Cheese aging traceability — every wheel leaves with its full history, not in fragments.

The problem

Months in the room, and a shift report to account for it.

Aging a cheese is an exercise in patience: the batch goes into the room, and across the weeks and months it spends inside, thousands of small things happen — a thermocouple that drifts, a door left open too long, a night when the climate control resets itself, a turning the operator did properly but never wrote down because the shift changed at exactly that moment. None of those things, on its own, ruins anything. All of them together do.

The classic system — shift reports every 4 hours, a manual round, a spreadsheet at the end of the month — covers the visible points. What happens between the points is recorded nowhere. And what happens between the points is exactly where these things get decided:

• The rind that comes out thin when it should have come out thick, because humidity dropped three nights running and nobody saw it.
• The batch that loses its PDO seal because the inspector asks for the full profile and all there is are 6 readings a day.
• The year sales climb and production realizes it cannot reconstruct which batch was the good one — because what it has is scraps of paper, not data.

The master cheesemaker knows — their eye carries it. The day they retire, half the plant walks out with them. That island of knowledge is the largest one, and the worst protected.

How it fits the IRIS system

iLEAN seals the cracks between the room, the turning robot and the shift report.

The problem with aging is not a lack of sensors — it is that the data lives on islands: the room's climate in the BMS, the turning robot in its own PLC, the operator's rounds on paper, the deviations in the master cheesemaker's head. iLEAN acts as the putty that fills the gaps between those systems, without asking you to change the climate control, the robot or the ERP.

Edge measures continuously where a shift report never reached. Connect captures manual turning just as it captures the robot's. The agent rebuilds each wheel's history. The master cheesemaker signs — and the knowledge stops walking out the door with them.

The three iLEAN pieces applied to aging traceability:

• Edge — a terminal with sensors distributed across the levels and zones of the room (the top and the bottom of a large room do not breathe alike), which also reads the signal from the existing climate control and from the turning robot where there is one. It works with no network: if the plant loses WiFi, it keeps measuring and recording on the panel's own power, because what is critical cannot depend on connectivity.
• Connect — captures manual turning through whatever route the team already uses: the batch code scanned with a phone or earpiece at the end of the pass, voice dictation, a photo of the shift report. And it also captures what arrives from outside (an email from the regulatory council, a WhatsApp from the importing customer with a question about last winter's batch) at second zero, without anyone forwarding anything.
• Agent — cross-checks the room profile, the turnings, the entries and exits, the cleanings and the incidents per batch. It rebuilds the batch file the moment it is asked for, not the following week. If there is a significant deviation, it warns the master cheesemaker before the pattern sets in. The person signs — the batch is never released on its own.

See the full IRIS architecture →

Before and after

Aging with shift reports vs. aging with a per-batch dossier

Aspect	Rounds + spreadsheet	With iLEAN Edge + Connect + Agent
Measurement frequency	Every 4 h, at one convenient spot	Continuous, several points per room
Turning record	Paper checklist, signed by the shift	Robot integrated + Connect for manual turning
Overnight deviation	Discovered the next day, if at all	Alarm at second zero, to the manager's channel
Dossier for IFS/BRC/PDO auditor	Rebuilt by hand, weeks	Signed per-batch file, automatic
The master cheesemaker's knowledge	Lives only in their head	Captured as a pattern, stays in the plant
Operation without network	n/a	Edge keeps measuring on the panel's own power

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.

• Aged-cheese dairy with several mid-sized rooms, semi-industrial profile, a mix of manual and robotic turning. Roughly what it takes to instrument one standard room as a pilot.
• First value expected within a few weeks: per-batch dossier running, deviation alarms going to the manager's channel, integration with the existing climate control.
• Indicative payback between 4 and 9 months, depending on the weight of deviation-driven scrap in your product mix and on the cost of rebuilding files by hand for the auditor or the PDO council.
• Hard levers: reduction in scrap per batch (out-of-spec rind, deformation, affinage problems) ≥ 30% as a conservative estimate; office hours spent rebuilding files eliminated; a stronger position in front of a PDO inspection.

And the technical director's reasonable doubt

“What if the AI invents a value that was never measured?” — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI simply reads a real sensor value and recontextualizes it (summarizing the room's day, generating the batch dossier, comparing against history), the best models brought error below 1.5% ^[1]. And even so, what is critical is never released on its own: iLEAN proposes, the person signs. The three safety rings exist precisely for this — the room's agent does not release the batch, it only prepares it.

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