Every condemnation gets its dossier

The problem

The history exists, scattered, and only gets assembled when somebody asks for it.

The official vet asks for a period's history and it has to be rebuilt piece by piece.

• Every condemnation is classified into a regulated category and has a destination that has to be documented, with its weight and its species.
• Today the record depends on somebody writing it down in the moment, in a book or a sheet, while the line keeps moving.
• When the official vet asks for a period's history, it is rebuilt piece by piece, and the reconstruction inherits every gap left by whatever nobody wrote.
• Worse, the pattern that would be most useful to the plant — which category rises, in which species, from which supplier — is never visible, because nobody assembles the history except under pressure. The record is kept as an obligation and never read as information, which is a waste of the only data that says something about the animals arriving at your ramp.

How it fits the IRIS system

Edge captures in the moment, Agents organizes by period and category.

Edge captures photo and data at the moment; Agents organizes the dossier by period and category.

The photo and the data are taken where the decision is made, at the inspection point, not written up later from a note. Agents then keeps the dossier organized by period, species and category, so a request for a history stops being a project somebody has to clear a week for and becomes a query somebody runs while the vet is still in the room.

See the full IRIS architecture →

Before and after

The condemnation record today versus the closed dossier

Aspect	Today	With Edge and Agents
When the record is made	Afterwards, from memory	At the moment of the decision
What it contains	A line in a book	Photo, category, species, weight, destination
A history for a period	Rebuilt over days	Queried in minutes
A category rising over months	Nobody sees it	Visible as a trend
Destination of the material	Noted separately	Part of the same dossier
Who classifies	The official vet	The official vet, unchanged

manual reconstruction → dossier available instantly.

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.

• Estimated payback 6-12 months, depending on how often inspections and history requests actually land.
• From a manual reconstruction each time it is asked for, to a dossier that already exists when the question arrives.
• Days of somebody's week recovered every time a period history is requested, usually from the person you can least spare.
• And a category climbing in one species becomes visible while it is still a trend, which is when a conversation with the supplier is still worth having.

Estimated payback of 6-12 months depending on inspection frequency. *Estimate to be validated*.

And the fair question from the production manager

“Does an automatic dossier mean the vet's decision is being second-guessed?” — no, and the boundary is deliberate: the classification is the official vet's and the system records it, never proposes it. What the system does is make sure the photo, the weight, the species and the destination are attached to that decision. Recording those fields against a known structure is an anchored task where the best models drop below 1.5% error [1], and the record is confirmed before it closes.

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