Clean before the next species

The problem

The sign-off that protects the next species is the one with least evidence behind it.

Today a supervisor signs off "clean" with no objective evidence.

• Every species change, and every switch between your own production and a toll client's animals, requires cleaning that has to be certifiable.
• Today a supervisor walks the line, looks, and ticks clean. That judgment is usually right and is almost never documented.
• When a complaint arrives weeks later, the tick proves that somebody signed, not what they saw, and the plant has no way to demonstrate the state of the line at that moment.
• Meanwhile the changeover sits on the shift's critical path, so the pressure to sign quickly is highest exactly where the evidence matters most. Everyone on the floor knows this, which is why the sign-off is the part of the procedure people are least comfortable defending.

How it fits the IRIS system

Edge with JIDOKA AI — cameras compare against the reference state.

Edge + JIDOKA AI. Cameras with a CNN trained to recognize "clean vs. leftover from the prior species/batch".

Fixed cameras at the critical points compare what they see against the reference for a clean line, and flag anything that looks like residue from the previous species or the previous client's batch. Startup stays blocked until quality signs, and the signature now rests on images taken at that moment rather than on a memory of having walked past ten minutes earlier.

See the full IRIS architecture →

Before and after

The changeover today versus the changeover with evidence

Aspect	Today	With iLEAN Edge
Basis for the sign-off	A walk and a judgment	Images against the reference
Residue from the previous species	Found later, or never	Flagged before startup
What the tick proves	That somebody signed	What they were shown
Startup of the new species	On the supervisor's word	Blocked until signed on evidence
Between toll clients	Same tick	Same evidence, per client
Who signs	The supervisor	The supervisor, on the images

blind sign-off → sign-off based on visual evidence.

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.

• We do not attach a payback range to this one, and it is fairer to say so: its return depends on what a non-conformity or a toll-client complaint has actually cost you, which is a number your quality lead has and we do not.
• What it protects against is concrete: a non-conformity raised on a changeover, and a toll client claiming their animals ran on a line that was not clean.
• From a blind sign-off to a sign-off backed by images of each critical point, kept with the changeover record.
• And the changeover stops absorbing the extra margin that gets added today purely because nobody can prove the line was ready.

Protection against non-conformity or a toll-client complaint. *Estimate to be validated*.

And the fair question from the production manager

“What if the camera blocks a changeover for no good reason?” — that is the right objection, because the changeover is on the critical path. The comparison is made against the reference state for that specific point, and when there is doubt nothing is blocked silently: it escalates to the supervisor, who signs on the evidence in front of them. On the reading itself, a fixed point with controlled lighting is an anchored task where the best models drop below 1.5% error [1]. It stops on a verifiable discrepancy, never on model uncertainty.

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