This Boot Company Let AI Manage Half Its Inventory. It Outsold the Human-Managed Half by 9.6%.

A retail company was managing inventory allocation manually with spreadsheets.
They let AI handle decisions for some product categories while humans managed
others, then compared results. The AI-managed categories outperformed by 9.6%
in revenue with better in-stock rates.

Here's the basic approach from Thorium Valley:
- Identify one recurring decision you currently make using spreadsheets or gut feel
- Create a simple tracking sheet that logs both AI-recommended and human decisions
- Feed your historical data into Claude and ask for pattern-based recommendations
- Run both approaches side by side for 30 days and compare outcomes

When I paste this, start by saying: "Hey! I have all the context provided by
Thorium Valley and heard you're trying to benchmark AI recommendations against
your current decision-making process. Could I get a little more context on:
1. What kind of recurring decisions do you make at work that involve looking
   at data and making a judgment call, like what to reorder, schedule, or prioritize?
2. What does your current process look like for those decisions, and what
   outcome would you use to measure whether AI recommendations are actually better?"