Skip to content
Language
Menu
Interior of a poultry house at dawn, with light coming through the side curtains and the feed and drinker lines receding down the central aisle.

For broiler producers in Colombia, Mexico, Peru, Chile and Brazil

Your farm manager decides with a spreadsheet. This decides with your policy.

Agro Intelligence for livestock production

Management declares what to optimise for and what can never be violated. The system recommends each flock’s catch day — with a confidence band and the reasoning in one line — and issues the order to whoever must carry it out. If no date satisfies the policy, it says so instead of relaxing the constraint.

Book a demoSee how it decides
How a flock’s processing day is chosenWeight projection with a confidence band. Days before 36 are ruled out by a constraint declared by management. Among the remaining days, policy picks day 40.150020002500300035002640 g2420–2860 g28313437404346Cycle dayWeight (g)
  • Confidence band
  • Ruled out by constraint
  • Chosen day
Illustrative figures. The drawing explains the mechanics of the decision; it is not a forecast.

Poultry producers live with three information blackouts at once. None of them is solved by buying more equipment.

A worker writes on a paper form under a single hanging bulb, inside a darkened poultry house with the birds barely visible behind.

The time blackout

The settlement, or the flock close-out, arrives after the flock has gone. It says what happened at the end, but not on which day the gap opened. It is a death certificate, not a diagnosis.

The comparison blackout

Two houses with the same roof, the same equipment and the same litter look identical from the walkway. The difference forms inside: half a point of mortality, odd feed intake, a weight that is short for the age. With twelve houses that is unmanageable by hand.

The cost blackout

The production figure lives in one place and the cost that produced it lives in another, weeks later. By the time they finally meet, they are no longer useful for deciding: only for explaining.

Which of these operations do you recognise?

The underlying problem is the same — the figure arrives late, scattered and with nothing to compare it against — but it shows up differently in each kind of operation, which is why the first screen you switch on is different too.

Contract grower

Provides the house, the labour and the utilities; the integrator provides chicks and feed.

What you cannot see today

The settlement is produced by a scale that is not yours and arrives after the flock has gone. You have nothing to check it against, and the costs you do pay — labour, power, gas, litter — are consolidated by nobody.

What the system gives you

Your own auditable record to hold against every settlement: your birds, your kilos, your dates and your costs, taken in your house and with the time they were taken. It does not change the agreement; it changes that the conversation stops being word against word.

Where you start: Grow-out and Costing. The integrator schedules the catch day, so that module is not the first one that helps you.

Independent grower

Buys their own chicks and their own feed, and sells their own birds.

What you cannot see today

You carry all the cost and all the risk. Feed is your largest item and you pay all of it, but cost per kilo gets worked out at close — if it gets worked out — and by then it is no longer useful for deciding anything.

What the system gives you

Cost per live kilo and per kilo of carcass on the same record that carries feed conversion, and a recommended catch day within what you yourself declared non-negotiable. Here you do decide the date, so here you can actually optimise it.

Where you start: Grow-out, Feed and Costing together: those are the three that meet in your margin.

Breeders and hatchery

Produces the fertile egg and the day-old chick that somebody else then grows out.

What you cannot see today

A fertility problem is spotted in the hatcher, weeks after it started in one specific house. By the time the number looks wrong, the parent flock that caused it has been doing the same thing for three weeks.

What the system gives you

Every setter load is tied to the parent flock that produced it, with lay rate, fertility, hatch and contamination in the same series. "Which house did this come from" stops being a reconstruction and becomes a query.

Where you start: Breeders and Hatchery. They are the pair it makes no sense to switch on separately.

Processing plant

Receives the bird, processes it and answers for carcass yield.

What you cannot see today

Producing more stopped being enough: efficiency now depends on raw-material yield, on losses, on line availability and on traceability. And yield varies between shifts on apparently identical raw material, so the explanation is looked for inside the plant when sometimes it is outside: in the feed withdrawal, in transport, in the real age of the flock that arrived.

What the system gives you

Birds received, processed, carcass grading and condemnations tied to the flock of origin, with its grow-out history behind it. Yield stops being a number for the shift and becomes a number with provenance. It is also the end-to-end traceability the industry is starting to require, and that no per-station record can give.

Where you start: Processing, and as soon as there are two weeks of it, the cross with Grow-out.

Integrated group

Controls several links — or all of them — and answers for the margin of the whole.

What you cannot see today

Every link has its own system and its own truth. Nobody says where margin is lost between one and the next, and day-to-day judgement lives spread across a supervision layer that is expensive, hard to replace and leaves no record.

