BreedLogic

Breeding decisions are commercial decisions.

For cattle breeders, choosing the right mating pair is not just a technical exercise. It shapes future herd quality, profitability, trait performance, and long-term breeding direction. The problem is that those decisions are often made through a messy combination of experience, historical records, sale catalogues, breeder instinct, and time-consuming comparison across multiple sources of information.

That works, up to a point.

But it also creates friction, inconsistency, and a heavy cognitive burden, especially when breeders are trying to weigh several traits at once and make sense of increasingly complex genetic data.

This project explored how AI-assisted prediction could help.

Working with a major Australian breed association, geneticists, and data specialists, I helped shape a product direction that used genetic and breeding data to predict likely offspring performance from mating pairs. The goal was not to replace breeder expertise. It was to make complex genetic insight more usable, more comparable, and more actionable inside a real breeding workflow.

Role Lead Product Designer

Capability AI-assisted workflows, Data visualisation, Product strategy

System AI system, Enterprise SaaS

The product opportunity sat in a very specific gap.

Breeders already had access to EBVs and other genetic indicators, but translating that information into a practical mating decision still required a lot of manual effort. Users had to compare sires, think across multiple traits, interpret genetic signals, and imagine how a mating might play out commercially and biologically. That process relied heavily on personal expertise, and while that expertise was valuable, it did not always make the work efficient or consistent.

The DGV concept aimed to change that by using available pedigree, phenotypic, genotypic, and SNP data to provide a more predictive view of likely outcomes. In practical terms, the product needed to help breeders explore which sires might best match their goals and what trait profile a future calf might be expected to show.

What made the challenge interesting was that this was not simply a data problem. It was a trust and usability problem.

1. Discovery & Framing

The first thing this project made clear was that breeders were not looking for abstract “AI.”

They were looking for confidence.

Confidence that a mating choice aligned with their breeding goals. Confidence that a trait profile would likely move in the right direction. Confidence that the numbers were useful enough to inform a real commercial decision.

That meant the design challenge was not to impress users with technical sophistication. It was to turn prediction into something breeders could work with.

Traditional mate selection was powerful, but heavy

Breeders already had processes. Many were deeply knowledgeable, with a strong feel for bloodlines, phenotype, and herd direction. But those decisions could still be time-intensive and mentally demanding. The more variables involved, the more effort it took to compare options, remember past outcomes, and judge which sire might best fit a particular breeding objective.

The difficulty was not a lack of expertise. It was the amount of synthesis required.

Users might want to optimise for:

• carcass quality

• calf weaning weight

• maternal performance

• calving ease

• fertility

• growth

• disease resistance

• overall profitability or index performance

Trying to hold all of that together manually is hard, even for experienced breeders.

The product vision needed to support breeders, not override them

This became one of the most important framing decisions in the work.

There is a lazy version of AI product thinking where the model becomes the answer. That would have been the wrong approach here. Breeders did not want to be told, blindly, what to do. They wanted stronger decision support.

The product therefore needed to:

• surface likely genetic outcomes clearly

• support side-by-side comparison

• allow breeders to adjust priorities and explore alternatives

• make the reasoning legible enough that users could trust it

• fit into the way breeders already think about animals, traits, and herd goals

That is a very different product from “an AI recommendation engine.”

Research needed to stay close to breeder reality

This was not purely a desk-based discovery effort.

Part of the work involved getting out onto farms and seeing how breeding decisions sat in real conditions, with real animals, real operational constraints, and real conversations about herd direction. That context mattered because it kept the product grounded in the lived reality of breeder work rather than drifting into abstract optimisation.

The discovery work also had to connect breeders, geneticists, and technical experts, since the product would only be useful if the prediction logic and the interface supported one another.

2. Designing Trust Into AI Outputs

This was the heart of the UX challenge.

The more technically sophisticated the model became, the more carefully the interface had to work to preserve confidence. Breeders were interested in better prediction, but many were understandably cautious about AI-driven outputs that felt opaque or overly absolute.

That meant the design had to carry a lot of explanatory weight.

