AI insights

Snapshot (last 30 days)

Scans analyzed

4,812

Across all sites and handlers.

Breed match confidence ≥ 95%

78%

Of scans produced high-confidence matches.

Health alerts from scans

39

require follow-up

Animals with repeated issues

11

Candidates for vet review.

Narrative insight (sample)

"Herd activity is concentrated in the morning scan rounds at the north ranch, with a secondary peak around loading times in the yard. Most health-related notes come from a small subset of animals, mainly goats in South paddocks.

Breed identification works well for Holstein and Boer animals, but confidence is lower on local crossbreeds where reference photos are limited. Health alerts tend to cluster around respiratory and lameness observations.

Focusing on better photos during scans and collecting more reference images for crossbreeds would improve model performance and reduce manual follow-up checks."

Suggested focus
  • Capture clearer front and side photos when scanning crossbred animals.
  • Tag scans that include health issues so the model can learn better patterns.
  • Standardize short health notes to make alerts easier to group.

This page is UI-only; in a real setup the AI engine would read scans, animal data, and health notes to generate these insights.

Patterns by segment (sample)

Bullets instead of charts for now.

  • Young stock: more frequent scans around growth checks and vaccinations.
  • High-yield cows: scanned consistently, but few health alerts.
  • Remote partner farms: fewer scans but higher share of flagged notes.
Model feedback ideas

What you might feed back into training later.

  • Collect more labeled examples for local crossbreeds under different lighting conditions.
  • Log when handlers agree or disagree with breed suggestions.
  • Capture structured tags (e.g. coughing, limping) instead of only free text.

All numbers and bullets here are placeholders; this prototype doesn’t run any real AI models.