Part of PoultryPlan
Your planning data can answer questions
PoultryPlan AI is an AI layer inside PoultryPlan. It reads the planning and production data your team already keeps in the system, and helps you work out why a number moved, without exporting anything to a spreadsheet first.
PoultryPlan AI
Ask
Why is expected feed demand higher than planned?
Discover
Three sites moved their schedules forward after a placement change upstream, pulling demand into this week.
Recommend
Re-run the delivery plan for those sites and bring one bulk delivery forward to avoid a stock shortfall on Thursday.
Illustrative example of a PoultryPlan AI interaction. Figures shown are not customer data.
The problem
The number moved. The reason is somewhere else.
A planner opens Monday’s report and sees uniformity down in House 4. The number is right there. The reason is not. It happened three weeks earlier, one step removed from anything the report shows.
None of that is a data problem. PoultryPlan already holds every fact in the chain below. It is a reasoning problem: someone has to walk it backwards, under time pressure, and only the most experienced planners do it quickly.
That walk is what PoultryPlan AI is being built to do.
Today
Where the trail actually starts
1
Three weeks ago
A hatchery batch for House 4 arrives with a wider hatch-weight spread than usual.
2
Two weeks ago
That spread carries into early growth: the lighter birds fall further behind the rest of the flock.
3
Last week
The mid-cycle weigh-in flags the gap, but it’s logged as a routine check, not an alert.
4
This Monday
Uniformity comes in below target for House 4. This is the only step anyone sees without digging.
Being built to
Ask, discover, recommend
Planner asks
Why is uniformity down in House 4 this week?
PoultryPlan AI discovers
A hatchery batch arrived with a wider weight spread three weeks ago than House 4 normally sees. Left alone, the gap between the lightest and heaviest birds keeps widening every week. It won’t close on its own.
PoultryPlan AI recommends
Split the house into two feeding groups by weight band now, before the next grading round, the fastest way to close the gap before slaughter.
Illustrative example. PoultryPlan AI is in development. This shows the kind of question it is being built to answer, not a released feature.
How it works
Four things have to be true before an answer is worth anything
An AI layer on top of a planning system is easy to demonstrate and hard to make useful. These are the conditions we are designing against.
01
It reads what is already there
PoultryPlan AI does not ask your team to log anything new. It works on the flock records, capacity figures, schedules and forecasts your people already keep in PoultryPlan. If a fact is not in the system today, PoultryPlan AI will not know it either. That is a limitation, but it is also the reason an answer can be checked against the same screens your planners already trust.
02
It knows the chain, not just the table
A general assistant can read a column of numbers. What it cannot know is that a placement change at parent stock lands in the hatchery three weeks later, and in the processing plan after that. PoultryPlan AI is being built on PoultryPlan’s model of the chain, so a cause can be followed across stages instead of stopping at the edge of one module.
03
It reasons about the deviation, not the number
The useful question is rarely what feed demand is. It is why feed demand is higher than the plan said it would be. That means comparing actuals against the plan, finding where the two diverged, and ranking the candidates by how much of the gap each one explains, which is exactly the work that takes a planner an afternoon.
04
It ends with a decision, not a dashboard
An explanation that leaves the planner to work out what to do has moved the work, not removed it. PoultryPlan AI is being built to close with a proposal: re-run this plan, move this delivery, flag this site. The planner accepts it, adjusts it, or throws it out.
Worked example
One question, followed all the way through
Rather than a list of things an assistant might say, here is a single question taken from start to finish, including the part most demonstrations leave out, which is what the planner does with the answer.
PoultryPlan AI
The question
Why is expected feed demand higher than planned for week 34?
What it reads
Placement records: 3 sites
Rearing site schedules
Feed forecast versus plan
Delivery plan, week 34
What it finds
1
Farm 3 and Farm 7 brought placement forward by five days after a hatchery reschedule.
2
Both sites therefore entered a higher feed phase earlier than the plan assumed.
3
That accounts for most of the gap. The remainder sits with a delivery that was split on Tuesday.
What it proposes
Re-run the delivery plan for Farm 3 and Farm 7, and bring Thursday’s bulk delivery forward to Wednesday to avoid a shortfall.
What the planner does
Accepts it, changes the date, or rejects it outright. PoultryPlan AI does not move the delivery itself, and it shows which records it read, so the reasoning can be checked rather than taken on faith.
Illustrative example. Figures and site names are not customer data. PoultryPlan AI is in development; this shows the kind of question it is being built to answer, not a released feature.
Comparison
Line the same number up across every site
Not every question is about why something changed. Plenty of them are simply: who is doing better, and by how much? Hatchability across your hatcheries. Laying percentage across your layer houses. Feed conversion across flocks.
