No data enters your master without a person having seen it first

The problem

The fear that AI will write data on its own is reasonable, and it blocks everything.

In a business where the product is defined by its identity — grade, coating, gauge and heat — the fear that artificial intelligence will write data on its own and leave somebody cleaning up the ERP master is the single biggest cultural brake on digitalisation. And it is a reasonable fear. A contaminated master is worse than having no data at all: it produces wrong decisions that look well founded. But the operator does not want to type either. Every second on a keyboard is a second away from the line. Between "let the AI write on its own" and "make the operator type" there is no need to choose: there is a third option.

• In a business where the product is defined by its identity — grade, coating, gauge and heat — the biggest cultural brake on digitalization is the fear that an AI will write data on its own and leave somebody cleaning up the ERP master.
• And it is a reasonable fear: a contaminated master is worse than no data at all, because it produces wrong decisions that look well founded.
• But the operator does not want to type either. Every second on a keyboard is a second away from the line, with the crane waiting.
• Between letting the AI write on its own and making the operator type, most plants choose neither and the data stays on paper, which is the worst of the three outcomes and the most common.

How it fits the IRIS system

Connect with early human verification — two taps at the line.

Connect in early human verification mode. An industrial tablet at every line. When the system reads the cutting order, interprets the machine panel, parses a schedule change that arrived by email or picks up a dictation from the supervisor, it all appears there as a two-second visual summary: coil, heat, grade, coating, gauge and strip pattern. Two taps: ✓ it enters the system, ✏️ it gets corrected first. The interface is built to validate in two seconds, not to enter data. Humans in command, implemented as architecture rather than as a promise.

Humans in command implemented as architecture rather than as a promise: whatever the system reads — the cutting order, the panel, an email, a dictation — appears on the tablet as a two-second visual summary, and it does not exist in the master until a person has confirmed it. It is the piece that gets the other eleven approved, because it answers the steering committee's first question before it is asked.

See the full IRIS architecture →

Before and after

Today's data entry versus the validated entry

Aspect	Today	With iLEAN
Who writes to the master	A person, typing	A person, confirming
Time at the tablet per coil	Minutes at a terminal	Two seconds, two taps
What appears as the author	Whoever typed it, later	Whoever looked at the coil
A wrong grade or coating	Discovered in the customer's press	Corrected before it enters
Where the summary comes from	—	Order photo, panel, email or dictation
Can the AI write on its own	The fear	No, by architecture

from unvalidated data contaminating the master — or data never captured out of distrust — to data captured and validated within seconds, with effortless traceability of who signed what.

Impact estimate

Impact estimate — to be validated with your numbers.

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.

• No standalone payback — this is an enabling piece, and the brief says so: it is not directly monetizable.
• Its value is strategic: without early human verification the other eleven pieces never get approved at steering level, because no sensible management team lets a system write into its master with nobody in front of it.
• From unvalidated data contaminating the master — or data never captured out of distrust — to data validated within seconds.
• And effortless traceability of who signed what, on which line and at what time, which is what gets asked for when a claim lands and what today has to be reconstructed from a folder.

an enabling piece, not directly monetizable. Without early human verification the other eleven pieces of the matrix never get approved at steering level, because no sensible management team lets a system write into its master with nobody in front of it. *Strategic value*.

And the fair question from the production manager

“If a person signs in the end, what is the AI saving?” — the work is not in deciding, it is in assembling. Today somebody would have to go to a terminal, find the coil and type grade, coating, gauge, width and heat. With the gate, the extraction — an anchored task where the best models drop below 1.5% error [1] — arrives assembled, and the operator only confirms or corrects. Every correction is stored: the system learns where its reading fails on your documents.

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