Broiler chicken — Welfair welfare is earned every day, and the audit collapses if the data isn't there.

The problem

Welfare happens every single day, the data lives across 40 farms, and the dossier is rebuilt the day before.

A poultry integrator working with several dozen family farms faces the same pattern in every batch:

1. The environmental sensor in the house records temperature, humidity, ammonia and CO₂ locally; very often it never even reaches the ERP.
2. The feed scale is connected to the feed manufacturer's system, not the integrator's.
3. The grower writes down daily mortality, observations and any management event in a notebook or a messaging-app thread.
4. The integrator's vet visits the house once or twice per batch — and the photo taken on their phone is seen by nobody until it lands in the ERP.
5. The processing plant contributes live weight, transport mortality and post-mortem inspection findings (footpad dermatitis, hock burn) in yet another system.

On Welfair audit day, someone at the integrator rebuilds five spreadsheets by cross-checking phone calls, messaging-app photos and paper tickets — only to discover, far too often, that the data now needed was never recorded. The classic system works 99% of the time. That 1% is what loses the rating, and with it a retail channel.

How it fits the IRIS system

iLEAN Tracer doesn't add one more system for the grower — it seals the cracks between the ones you already have.

The problem in broiler is not a lack of data: the environmental sensors, the feed scale, the grower's notebook, the vet's phone and the processing plant's ERP all have the information. The problem is that it lives in five heterogeneous islands, spread across dozens of farms, and on audit day nobody has time to cross-check them. Tracer is the filler that closes those gaps, without asking you to change your ERP, your sensors or the feed manufacturer's system.

Edge samples vision in the house. Connect captures sensors, scales and notebooks. The agent stitches the five events per batch against the Welfair protocol and signs the dossier. The person validates.

The three iLEAN pieces applied to multi-farm broiler welfare:

• Edge — a terminal with vision (CNN) in the house for periodic sampling: real stocking density, flock gait (gait score), presence and use of environmental enrichment, mortality hotspots. It works without a network — on a farm, connectivity comes and goes, and Edge keeps recording and processing on the power from the panel.
• Connect — captures the environmental sensors via API or serial port, the feed scale tickets (printed ticket → photo → data), the grower's notebook (photo + transcribed voice) and the vet's messages. And it captures on the side whatever arrives as a retailer email or an integrator technical note.
• Agent — stitches the indicators per house and per batch against the Welfair protocol, flags deviations while they can still be corrected (not at the end of the batch) and generates the auditor-ready welfare dossier per farm and per batch. If an indicator drifts, it raises the flag to the vet and the technical director before the batch closes.

See the full IRIS architecture →

Before and after

Manual Welfair documentation vs. a per-batch dossier with iLEAN Tracer

Aspect	Sensors + notebooks + messaging apps	With iLEAN Tracer
Welfair indicators per batch	Rebuilt the day before the audit	Live throughout the batch, per house
Detection of welfare deviations	At batch close or at the processing plant	While they can still be corrected on farm
Multi-farm coverage	Limited to the vet's visits	Continuous, with vision sampling and sensors
Dossier for the Welfair auditor	Five spreadsheets rebuilt by hand	Per-farm, per-batch dossier, signed
Passport to the retailer's shelf	The integrator's summary sheet	Indexable per-batch dossier, with signed data
Workload on the grower	Notebook + apps that duplicate work	Their usual notes, collected by Connect

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 from your farms and your batches. We set it out so the committee has an order of magnitude; we refine it during the diagnostic.

• Poultry integrator with 30-80 family farms, standard and/or slow-growing broiler, European retailers demanding Welfair or an equivalent as a condition of channel access.
• Tracer pilot on 3-5 pilot farms (environmental sensors + Edge for vision sampling + Connect against the integrator's ERP, the feed scale and the grower's notebook + Welfair agent). First value expected within a few weeks.
• Indicative payback between 5 and 9 months, depending on the number of batches per farm per year and the cost of a documented Welfair incident in recent years.
• The hard lever is a single batch held on the shelf, avoided + the reduction in vet time spent rebuilding spreadsheets (which goes back into field management). Expected reduction in audit time ≥ 30%.

And the technical director's reasonable doubt

«What if the AI gets it wrong and gives me a badly measured stocking density?» — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI limits itself to counting and to recontextualising data from a sensor into a dossier, the best models brought the error below 1.5%^[1]. And even so, the dossier only goes out with a human signature: the vet and the technical director validate before anything reaches the Welfair auditor.

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