Proprietary Opacity

LinkedIn Post 06: Proprietary Opacity

Your building's AI vendor controls the optimization logic for your HVAC system.

You ask: "How does the model decide to heat or cool my building?"

Vendor response: "The algorithm is proprietary. We can't disclose trade secrets."

You're managing a $20M building. An AI system controls 40% of your operating expenses. You have zero visibility into how decisions are made.

You're managing by faith.

What if the model is optimizing for vendor interests instead of yours? You'd never know.

What if the model contains errors that degrade performance? You couldn't verify it.

What if the model's training data was biased? You couldn't audit it.

Proprietary claims shouldn't exempt vendors from transparency.

Solution: Require vendors to provide:

Feature importance reports (which variables drive decisions)
Training data disclosure (what data periods, weather sources, building variables)
Model explainability reports (how the model weights variables)
Audit rights (third-party evaluation of training and logic)

Other industries (finance, healthcare) require this transparency. Facilities management should too.

Demand explainability. You own the building. You should understand how it's managed.