Those figures already exist in PoultryPlan for every unit. What costs time is bringing them together: opening each one in turn, writing the numbers down, working out the average, then spotting which site sits below it. PoultryPlan AI returns it as a table.
Hatchability by hatchery
Hatchery
Hatch %
vs avg
Hatchery North
84.2%
+1.1
Hatchery East
83.6%
+0.5
Hatchery South
82.4%
−0.7
Hatchery West
81.9%
−1.2
Average 83.1% across four sites
Laying percentage by house
House
Lay %
vs avg
House 1
93.4%
+1.6
House 2
92.1%
+0.3
House 3
90.8%
−1.0
House 4
90.2%
−1.6
Average 91.8% across four houses
Figures illustrative, not customer data. The same question works for first-week mortality by site, feed conversion by flock, or uniformity by rearing house.
Turnaround
What closing the turnaround gap is worth, across the whole site
Bringing the site’s average down by a couple of days doesn’t feel like much house by house. Multiplied across every broiler house on site, it adds up to real production. PoultryPlan already logs turnaround per house, so PoultryPlan AI can model the site-wide effect of hitting a new average, not just flag one outlier.
PoultryPlan AI
The question
What would a 50-house broiler operation gain by bringing average turnaround down from 12.4 to 10 days?
What it finds
1
Across the 50-house operation, turnaround currently averages 12.4 days between flocks: some houses run faster, some slower, but that is where the site sits today.
2
At a 40-day grow-out, moving the average from 12.4 to 10 days lifts every house from roughly 6.97 cycles a year to 7.3, worth close to 10,000 more birds per house.
3
Multiplied across 50 houses, that shift is worth close to 500,000 more birds placed and grown across the site in the same twelve months.
The numbers
12.4 → 10
days, site average turnaround
+500,000
more birds a year, across 50 houses
50 houses
illustrative broiler operation
What it proposes
Rank every house on site against the 10-day target, and prioritize the ones furthest behind. That is where the average moves fastest.
Illustrative example. Figures assume a 50-house broiler operation, a 40-day grow-out, 30,000 birds per house and a current average turnaround of 12.4 days, and are indicative of the type of comparison PoultryPlan AI is being built to make, not customer data or a guaranteed outcome. Actual gains depend on flock size, grow-out length, house count and site-specific costs.
Why here
A general-purpose assistant cannot answer these questions
Not because the model is not clever enough. Because the questions are not answerable from a spreadsheet, and the context they need does not exist outside the system that recorded it.
The data is already connected
PoultryPlan links parent stock, hatchery, rearing, laying, broilers, feed and processing inside one system. The relationships between stages are recorded, not guessed at. That is what makes it possible to follow a cause from one end of the chain to the other instead of correlating two spreadsheets and hoping.
The domain is built in
Feed phase, uniformity, hatchability, first-week mortality, placement density. These are not generic business metrics, and their meaning does not survive being exported to a generic tool. PoultryPlan’s model already knows what they are and how they relate, so an answer comes back in the vocabulary your planners actually use.
The history is operational
PoultryPlan has been in daily use across the poultry chain since 2008. What it holds is what actually happened on real sites, recorded by the people responsible for it: not a sample, not a survey, and not a demo dataset.
Boundaries
Just as important: what it will not do
An AI layer inside a planning system earns its place by being predictable. The limits below are deliberate design decisions, not gaps we are planning to close later.
It will not act on its own
PoultryPlan AI proposes. A person decides. No plan is re-run, no delivery moved and no schedule shifted unless someone accepts it.
It will not invent a number
Every figure in an answer traces back to a record in PoultryPlan. Where the data does not support a conclusion, the honest answer is that it does not, and that is the answer you will get.
It will not hide its reasoning
An answer arrives with the path it took: which records were read, which findings were ranked highest, and how much of the gap each one explains. A planner should be able to check it, not trust it blindly.
It will not replace your planners
Whether a proposal is right for this site, this customer and this week is judgement, and it stays where it belongs. PoultryPlan AI removes the reconstruction work, not the decision.
Status
Where PoultryPlan AI stands today
In development
PoultryPlan AI is not a released feature, and nothing on this page should be read as one. Every example here describes the kind of question it is being built to answer, using data PoultryPlan already holds. We have published it at this stage on purpose: the questions worth answering first are the ones costing your planners time right now, and we would rather hear those before the work is finished than after.
If you run a poultry operation on PoultryPlan and there is a question you keep having to reconstruct by hand, that is exactly what we want to know about.
Help decide what PoultryPlan AI answers first
Tell us which question costs your planners the most time today.