Deluisio Angus uses AI to guide Angus breeding decisions

Deluisio Angus near Leechburg, Pennsylvania, says it has built an AI system that pairs a reasoning model with deterministic genetic calculators to plan matings without letting the language model invent numbers. The setup is designed to help owner Cody Deluisio rank sires, screen for inbreeding and allocate semen based on real-time inventory and breeding goals.

Why it matters: - Deluisio Angus is trying to make AI useful for real breeding decisions, not just analysis and paperwork. - The system is designed to reduce bad matings by combining plain-English planning with deterministic genetic math. - The operation says the approach helps balance maternal traits, carcass merit and long-term seedstock value.

What happened: - Deluisio Angus, a registered Angus seedstock operation near Leechburg, Pennsylvania, has put an artificial-intelligence system behind its artificial-insemination program. - Owner Cody Deluisio runs the system using a plan-execute-narrate workflow. - A reasoning model reads a breeder's question, chooses the analysis and then explains the result in plain English. - The language model never computes the numbers itself.

The details: - The deterministic engines handle mating projections, midparent EPDs, dollar values and confidence bands. - The system also calculates coefficients of inbreeding using Wright's path coefficients. - In one planning session, the inbreeding engine rejected a mating at 13.2% and flagged two more crosses with excess common ancestors before semen was ordered. - Advanced modules add covariance-aware weighting for correlated-trait drag and Pareto multi-objective optimization. - The system can chain tasks on its own. - When asked to rank sire options for a cow, the planner first ran a pedigree valuation and then fed that into sire ranking. - Heavy reasoning runs on Claude Opus 4.8 with high reasoning effort and extended thinking. - Routine requests move to a lighter local model. - The system retries or degrades gracefully during transient failures and rebuilds context from a durable log. - The AI setup is inventory-aware. - It tracks semen tank contents, including sexed-female counts. - The system allocates scarce straws so the best cows get the rarest sires first. - In one session, that allocation matched Deluisio's handwritten chute sheet animal for animal. - The system also measures seedstock value that standard dollar indexes may miss, using ancestor influence and cow-family depth to identify females worth breeding for replacement daughters.

Between the lines: - The core design choice is separation of reasoning from calculation. - That reduces the risk of fabricated outputs while still letting AI organize the workflow. - Deluisio is positioning the operation around a hybrid of technical accuracy and breeder judgment. - The emphasis on maternal longevity and carcass merit suggests the AI is being used to reinforce an existing breeding philosophy, not replace it.

What's next: - Deluisio has published a technical account of the system's architecture and breeding philosophy on the Deluisio Angus and Deluisio technology sites. - The operation is likely to keep using the system for sire selection, inbreeding checks and semen allocation as breeding decisions continue.

The bottom line: - Deluisio Angus is betting that AI works best in cattle breeding when it explains decisions, while deterministic engines do the math.