.webp)
Before Letting Brainbox AI Optimize Anything, Amazon Spent Six Months Finding Internal Owners and Building M&V to Survive Parallel Operational Changes
The Setup Phase is the most commonly overlooked aspect of the HVAC Sequence Optimization Playbook, but Amazon's energy team was intentional about it. Before Brainbox AI touched a single setpoint, Niharika Kishore, Senior Sustainability Specialist at Amazon, and her team spent six months on the parts of the project that aren't technical at all: finding the right internal owners, and designing M&V that wouldn't be destroyed by parallel decisions elsewhere in the company.
At NexusCon 2025, Kishore walked through how that pre-work shaped the eventual deployment. Identifying the right POCs across legal, FM, BAS technical owners, on-site technicians, and M&V owners took six months on its own.
The M&V design was another major workstream the Amazon team wanted to get sorted out before they executed. With so many different teams within Amazon, Kishore was worried that the sustainability team could deliver real savings, but the operations team might add or adjust equipment at the same time, increasing energy consumption and masking the optimization savings.
So Amazon designed the M&V to measure the intervention itself. Brainbox's on/off toggle handles the short-cycle verification — one week with the AI engaged, one week without, control system data pulled straight from the BMS. The formal six-month plan goes deeper, with granular submetering at the equipment level so the savings can be attributed to the AI rather than buried in the utility bill.
Buy vs. build was another consideration the Amazon team had to navigate. Brainbox showed domain expertise in the coupling of HVAC and refrigeration systems: a setpoint tweak on the HVAC side affects refrigeration performance, an interaction that applies across air conditioning and coolers. Leadership's recurring question was "Why can't we build this in-house?" Kishore's answer focused on Brainbox's domain expertise and showing leaders that the pilot delivered what the team promised.
For any enterprise running an optimization pilot at scale, the setup phase is critical: identify the roles, design the M&V, then optimize. Reverse that order and the savings get lost in the noise of everything else the company is doing.
Register for the next Nexus Labs event.
Sign up for the newsletter to get 5 stories like this per week:
The Setup Phase is the most commonly overlooked aspect of the HVAC Sequence Optimization Playbook, but Amazon's energy team was intentional about it. Before Brainbox AI touched a single setpoint, Niharika Kishore, Senior Sustainability Specialist at Amazon, and her team spent six months on the parts of the project that aren't technical at all: finding the right internal owners, and designing M&V that wouldn't be destroyed by parallel decisions elsewhere in the company.
At NexusCon 2025, Kishore walked through how that pre-work shaped the eventual deployment. Identifying the right POCs across legal, FM, BAS technical owners, on-site technicians, and M&V owners took six months on its own.
The M&V design was another major workstream the Amazon team wanted to get sorted out before they executed. With so many different teams within Amazon, Kishore was worried that the sustainability team could deliver real savings, but the operations team might add or adjust equipment at the same time, increasing energy consumption and masking the optimization savings.
So Amazon designed the M&V to measure the intervention itself. Brainbox's on/off toggle handles the short-cycle verification — one week with the AI engaged, one week without, control system data pulled straight from the BMS. The formal six-month plan goes deeper, with granular submetering at the equipment level so the savings can be attributed to the AI rather than buried in the utility bill.
Buy vs. build was another consideration the Amazon team had to navigate. Brainbox showed domain expertise in the coupling of HVAC and refrigeration systems: a setpoint tweak on the HVAC side affects refrigeration performance, an interaction that applies across air conditioning and coolers. Leadership's recurring question was "Why can't we build this in-house?" Kishore's answer focused on Brainbox's domain expertise and showing leaders that the pilot delivered what the team promised.
For any enterprise running an optimization pilot at scale, the setup phase is critical: identify the roles, design the M&V, then optimize. Reverse that order and the savings get lost in the noise of everything else the company is doing.
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.