AI agent for agricultural scientists
Variety Trial Ranking Agent
A consistent, documented variety ranking with uncertainty and a clear note on any site-driven changes.
What it does
A multi-site trial brings data from eight locations in different formats, and each year the ranking method changes slightly. This agent merges the site data, checks for missing or damaged plots, and runs the agreed statistical model every time. It ranks varieties with uncertainty intervals, so a close call looks like one. If a site behaves as an outlier, for example because of a flood or a poor stand, it reruns the model with and without that site and compares the ranking. It lists the varieties whose rank depends on the choice. You approve the ranking before it is shared. Edge case: a variety missing from some sites is ranked on a separate table so it does not distort the others.
How it works
Follow the arrows from top to bottom. The orange dashed arrow is the loop: when a check fails, the agent goes back and tries again.
Read the steps as a list
- Site data submitted
- Merge site files and align variety names
- Are all plots present and valid?If not: list missing or suspect plots and ask the site manager for corrections. Back to step 2.
- Run the agreed model on the combined data
- Rank varieties with confidence intervals
- Is any site an outlier in residuals or interaction?If not: rerun without that site and compare the rankings. Back to step 4.
- List varieties whose rank changes with the site choice
- Write the ranking table and a short note on method and exclusions
- Scientist approves the rankingThe agent waits here for your OK.
- Final ranking and model record
How it decides
It uses the agreed model unchanged. A site is treated as an outlier when its residual variance or variety interaction is far beyond the other sites and then tested by exclusion.
- Plot missing more than 15 percent at a site: ask for explanation before including
- Interval overlap between two varieties: report as a tie
- Site interaction above the agreed threshold: show results with and without it
- Variety not in all sites: rank in a separate table
Make it yours
Every agent is a starting point. You choose these settings for your own situation.
- Statistical model and trait
- Outlier test and threshold
- Tie rule
- Report format
- Sites to include
What keeps you in control
It always asks you first
- Excluding any site
- Release of the ranking
Hard limits
- Never change the agreed model without approval
- Never drop a site silently
It stops when
- Done: ranking approved
- Stop: data too incomplete for the agreed model
Set it up
We guide you through the set-up, step by step
Members get the full set-up guide for this agent. No technical skills needed: you copy, paste and upload.
- One set of instructions to paste into your AI, with the clicks for ChatGPT, Claude, Microsoft 365 Copilot, Gemini and Grok
- The agent then walks you through connecting your own data, one source at a time
- A downloadable copy with the flow chart, the rules and the full guide
