AI for veterinary pharma — multi-species, multi-dose, a single safety net.

The problem

A veterinary plant is a pharma plant with one more layer on top.

A veterinary factory carries everything any human pharma plant carries — GMP, process validation, data integrity, traceability — and adds a layer that has no exact equivalent in human pharma:

1. Multi-species with radically different doses of the same molecule — the same antibiotic goes to a small dog, to cattle, to swine. One decimal place in the doser is the difference between efficacy and residue.
2. Atypical dosage forms — medicated premixes incorporated into feed, oral solutions in drums, intramammary mastitis products, pour-ons, ruminal boluses. None of them fits a classic blister line.
3. EU Regulation 2019/6 and CVMP/EMA — antimicrobials under watch, batch-to-animal traceability via the Union Product Database, withdrawal period visible on the label. The penalty for “residue in meat or milk” reaches the human consumer: it is not a “minor” failure.

The shift lead, the quality manager and the QP are juggling the ERP, the MES, an R&D spreadsheet with the latest formulation, a doser panel nobody ever integrated and an email from the supplier with the new CoA. The classic system works 99% of the time. That 1% is what reaches the CVMP — and each time it is a recall, not a deviation.

How it fits the IRIS system

iLEAN does not add a fourth system — it seals the cracks between the ones you already have.

The problem in veterinary pharma is not a lack of information: it is information living on islands that, at the critical moment (the multi-species SKU changeover, the premix recipe adjustment, the intermediate cleaning that finished with high conductivity), does not reach the decision-maker in time. iLEAN acts as the putty that fills those gaps, without asking you to change your ERP, your MES or the packaging line.

Edge sees the container, the fill and the label. Connect reads the order, the cleaning record and the old doser's panel. The agent cross-checks against EU Regulation 2019/6 and, if something does not add up, holds the batch. The person signs — never the other way round.

The three iLEAN pieces applied to a veterinary plant:

• Edge — a terminal with machine vision (CNN) over the line. It reads the fill of the oral drum, the tamper-evident closure and the label (species, dose, withdrawal period). If the label declares “swine 10 mg/kg” and the active order was “canine 5 mg/kg”, it fires the rejector. It works without a network. If the plant loses WiFi, Edge keeps reading and holding.
• Connect — captures the work order whether it comes from the ERP, the vertical MES or a spreadsheet from the galenic department; it reads the doser's old panel every few minutes; it digitizes the supplier CoA arriving by email and the WhatsApp from the regulatory veterinarian with a new antimicrobial requirement. Everything enters at second zero.
• Agent — cross-checks the order + the cleaning history + the container image + EU Regulation 2019/6 and CVMP. If there is a deviation, it does not send an email at 10 p.m.: it holds the batch and alerts the QP through whatever channel they use. It generates the batch dossier ready for inspection without rebuilding it by hand.

See the full IRIS architecture →

Before and after

Manual release vs. cross-checked release with iLEAN

Aspect	Classic release + checklist	With iLEAN Edge + Connect + Agent
Multi-species SKU changeover	Paper checklist, signed by the shift lead	Edge verifies the label + the agent verifies certified intermediate cleaning
Medicated premix recipe	Galenic spreadsheet → printout → taken to the line	Captured in the ERP/MES/spreadsheet at second zero, no forwarding
Withdrawal period on the label	The operator confirms it by eye	Edge reads it, the agent cross-checks it with the target species and the order
Batch dossier for CVMP/inspection	Rebuild by hand, weeks	Generated continuously, with an image and signed JSON
Operation without a network	n/a — depends on WiFi and servers	Edge keeps running on the panel's own power
Decision on a suspect batch	A call to the QP at 10 p.m.	Batch held; the QP signs once they have the full context

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 veterinary plant, 1 oral drum line + 1 medicated premix line, 60+ multi-species SKUs (dog, cat, cattle, swine, poultry).
• Edge pilot on the oral line (camera over the fill + label + actuator + integration with the ERP/MES order). First value expected within a few weeks.
• Indicative payback between 4 and 9 months, depending on the history of CVMP non-conformities, the frequency of rework from labeling or dose, and the cost per inspection finding.
• The hard lever is a single recall or serious deviation avoided: rework, non-conformity, brand damage with veterinary distributors. One recall pays for the pilot and leaves the line ready.
• Reference standard borrowed from automotive to anchor the quality bar: on the order of 25 PPM in defects per million ^[1]. Veterinary is not automotive, but the logic of the bar is the same.

And the QP's reasonable doubt

“What if the AI gets it wrong and releases a mislabeled batch?” — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI merely recontextualizes a piece of data from one system to another (reading the label and comparing it with the work order), the best models brought error below 1.5% ^[2]. And even so, what is critical is never decided alone: iLEAN holds the batch and the person signs. The three safety rings exist precisely for this.

^[1] 25 PPM standard in automotive quality — Symestic.

^[2] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.

Related: AI OTC manufacturing · Drug-device combination products · EBR — electronic batch record.