Jun 26, 2026
4 min read
Chilled water plants are dynamic systems
Chilled water plants are dynamic systems that need supervisory control, not isolated spreadsheet tuning.
Read field noteField Notes
Practical notes on AI HVAC optimization and building operations.
Field notes for teams evaluating whether AI can safely reduce HVAC energy use in existing buildings, especially central plants already running on BMS infrastructure.
Archive
Latest field notes
Page 1 of 2. 12 posts per page.
Jun 24, 2026
4 min read
Stop paying to retune static HVAC rules
If a building keeps fighting its static BMS rules, AI supervisory control can be a better path than another round of manual retuning.
Read field noteJun 22, 2026
4 min read
A better HVAC optimization pilot reduces procurement risk
A better HVAC optimization pilot should start with one plant, prove the operating evidence, and reduce procurement risk before expansion.
Read field noteJun 19, 2026
5 min read
The real sales objection in HVAC AI is not savings. It is risk transfer.
The real HVAC AI sales objection is whether operational risk is bounded, visible, and accountable before software changes the plant.
Read field noteJun 16, 2026
4 min read
Why ClimaMind will not bypass the BAS
ClimaMind keeps HVAC optimization inside the existing BAS control path, so better setpoint decisions do not bypass the system operators already trust.
Read field noteJun 14, 2026
5 min read
Data centers need cooling control, not just more cooling capacity
Data center cooling optimization should improve cooling operation inside an approved control envelope, so efficiency work does not compromise reliability.
Read field noteJun 12, 2026
5 min read
Shared savings fails when the measurement boundary is vague
Shared-savings HVAC optimization only works when the measurement boundary is defined before control begins, so savings can be tied to approved actions and defensible evidence.
Read field noteJun 10, 2026
5 min read
The hardest gate in HVAC optimization is write permission, not algorithm quality
HVAC optimization stalls when a system can read the BAS but cannot write approved setpoints back. The control loop only closes after bounded write permission is earned inside an operator-approved envelope.
Read field noteJun 6, 2026
5 min read
Human-in-the-loop should not mean reviewing every AI decision
Human-in-the-loop HVAC AI needs accountability for 15-minute control decisions without turning facility teams into the review queue.
Read field noteJun 4, 2026
6 min read
What ClimaMind actually controls
ClimaMind acts as a supervisory optimization layer above the existing BMS. It reads plant conditions, recommends bounded control actions, writes back only to approved points, and preserves the evidence needed to verify savings.
Read field noteJun 2, 2026
5 min read
The hardest part of HVAC optimization is split accountability
HVAC savings stall when asset ownership, utility bills, maintenance, BMS authority, and financial risk sit with different teams. Supervisory control has to be designed for that operating reality.
Read field noteMay 21, 2026
5 min read
Time-series prediction is not end-to-end HVAC AI
Forecasting building behavior is useful, but HVAC optimization is not end-to-end until AI stays in the loop from BMS data to control decisions, approved envelopes, write-back, operator review, and savings evidence.
Read field noteNew notes are added as the archive grows; pagination keeps each index page scannable.