The smart money is on the BMS

The smart money in retail IoT is on the building management system. Not the LLM-powered chatbot, not the computer-vision shelf-scanner, not the predictive demand model. The BMS. The thing that's been quietly running refrigeration cycles and HVAC setbacks for thirty years.

We watched a single retail store's BMS telemetry for seven days. One line on a chart, three peaks. Store opening, refrigeration defrost, HVAC fault. The fault was a real one — a stalled compressor that would have taken the chiller offline by evening trade. The BMS flagged it at 14:08. By 14:35 a technician was on site. No dashboard was opened. No LLM was queried. No predictive model was retrained.

That is what an AI-adjacent system looks like in production: deterministic logic anchored on real telemetry, surfacing the right anomaly at the right moment, integrated into an existing operations muscle that already knows what to do when the alarm rings. The BMS is not a smart-building experiment. It is an instrumentation-grade operational tool that predates the AI hype cycle by two decades.

The LLM projects will spend their first year collecting the telemetry that the BMS has been logging since the day the store opened. The computer-vision systems will spend their second year integrating with the point-of-sale data the BMS already correlates against. The predictive demand models will spend their third year learning what the refrigeration controller has known all along: a defrost cycle has a signature, the signature looks like this, and nothing on the sales floor is going to fix it.

The smart money goes where the signal is. The signal is on the wire the BMS already speaks.