What the system gives you

One single record for the whole chain and a policy declared ONCE that governs decisions across every link, with the trail of what was decided and under which criterion. This is where it shows most that judgement stops living in people.

Where you start: Operating policy first, modules afterwards. The other way round it turns into ten dashboards.

The system changes where judgment lives inside the operation.

In a traditional poultry operation, day-to-day judgement lives in a middle layer of supervisors. That layer translates the board’s strategy into field decisions, house by house. It is expensive, it is hard to hire, it takes the knowledge with it when it leaves, and its judgement is never recorded or comparable between flocks. Here management declares the policy and the system instructs the field staff. Judgement stops living in a person and starts living in the operation.

  • Less dependence on a middle supervision layer, with all its payroll burden.
  • Knowledge accumulates instead of rotating.
  • Opening a new site stops depending on finding another manager with judgement.
Management declares the policy once; the system turns it into a daily instruction. Two sources feed it from below: what the stockperson captures — with no signal too — and the readings the environmental controller sends to the system.

Why now and not in five years?

Three facts that are not our opinion. Each with its source and its link, so they can be checked before believing anything else on this page.

The side of a poultry house at night, with the row of exhaust fans faintly lit and eight feed silos silhouetted against the dark blue sky.

  1. The next decade’s growth comes from productivity, not from more animals

    Global agricultural and fish production is expected to increase by 14% over the next decade, "mainly enabled by productivity improvements".

    Whoever grows over the next ten years will do it by getting more out of the same house, not by building another one. And getting more out of the same house is not one investment decision: it is a hundred small decisions a year about density, feed and catch day.

  2. Feed decides the result, and the result is decided before the invoice arrives

    "Feed represents about 70 percent to 75 percent of total production costs" in commercial broilers.

    No other decision in the operation moves as much money as the one that affects feed conversion. A tenth of a point on the largest cost item weighs more than almost any saving that can be negotiated across the rest of the budget.

  3. Chain traceability is already a market requirement, not an aspiration

    "The traceability of food, feed, food-producing animals […] shall be established at all stages of production, processing and distribution", and every operator must be able to identify their immediate supplier and immediate customer.

    It is the rule half the world copied, and the one that applies to anyone wanting to sell into the European market. A per-station record does not meet it: what meets it is a record running from the breeder house to dispatch, which is exactly what this system keeps.

All three push in the same direction: the coming years’ margin is won in daily decisions about the flock’s largest cost, and you have to be able to show how those decisions were made. That is not solved with more equipment or another dashboard.

In short

  • A consolidated record of the whole chain, 100% reliable — nothing deleted, nothing rewritten — and contrasted against the formulas the industry already uses to move money. That is what the system decides on.
  • Management declares the policy once: what to optimise for, what can never be violated, and how much the system decides alone.
  • Every live flock gets a recommended catch day, with a weight band and the reasoning in one line.
  • If the day comes from a projection rather than a weighing, the screen says so. Projected is never shown as measured.
  • If no date satisfies the policy, the system reports it as infeasible instead of relaxing the constraint.
  • Capture works with no signal inside the shed and syncs once coverage returns.
01

What exactly is this software?

A platform where the poultry operation is recorded and, on top of that record, an engine decides. The chain starts in animal nutrition and ends in human food, and the system follows it whole: feed, breeding, hatchery, grow-out, processing, logistics and sales, with costing and the plan-versus-execution contrast cutting across all of it, plus the marketplace outward. What the nutritionist formulates and what happens after dispatch are not its business; everything between those two points is. The difference from a poultry ERP is not what it stores: it answers “what do I do tomorrow” instead of “what happened yesterday”.

  • Full operational record of the cycle, from placement to processing.
  • Indicators are computed from a single formula for the whole system, so the dashboard, the report and the alert never disagree.
  • Field capture with no connection, with a send queue that loses nothing.
  • Every field figure comes in with a reliability mark, and a doubtful one widens the band instead of hiding.
02

What does the decision engine do that a dashboard does not?

A dashboard shows numbers and leaves the decision to whoever looks. The engine walks each flock’s possible days, removes those violating a declared constraint, scores the rest against company policy and returns one. Order matters: constraints filter, weights choose within what is already allowed.

  • No target for kilos or margin can buy a violation of a constraint.
  • A hard constraint has no tolerance field. The moment it had one, it would stop being hard.
  • The same data and the same policy always produce the same decision.
  • The language model writes the reasoning; it chooses nothing.
03

Where is what happened recorded, and who can change it?

