Pennsylvania Angus breeder uses AI to route decisions, not crunch genetics

Deluisio Angus near Leechburg, Pennsylvania, is using artificial intelligence to guide breeding decisions while keeping the actual math inside deterministic tools. The setup highlights a growing concern in livestock genetics: AI can explain numbers, but breeders still need guardrails so the model does not invent them.

Why it matters: - Deluisio Angus is showing how AI can support cattle breeding without taking over the core genetics work. - The approach is aimed at reducing the risk that a language model produces confident but false breeding numbers. - Modern Angus selection depends on expected progeny differences, dollar-value indexes and genomic data that can improve the accuracy of young animals before they produce calves. - The system is designed to help breeders compare matings, manage inventory limits and account for pedigree depth that standard spreadsheets may miss.

What happened: - Cody Deluisio, who owns Deluisio Angus near Leechburg, Pennsylvania, built an AI breeding system that can choose which analysis to run and explain the result in plain language. - The system does not calculate the genetic figures itself. - The operation says the setup uses a plan-execute-narrate sequence. - A reasoning model interprets a breeder's question and selects the right tool. - Deterministic math engines perform the actual computation.

The details: - The system handles mating projections, accuracy-based confidence intervals and inbreeding analysis using Wright's path coefficients. - In one planning run, the inbreeding screen rejected a proposed pairing that reached 13.2 percent. - The same run flagged two additional crosses that shared an unusual number of common ancestors before any semen was purchased. - Other modules weigh correlated traits against one another and rank matings across multiple goals at once. - Deluisio Angus says the tools also factor in limited or sexed-semen inventory. - The operation says the system tracks pedigree depth that standard dollar indexes do not capture. - Deluisio Angus describes its breeding target as producing cattle for both maternal longevity and carcass quality. - Deluisio says genomic information can lift the accuracy of a young sire's EPD on a key trait from roughly 0.15 to 0.45 or higher. - Deluisio said the main concern with AI in breeding is a model that states a number that was never actually calculated. - Deluisio said keeping the math in deterministic engines and using the model only to route and explain it prevents that failure.

Between the lines: - The system reflects a cautious view of AI in agriculture: use it as an interface and decision aid, not as the source of truth for hard numbers. - That distinction matters because breeding decisions can affect genetics, semen purchases and long-term herd value. - The architecture is built to preserve breeder oversight while making complex selection analysis faster to use. - Deluisio has published a technical account of the system's architecture and a separate overview of the breeding goals for readers who want more detail.

What's next: - Deluisio Angus is likely to keep refining the system as more breeding scenarios and constraints are added. - Readers can review Deluisio's technical and breeding-goals writeups for additional background on the methods. - The operation will continue using AI to explain and route breeding analysis while the breeder makes the final call.

The bottom line: - Deluisio Angus is betting that the safest role for AI in cattle genetics is to assist with analysis, not to generate the numbers breeders rely on.