Silent AI Updates
- James W.
- May 5
- 1 min read
LinkedIn Post 02: Silent AI Updates
Your vendor pushed an AI model update yesterday.
You didn't approve it. You weren't notified. You certainly weren't asked to test it.
Today, your building's energy consumption is 15% higher than last week.
When you call the vendor: "Oh, that was a global optimization update. Sometimes individual buildings see different results."
Translation: We changed your building's operational logic without asking. Your utility bills went up. This is normal.
No. It's not.
Good governance means:
14-day advance notification before updates
Pre-update testing in parallel with current model
Baseline comparison ensuring new models don't degrade performance
Automatic rollback rights if performance drops post-update
Vendors are optimizing globally for portfolio-level efficiency. This often masks poor individual building performance.
You need contractual protection against silent updates.
Comments