Livestock Feedlot Data Management

Replacing a Failing Platform

Ausvet is a recognized leader in veterinary epidemiology and animal health data systems. Their platform, Animal Health Data (AHD), was essential for livestock feedlots to track health and production. But AHD was failing. It was slow, difficult to use, and could not keep up with client needs. Long-term clients worth over $2M annually were walking.

Ausvet decided to rebuild. The initial plan was to work from a long list of feature requests collected over years. This risked repeating past mistakes by focusing on features instead of user needs.

I pushed back. Working from a feature list would build the wrong thing faster. I asked for two weeks to run research, then we'd build from a real strategy instead of a wish list. They agreed.

Discovering the Real Problem

Facing a tight deadline and no prior research, I focused on understanding the people and their workflows. In two weeks, I ran system mapping and stakeholder interviews in parallel. AI tools helped process complex spreadsheets, transcribe interviews, and surface patterns across operational documents.

Validating the Structure

I tested the proposed structure with clickable, grey-box prototypes. The veterinary consultant helped recruit users and run sessions. Key findings included managers thinking in maps not spreadsheets, consultants needing "quiet" alerts that only surfaced when necessary, and terminology being critical to building trust.

Designing for AI-Powered Predictions

The Manager dashboard includes AI predictions. Unlike traditional analytics showing what happened, this predicts what might happen, requiring users to trust and act on uncertain information.

I made the AI's reasoning visible, not hidden. Each prediction shows current metrics, contributing factors, and similar historical cases. The AI provides a confidence level for each prediction, and users can set their own alert threshold, choosing to see only high-confidence predictions or surfacing earlier warnings.

Managers can dismiss predictions that don't match field observations.

This approach turned AI into a decision aid, not a black box. Managers get early warnings while staying in control.

Building Under Pressure

The business needed a working platform in six months to retain major clients. We had just eleven weeks to design four distinct user experiences.

My strategy was to invest upfront in a flexible design system. I used the Ant Design library as a foundation to speed up the creation of our modular component library.

This approach required initial work that didn't look like visible progress, but it was crucial. It ensured visual consistency and allowed engineers to build efficiently. The same core components could be arranged into different dashboards for each role, saving immense time later in the process.

The Platform We Shipped

We delivered a unified system built around three core tools that served every role.

Custom Reporting Engine where users select metrics, choose visualizations, and save their own report templates. This created a flexible analysis tool instead of a library of static reports.

The Results

Four months in, the project moved from a confusing feature list to a complete blueprint for a complex enterprise platform.