks of reconstruction	Zero: the pack exists
Proof of cleaning between batches	A signature in a binder	Before and after images with signatures
Receiving of strains and extracts	Delivery notes in folders	Lot, expiry and validation, by date
Organization	By whoever filed it	By certifying standard and period
Gap in the evidence	Discovered by the auditor	Flagged by Agents beforehand
Access for the auditor	Boxes on a table	A download of the period's pack

Agents organizes all that evidence by certifying standard. The auditor logs into the site and downloads the period's evidence pack, with nobody at the plant spending hours reconstructing anything. Estimated payback 6 to 12 months.

Impact estimate

Impact estimate — to be validated with your audit calendar.

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 6-12 months.
• Continuous protection of both certifications — sanitary registration and organic certifier — instead of a sprint before each visit, whether the visit is announced or not. An unannounced inspection finds the same pack as a scheduled one.
• Weeks of reconstruction per audit, often for more than one person, returned to the quality team, who can spend them on the plant instead of on binders.
• And documentation-gap nonconformities, the ones raised when nothing actually went wrong, disappear, because the gap is flagged when it happens, not when the auditor finds it, and is closed by the person who can still close it.

Continuous protection of certifications (sanitary registration and organic certifier) · estimated payback 6-12 months.

And the fair question from the production manager

“Will the certifier accept a digital pack?” — what the auditor asks for is evidence that is complete, traceable and signed, and the pack is exactly that, with original images and timestamps. Classifying each piece of evidence into the right standard and period is an anchored task where the best models drop below 1.5% error [1], and quality reviews the pack before it is shared. Nothing is altered on the way: the images and signatures are the originals, with their timestamps intact.

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