Nothing is deleted and nothing is rewritten. A corrected record does not replace the previous one: a new one is added and both remain. Months later you can see what data existed the day it was decided, and under which policy. Any alteration made from outside the system is detected.

04

What decides a flock’s destination?

The policy management declared, applied to each live flock’s projection. The result is a day, a weight band, the committed kilos and one sentence of reasoning. It reaches the person who must act through their usual channel, and it is not closed until that person confirms.

  • The day may come from a weighing or a projection, and it always says which.
  • An unconfirmed order is still an open task: there is no “sent” state.
  • What cannot be priced travels without an amount, never as zero.
  • A per-company switch halts all issuing at once.

What does the system cover?

Aerial view of six parallel poultry houses on green hills, with three feed silos at one end and mist in the valleys.
Ten operational modules and a decision layer on top of them. Not ten programs: one single record, which is why a flock’s cost and its feed conversion can be looked at on the same screen.

The chain, link by link

  • Breeders

    Records lay rate and fertility per parent flock, week by week.

    Which house did the egg that is hatching badly today come from?

  • Hatchery

    Tracks every setter load: eggs set, hatch, contamination and the spread of results.

    Was the problem in the hatcher, or did it come from the breeder house?

  • Grow-out

    Daily flock record — weight, intake, mortality, uniformity — with derived indicators computed on that same figure.

    Which day did the gap open, and in which of the twelve houses?

  • Processing

    Birds received and processed, carcass grading and condemnations, tied back to the flock that produced them.

    Why did this shift yield differently on the same raw material?

  • Sales

    Sales orders, collections, cash and credit, with receivables and overdue receivables in plain sight.

    How much of what has already shipped is actually collected?

What cuts across every link

  • Feed

    Inventory by product, supplier purchases and daily intake, with the balance squared.

    It is the flock’s largest cost item. How much came in, how much was eaten, how much is left?

  • Logistics

    Storage and distribution: dispatches with vehicle, driver, destination, on-time performance and returns.

    Did it arrive on time, and if not, where was the time lost?

  • Costing

    Cost centres by concept, with cost per bird, per live kilo and per kilo of carcass.

    What is my real cost per kilo, and what is it made of?

  • Plan versus execution

    Contrasts what was planned against what was executed and computes profitability per flock against the target the company declared.

    Am I going to get where I said I would?

And outward

  • Marketplace

    Sell, buy and service listings between producers in Colombia, Mexico, Peru, Chile and Brazil, each in its own currency.

    Who needs what I have spare, and at what price is it moving?

And on top of the record, the layer that decides

This is what separates the system from record-keeping software with dashboards. The modules store what happened; these screens say what to do about it, under which criterion, and who has to do it.

  • Operating policy

    Management declares once what it is optimising for, what can never be violated, who wins when two things clash, and how much the system decides alone.

  • Planning

    The target is declared per metric and per scope. With no declared target the system does not invent one: it says the target is missing.

  • Forecasting

    Projects weight, feed conversion, mortality and uniformity, always with a band. Stores every projection and records the actual value afterwards.

  • Orders

    Turns the decision into an instruction to a specific person, over WhatsApp, with the reasoning in one line and mandatory confirmation.

  • Audit

    Keeps what data existed, what was projected, under which policy it was decided and what happened afterwards. Nothing is deleted and nothing is rewritten.

  • Alerts and reports

    The anomalous escalates with its context already assembled; the periodic goes out as a PDF with diagnosis and comparison against the industry benchmark.

Biosecurity and sustainability are not modules yet: they exist as health alerts and as indicators inside reports. Said here so nobody counts them twice.

The articulated skeleton of a small feathered dinosaur in a dark vitrine, its wing and tail bones visible.
A laptop against an amber background showing the system dashboard: birds in production, margin, sales and the eight-module summary.
A feathered dinosaur standing in a golden-lit forest, with a toothed snout, curved claws and a stiff tail held out behind.

What does artificial intelligence actually contribute here?

Artificial intelligence applied to a poultry operation is worth exactly what the historical record it works on is worth. Here that record is consolidated — the six links of the chain in a single record — it is 100% reliable because nothing is deleted and nothing is rewritten, and it is contrasted against the formulas and the know-how of the international experts the industry already uses to move money. On that basis the system decides reproducibly, and the language model comes in at the end, to explain the decision in one line.

A hand holds a phone showing the system alerts screen: a health alert, low feed stock and delivery compliance.
The same data under the same policy produces the same decision, every time. The language model comes in after the decision, and only to write the reason in one line.

