Evidence pack for gluten / gluten-free changeover cleaning — the AOECS audit stops being a sprint.

The problem

The annual sprint before the auditor — and the batch nobody can reconstruct when an alert lands.

In a food products factory that shares a line between wheat-containing and gluten-free product, the operational reality of cleaning between batches usually looks like this:

1. The operator takes photos with his personal phone at every cleaning — they sit in his camera roll, mixed in with his birthday pictures. When he goes on vacation, those photos exist nowhere in the plant.
2. The shift lead fills in a spreadsheet at the end of the shift, ticking a “cleaning OK” box. The signature comes after the fact, sometimes the next day, almost always without going back to the workstation to verify.
3. Three weeks before the AOECS, IFS or BRC audit, the sprint begins: the quality manager chases photos on phones, rebuilds spreadsheets, requests missing signatures and assembles a dossier that looks coherent. Two weeks of extra work for someone who was already stretched.
4. And if a health alert lands mid-year — a celiac consumer reporting a reaction, an analysis coming back positive on a specific batch — the authority's question is very specific: “show me how the line was cleaned on the day of batch 247-B”. Reconstructing that particular changeover, weeks later, is often impossible. And not being able to prove it is the same as having done it badly.

The classic system works 99% of the time. That 1% is what puts the Crossed Grain symbol on the pack at risk — and with it, the contract with the retailer that demands it.

How it fits the IRIS system

iLEAN does not add a fourth system — it turns the cleaning into a file that closes itself.

The problem with the evidence pack is not that the cleaning is done badly: the problem is that the proof that it was done well lives in places incompatible with an audit (personal phones, spreadsheets, the shift lead's memory). iLEAN acts as the putty that picks up that proof at the exact moment it is being generated, packages it, signs it and deposits it in your QMS. Without anyone changing the way they clean.

Edge captures the before/after photo at every critical point. Agents assembles the per-changeover dossier with timestamp, owner and hash. The person signs on the spot — the file is closed before the next batch. The audit stops being a sprint.

The two iLEAN pieces applied to the cleaning evidence pack:

• Edge — a terminal with machine vision (CNN) that runs a pre-programmed route through the critical points of the food products line: mixer, flour hopper, forming belt, oven, cooling belt, packing outfeed. At each point, a pre-cleaning photo and a post-cleaning photo with a timestamp. It works with no network: if the plant loses WiFi, Edge keeps capturing and storing locally; when the network returns, it syncs. What is critical cannot depend on connectivity.
• Agents — an agent picks up the images from Edge, cross-references them with the cleaning SOP in force, the order changeover recorded in the MES (which batch ended, which batch begins) and the signature of the person who validated on the spot, and packages the per-changeover dossier as a PDF with an integrity hash. It deposits it in the QMS or the repository you already use. The file lives in your system — iLEAN does not build a fourth silo, it feeds the one you already have.

There is a cross-cutting argument worth making explicit here: the “boring, concrete agent” that closes the file for the inspector is not a marketing promise. We apply it in aerospace — with the evidence documentation an inspector demands on a landing gear before the part is released — and it is exactly the same problem an AOECS auditor faces with a batch changeover: show me, with a photo and a signature, that this critical point was properly closed on that specific day. The industry with the most at stake in every single part (aerospace) has been solving it this way for years. Food is arriving at the same place.

See the full IRIS architecture →

Before and after

Annual sprint vs. a file closed before the next batch

Aspect	Cleaning documented by hand	With iLEAN Edge + Agents
Capturing the photo	Operator's personal phone, private camera roll	Edge terminal on the floor, pre-programmed route, central archive
AOECS / IFS / BRC pre-audit	A 2-3 week sprint rebuilding dossiers	Zero. The file is already closed and signed for every changeover
Signature of the person in charge	After the fact, sometimes the next day	On the spot, in situ, anchored to the photo pack
Reconstructing a specific batch months later	Hunting for photos on phones, remembering who was on shift	Filter by batch number → signed PDF in 3 clicks
Operation with no network	n/a (the operator's phone does not depend on it either, but his camera roll does)	Edge keeps capturing on its own power; syncs when the network returns
Integrity of the file	A spreadsheet that can be edited after signature	PDF with a cryptographic hash — any change shows up

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.

• SME-sized food products factory, 1 shared line with daily or weekly gluten / gluten-free batch changeovers, current AOECS certification plus IFS Food.
• Edge + Agents pilot on that line (terminal with cameras at the 5-6 critical points + integration with the MES/QMS). First value expected within a few weeks — from the very first audited changeover.
• Indicative payback between 4 and 9 months, depending on the frequency of batch changeovers, the current cost of the pre-audit sprint (quality hours + plant hours) and the cost of a potential non-conformity or re-audit.
• The hard lever is not losing the Crossed Grain symbol: re-audit, explanations to the retailer that demands it, risk of being pulled from the shelf over a “gluten-free” claim you cannot defend. A single major non-conformity avoided pays for the pilot.

And the quality manager's reasonable doubt

“What if the AI assembles a dossier that looks right but is not?” — the evidence pack iLEAN generates is an anchored task, not free generation: the agent takes real photos that Edge captured, data from the MES and the operator's human signature, and puts them together in a template. It does not invent anything. In anchored tasks, the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: the person signs on the spot before the dossier is closed. The three safety rings exist precisely for this.

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