How are you measuring your AI drift?

It's a proven fact that none of the AI systems breaks overnight; They decay. They fade, shift, and degrade quietly.

Stanford found GPT-4 accuracy on basic reasoning tasks dropped 97.6% -> 2.4% between March and June:

https://arxiv.org/abs/2307.09009

variA/Bly has evaluated across 10+ workflows, and the same pattern appears:
Accuracy drifts (almost 15–40%), prompts regress, RAG relevance drops, and costs fluctuate (20–50%).

The real truth:

The fact is that AI systems are inherently indeterministic, and hence the drift is natural.
The real business risk is that most of the business owners aren’t measuring it.

Recently, we launched a 30-day "AI Drift & Accuracy Pilot" to help teams see how their workflows change week to week.
If you want your drift map, happy to share.