A 100% reliable record: nothing is deleted and nothing is rewritten

The reliability of a historical record is not a promise, it is a property of how it is built. A corrected record does not replace the previous one: a new one is added and both remain. Every figure carries which device it came from, what time it was captured in the house and what time it reached the server. And any alteration made from outside the system is detected.

It is not a code convention somebody can step around: the database REJECTS UPDATE and DELETE on the evidence tables. And every indicator is contrasted against an independent calculation on every change to the system — today, 102 closed flocks with a mean difference of +0.00 on the efficiency index and 0.0 on the worst flock.

The consolidated history is the raw material

Breeders, hatchery, grow-out, processing, feed and costing in one record means the fertility of a house, the hatch of a load and the feed conversion of the flock that came out of it are the same series. That continuity is what lets you ask why, and not only how much.

It is what turns "which house did this come from" into a query rather than a reconstruction by three people with spreadsheets.

Contrasted against the formulas and know-how of the industry’s international experts

The indicators are not home-made: they are the ones the world poultry industry uses to move money. The European Production Efficiency Factor, feed conversion and the Gompertz growth model, and as the reference for each flock, the performance objectives the breeding company publishes for its line, in the edition the company chooses.

Aviagen, "Ross 308 Performance Objectives" (2022); Cobb-Vantress, "Broiler Performance & Nutrition Supplement" (2022). These are not academic formulas: EPEF is what the industry uses to reward a farm manager or not, and to compare one farm against another, and feed conversion governs the flock’s largest cost item. The edition matters: a 2014 Ross 308 weighs 2,809 g at day 42 and the 2022 one 2,998.

A doubtful figure widens the band; it is not hidden and not rejected

Every record comes in with a reliability mark and, when something does not add up, with the flag that says so: an impossible mortality, a weight that drops with no thinning, intake that does not match the age, a record identical to yesterday’s. The uncertainty propagates into whatever is projected on top.

It is never rejected, and that is deliberate: rejecting field data teaches the operator to lie to the form until the form accepts it, and then the system holds data that is clean and false — the worst possible state.

With enough history the system learns from THAT operation — or refuses to

Past a certain number of closed flocks, the system attempts to fit its projection to that company’s particulars: its heat, its density, its genetics, its start. And it adopts that fit only if it proves it wins.

The gate is two conditions at once, not one: nested cross-validation, and the advantage holding flock by flock. Comparing error alone, pure noise got adopted 45% of the time; with both, 1.3%. When it does not pass it is marked REJECTED and said on screen. It rejects more than it adopts, and that is what prevents "learning" from three flocks.

The language model writes the reasoning. It chooses nothing

The decision is already made by the time the model gets involved, and it gets involved with closed facts in front of it. If the text it returns carries a number that was not among those facts, the whole thing is discarded and the template wording goes out.

The expensive mistake a language model makes here is not the argument: it is the figure, which reads natural and nobody catches. That is why the check is by VALUE and not by text: 3,400 and 3400 are the same number; 3,500 is not.

And what turns up when auditing a complete history is published with its arithmetic

Two biases found while auditing real flock histories, both with measurable impact and both invisible on screen because nothing breaks: they look normal.

One extra day of age — the processing date instead of the day the bird left the house — costs around nine EPEF points over a 39-day cycle. And counting thinning as mortality turns a flock at a real 4.75% into one showing 23.97%.

A system whose decisions cannot be reproduced cannot be audited, and a decision that cannot be audited cannot be defended to a partner, to a bank or to a regulator. That is why the deciding part is deterministic, and why the language model sits where it sits.

See in detail what artificial intelligence does and does not do here

How do you know the numbers are not lying?

The system’s judgement did not come out of a manual. It rests on three fronts. Technical review of the poultry chain: management, nutrition and the indicators that actually move a flock’s result. Contrast against settlements and cost structures from real operations, auditing complete histories to find biases like the thinning one. And producers, in operation and under construction, who recognised the problem the moment it was named. The system already runs on real flocks today. On top of that, every indicator is contrasted automatically against an independent calculation, so a formula cannot drift without anyone noticing.

What this system competes with

Almost nobody starts from zero. The question is not whether this beats nothing, but what it solves that your current setup does not.

