.webp)
Amazon is reorganizing facilities work around AI triage after finding that reactive staffing left too little bandwidth to act on stuck dampers, bad discharge temperatures, and water leaks
Amazon is shifting facilities operations toward an AI-assisted model after concluding that reactive staffing was leaving too many building problems untouched until failure. The company described a starting point where “one technician [is] managing 50 buildings,” with HVAC and refrigeration issues often sitting for “24 to 48 hours” before someone gets on site, said Niharika Kishore, Senior Sustainability Specialist at Amazon.
That operating model created a gap between what Amazon’s buildings could see and what its teams could act on. Kishore said much of Amazon now sits in what they call “site operations 2.0”: buildings already have a cloud-connected BMS, but teams don’t have the labor bandwidth to actually use any of that information.
Amazon’s BMS already surfaces common issues like stuck dampers and discharge air temperatures that are off setpoint, but site teams are too bogged down with urgent calls to work those issues unless something has already broken.
Amazon’s is now evolving to “site operations 3.0,” where AI identifies impending failures and cuts tickets before associates are uncomfortable or refrigerated product is lost. Kishore said the company has built in-house AI tools for refrigeration, HVAC, and water, including defrost monitoring, fan cycling, base-building HVAC monitoring, and water-leak indications.
Amazon’s labor problem is a market-wide one: too few people, too many buildings, and more alarms than teams can realistically work. Its “site operations 3.0” model is an early sign that AI may be maturing into a practical triage layer for stretched facilities teams.
Watch the full recording
Register for the next Nexus Labs event
Sign up for the newsletter to get 5 stories like this per week:
Amazon is shifting facilities operations toward an AI-assisted model after concluding that reactive staffing was leaving too many building problems untouched until failure. The company described a starting point where “one technician [is] managing 50 buildings,” with HVAC and refrigeration issues often sitting for “24 to 48 hours” before someone gets on site, said Niharika Kishore, Senior Sustainability Specialist at Amazon.
That operating model created a gap between what Amazon’s buildings could see and what its teams could act on. Kishore said much of Amazon now sits in what they call “site operations 2.0”: buildings already have a cloud-connected BMS, but teams don’t have the labor bandwidth to actually use any of that information.
Amazon’s BMS already surfaces common issues like stuck dampers and discharge air temperatures that are off setpoint, but site teams are too bogged down with urgent calls to work those issues unless something has already broken.
Amazon’s is now evolving to “site operations 3.0,” where AI identifies impending failures and cuts tickets before associates are uncomfortable or refrigerated product is lost. Kishore said the company has built in-house AI tools for refrigeration, HVAC, and water, including defrost monitoring, fan cycling, base-building HVAC monitoring, and water-leak indications.
Amazon’s labor problem is a market-wide one: too few people, too many buildings, and more alarms than teams can realistically work. Its “site operations 3.0” model is an early sign that AI may be maturing into a practical triage layer for stretched facilities teams.
Watch the full recording
Register for the next Nexus Labs event
Sign up for the newsletter to get 5 stories like this per week:
This is a great piece!
I agree.