A 1998 tunnel oven with an isolated Windows 2000 PC — and a per-batch baking dossier the IFS auditor wants tomorrow.

The problem

The oven works perfectly. The problem is what never comes out of it.

In a food products factory, the tunnel oven is the most expensive and the most reliable asset on the floor. It has been baking for 15, 20, sometimes 25 years. The baking curve was fine-tuned back in the day by the master baker — together with a technician from a manufacturer that has since closed down — and the oven has delivered ever since. The problem is not the oven; it is everything around the oven:

1. Only the master baker can see the baking curve. It lives in his head, in a dog-eared folder next to the panel, and in whatever the operator writes down by hand when he remembers. The day the master baker retires, so does the baking pattern for every product, the fine tuning of every line, and the trick that keeps the oven exit from coming out pale when the humidity in the plant shifts in October.
2. The industrial PC is cut off from the network. It is a Windows 2000 machine (sometimes Windows NT) with a proprietary driver from the oven manufacturer. It has no usable Ethernet, no Internet, and IT has spent years asking everyone to “please leave it alone”. And they are right: that PC controls the bake, and if it goes down, the line stops.
3. The IFS/BRC auditor asks for per-batch baking data and there is no record. There is a paper sheet the shift lead signs at the end of the day. No curve, no setpoint, no logged alarm. The non-conformity lands in the report and from the following year the customer demands corrective action — or moves to the competitor who does have it.

This is not negligence. Integrating old machines was unviable for decades: the project cost more than the savings, the drivers did not exist, and modernizing the whole oven cost more than buying a new one. That equation has changed.

How it fits the IRIS system

Connect, intermediate mode — the putty between your 1998 oven and the rest of the 21st century.

iLEAN Connect was designed with this kind of machine in mind. The philosophy is graduated capture: Connect adapts to what your machine can give, not the other way round. Three modes depending on the oven in front of you:

• Manual — a 100% analog panel, needles and pilot lights. A fixed camera is pointed at the panel and a neural network reads the values the way the master baker would. The installation work is a bracket and a power cable.
• Intermediate (your tunnel oven's case) — the machine has an isolated, old industrial PC with an RS-232/RS-485 serial port or a flat-file export. Connect hooks up to that port or reads the export folder, and extracts temperature by zone, belt speed, setpoint, alarms and events. Nothing new is installed inside the old PC — Connect is the external box that reads.
• Integrated — the oven has OPC-UA or modern Modbus. Direct integration into the SCADA or the MES.

Connect reads the old PC's serial port or the screen with a camera. The Agents cross the curve with the batch in the ERP. The per-batch dossier assembles itself. The master baker stops being the sole keeper of the baking pattern — without losing any of his authority over the line.

The effect in a food products factory:

• A real baking curve, per batch, in real time. Every tray coming out of the tunnel carries its temperature by zone, the seconds spent in each section and the alarms that were triggered. That is what the IFS auditor asks for, and it is what never existed on paper.
• The master baker's knowledge is captured as data. Not to replace him — so that on the day he is on vacation, the second in charge has the pattern from the last 200 good batches to work from. The veteran's eye finally stops being an island.
• It works with no network. Connect captures locally and syncs when it can. The plant can go three days without WiFi: baking continues, capture continues, the dossier is rebuilt afterwards with no gaps.
• Bidirectional, but with a brake. Connect carries, the Agents reason, the person signs. The system can suggest to the shift lead that the curve is drifting because the air in the plant is more humid than average — but the one who decides to slow the belt down is the master baker, not the agent.

See the full IRIS architecture →

Before and after

Tunnel oven with a notebook vs. tunnel oven with Connect

Aspect	Today: paper sheet + isolated PC	With iLEAN Connect in intermediate mode
Baking curve of the batch	In the master baker's head	Captured in real time, by zone, by minute
Data for the IFS/BRC auditor	Sheet signed at the end of the shift, no curve	Per-batch dossier with setpoints, real curve and alarms
The day the master baker retires	The baking pattern leaves with him	Pattern stored as a reference for the successor
Oven alarms	A pilot light someone sees if they walk past	Timestamped event, notified through a channel
Cost of touching the industrial PC	n/a — IT forbids going near it	Zero — Connect reads from outside, nothing installed inside
Integration with ERP/MES/spreadsheets	Manual, weeks after the batch	Automatic match against the work order

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.

• Mid-sized food products factory, 1-2 tunnel ovens aged 15-25 years, isolated Windows 2000/NT industrial PC or a mixed analog-digital panel.
• Connect pilot on one oven (Connect box + serial port reading or camera over the panel + integration with the batch in the ERP/spreadsheet). First value expected within a few weeks — the per-batch dossier starts being generated the day the box goes up on the wall.
• Indicative payback between 3 and 9 months, depending on how much weight the audit dossier carries with your end customer (retailer, UK/US export) and on the documented cost of your last baking incident.
• The hard lever is a single IFS/BRC non-conformity avoided at audit — or a single customer complaint about a badly baked batch that can now be traced to the exact minute. That alone pays for the pilot.
• Reference anchor: the powder coating case documented in the iLEAN base closed its pilot with a real investment close to €30,000 in year 1 and a conservative payback of ~3 months, with first value in weeks. Same mechanics, different sector: the order of magnitude applies.

And the plant manager's reasonable doubt

“What if Connect misreads the panel and logs a temperature that is not the real one?” — hallucination is a problem of free generation, not of anchored tasks. Reading a 7-segment display, a thermometer needle or a flat file from the old PC and carrying that value into a database is an anchored task: there is a source, and there is a value that can be verified. In this kind of task, the best models brought error below 1.5% ^[1]. And even so, what is critical is never decided alone: the data is stored, the agent proposes, and the master baker or the shift lead signs. The three safety rings exist precisely for this.

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