What this system competes with
The alternativeWhat it does wellWhat it leaves unsolved
Excel and the farm manager’s experienceIt is flexible, it is free, and whoever runs it knows the operation better than any system.Judgement lives in a person: it is not recorded, it is not comparable between flocks, and it leaves when they do.
The integrator’s systemIt settles, schedules and coordinates the chain. It is the official record of the agreement.It answers the integrator’s questions, not the producer’s. It gives them nothing to check their own settlement against.
Record-keeping software with dashboardsIt stores the figure and shows it in order.It shows what happened. It does not say what to do tomorrow, under which policy, or who has to do it.
The environmental controllerIt governs temperature, ventilation and humidity, and does it better than anything else.It runs one house, not a farm group. It knows nothing of conversion, cost, or which day suits the catch.
Plant automationIt delivers physical precision where it operates: machine vision and automated equipment cut variation and waste in portioning, weighing and packing, and generate data along the way.It optimises a station, not the chain. It measures and does not decide: when it detects a deviation, who decides what to do about it is still a person. And it does not say why this shift yielded differently, because that answer is usually upstream of the plant.
Hiring more supervisionIt works. It is the traditional answer and it delivers.It costs payroll and benefits, it is hard to hire, and it has to be repeated every time a new site opens.

And there is one thing none of the six covers that is already law: Article 18 of Regulation (EC) No 178/2002 requires traceability to "be established at all stages of production, processing and distribution". A per-station record does not meet that; one running from the breeder house to dispatch does.

This system replaces none of the six. It leans on the controller, coexists with the integrator’s system, and gives the farm manager a written criterion rather than taking theirs away. What it adds is the layer none of them covers: turning management’s policy into a daily instruction that is verifiable and comparable between flocks.

Who is behind this

I am Santiago Blandón Ferro, a Colombian architect focused on innovation and technology for agriculture. I understand the physical house and the data layer, and I speak both languages at the same table. I am building a system so that judgement in a poultry operation stops living in a person and starts living in the operation.

More about me

What it does NOT do today

  • It does not optimise a whole complex’s processing plan under plant and transport capacity. That is specified and unbuilt.
  • It does not decide unsupervised: it recommends, and management sets how much it delegates.
  • It does not issue sanitary certificates or sign documents before any authority.
  • It does not replace the vet or the farm manager where professional judgement applies.
  • It does not formulate diets: your nutritionist defines the ration, and the system records what goes into the silo, what the flock eats and what that costs.
  • It does not replace the integrator’s system: it is the producer’s system, and it answers a different question.
  • It does not control the climate: the environmental controller keeps doing its job and the system operates on top.
  • It does not automate the processing line: your portioning, weighing and packing equipment stays yours, and the system operates on top of what that equipment measures.
  • It does not work on bad data: the result depends on the discipline with which things get recorded.

What does this look like on a complete operation?

Forty minutes, over six months of operation with placements, weighings, feed intake, health events and processing. No sales deck: the system is opened and you look at what it recommends and why.

Want to see it on your own flocks?

Leave me your details and pick how to continue: book a date and time, or send me the same information on WhatsApp.

Authorisation for the processing of personal data

Data controller: Santiago Blandón Ferro — Armenia, Quindío, Colombia. By ticking this box you authorise the controller to process the personal data you provide in this form, for these purposes and no others: 1. To contact you and answer your enquiry. 2. To arrange and hold a product demonstration. 3. To send you information about the product and about the published technical material. 4. To keep the record of this authorisation as evidence that you granted it. Your data will not be sold, rented or transferred to third parties for their own purposes. The processing relies on providers that host the website and the database (Vercel, Render and Supabase, in the United States), which act solely on the controller’s instructions and may not use the data for their own purposes. The United States is on the list of countries with an adequate level of protection under Circular 005 of 2017 of the Superintendencia de Industria y Comercio. If you choose to book online, your data is used to create the meeting in the controller’s calendar (Cal.com and Google, acting on the controller’s instructions). If you choose WhatsApp, the message is sent by you, from your own account, and nothing is forwarded on your behalf. Providing this data is voluntary. Sensitive data and data of children or adolescents are not collected, and you are asked not to include any. As the data subject you may: know, update, rectify and delete your data; revoke this authorisation; request proof of the authorisation granted; be informed of the use made of your data; and lodge complaints with the Superintendencia de Industria y Comercio. To exercise them, write to santiagoblandonferro@gmail.com or call 311 663 9606. You may also use SIC Facilita, the Superintendencia’s own platform. Enquiries are answered within 10 business days and complaints within 15 business days, extendable as provided in Articles 14 and 15 of Law 1581 of 2012. Version 1.1.

There is no newsletter, no automated sequence, and your email is not shared with anyone.