Trust depended on comparison and context

One of the strongest insights from research was that users wanted visualisation tools that made comparisons easier. They did not just want a score. They wanted to see how one option stacked up against another and how the predicted profile sat relative to their current herd, trait ranges, or selected criteria.

This made comparison central to the product.

The interface needed to help users move beyond isolated numbers and toward a more structured judgement process:

• what is this sire likely to improve?

• where are the trade-offs?

• how does this compare with current herd performance?

• what sits inside or outside the desired range?

• which pairings look commercially stronger?

That is where the UI started becoming genuinely useful.

Explainability mattered as much as accuracy

Another clear insight was that users needed enough transparency to understand why an output looked the way it did. Even when breeders were open to AI-assisted recommendations, they wanted confidence that the result was grounded in meaningful data rather than hidden black-box logic.

So the UX needed to strike a balance:

• not overwhelming users with technical genetics detail

• not oversimplifying the output into unearned certainty

• making trait signals, ranges, and comparisons legible

• supporting interpretation rather than dumping prediction onto the screen

This is where AI products often fail. They either become too opaque to trust or too reductive to be useful. The design work here was about holding a more credible middle ground.

3. Prototyping The Breeder Workflow

With the product framing in place, the next step was to turn the concept into something breeders could actually interact with.

The prototype work focused on helping users move through a practical decision flow:

• review a herd or animal

• understand key DGV signals

• compare sires against breeding goals

• explore side-by-side differences

• assess where predicted offspring performance might improve

• make a judgement with more clarity and less manual effort

From herd-level insight to individual decision support

Part of what made the product direction compelling was that it could operate at multiple levels. At a herd level, users could see broader patterns and compare outcomes against herd ranges. At an individual animal level, they could dig into more specific performance signals and evaluate likely pairings more closely.

That created a more flexible decision environment. Users were not locked into one path. They could move between overview and detailed comparison depending on what they were trying to decide.

Comparison was where the product started to feel powerful

The comparison experience was especially important because it transformed prediction into choice architecture.

Rather than asking breeders to mentally synthesise trait tables and pedigrees across multiple sources, the product could bring likely outcomes into one visual space. This made differences more visible and helped users reason about which sire best matched the breeding direction they were pursuing.

That was one of the strongest practical value points in the product.

Integrating with breeder workflows and browsing behaviour

The work also explored how these AI-supported insights might sit alongside more familiar breeder behaviour, such as browsing bulls, comparing favourites, and moving between herdbook-style exploration and more advanced analysis.

That mattered because adoption was always going to be stronger if the product met users where they already were, rather than forcing them into an entirely new pattern of work.

4. Outcomes & Reflection

What made this project interesting was not only the AI. It was the way product design could make that AI more usable in a specialist commercial setting.

The prototype established a much stronger direction for how breeders could evaluate mating pairs using predicted outcomes rather than relying entirely on slower, more manual synthesis. It also showed that AI-assisted breeding decisions were more likely to gain trust when they were presented as structured support for breeder expertise, not as a replacement for it.

What this enabled

• a clearer product direction for AI-assisted mate selection

• a more usable way to compare sires and evaluate likely offspring performance

• stronger visual support for balancing multiple breeding traits at once

• a UX model that treated explainability and trust as central to adoption

• a scalable foundation for future genetic and AI-driven breeding tools

Why this mattered

This project sat in a genuinely interesting space: highly technical modelling with very practical on-farm and commercial implications.

That meant the UX work had to do more than make the interface clean. It had to help breeders move from complex data to confident action. It had to bridge genetics, AI, and real-world breeding goals without flattening the nuance that breeders actually care about.

That is where the design value sat.

Reflection

This remains one of the most interesting AI-related projects I have worked on because it was never just about prediction quality in isolation.

It was about how to make advanced prediction usable.

The strongest lesson from the work was that specialist users will engage with AI when it helps them think better, compare more clearly, and decide with more confidence. They do not need the product to pretend the machine knows everything. They need it to help them bring together data, expertise, and judgement in a more workable way.

That is what made this concept compelling.

Not that it replaced breeder intuition, but that it gave it a stronger tool to work with.