Measuring Savings from AI-Enabled Ventilation
A practical way to quantify how heat recovery ventilation plus AI scheduling improves indoor air quality, cuts energy waste, and reduces operational risk across real office portfolios in California.
Start with decision-grade baselines
Before you evaluate any AI scheduling changes, define a baseline that reflects how the building operated immediately before the intervention. For each floor, zone, or building, capture at least:
- Run schedules and setpoints (occupied/unoccupied logic, supply/exhaust targets, economizer mode behavior).
- Measured indoor air quality outcomes (CO2 or other occupancy proxies, differential pressure stability, comfort-related indicators if available).
- Energy consumption drivers (fan runtime, heating/cooling demand proxies, heat recovery utilization estimates).
- Weather context (outdoor temperature, humidity, and any forecast-driven control inputs).
Measure outcomes in two layers
Savings typically come from the interaction between better ventilation timing and more efficient conditioning. To keep results credible, measure both layers and then reconcile them into a single story:
1) IAQ and operating health
- How often IAQ thresholds are met during occupied hours.
- Stability metrics (variance of pressure/airflow setpoints, frequency of corrective overrides).
- Maintenance signals (alerts that precede faults, cycling frequency changes).
2) Energy and recovery efficiency
- Fan energy and conditioning energy impacts tied to occupancy patterns and weather conditions.
- Heat recovery utilization improvements under mixed load conditions.
- Reduced waste from over-ventilation or poorly timed ventilation schedules.
Use a clear event and attribution model
AI-enabled ventilation changes multiple controls, so attribution matters. A simple, reliable approach is to define three measurement windows:
- Pre: the baseline period with identical monitoring and no AI scheduling adjustments.
- Run: the first rollout period while the model learns and schedules are tuned.
- Stabilize: a later period once behavior settles and operational teams stop making manual compensations.
When a change is applied, record the exact logic version and when it goes live. Then report results as both absolute deltas and percent deltas, with at least a short explanation of the largest contributing zones or seasons.
Validate with predictive maintenance signals
Savings that last are savings that avoid downtime. Pair performance measurement with predictive maintenance alerts so you can show that AI scheduling and control refinements do not increase equipment stress. Track:
- Alert counts and severity trends for heat recovery components.
- Any correlating changes in cycling, filter load indicators, or known fault precursors.
- Time-to-repair improvements from earlier detection during peak and shoulder seasons.
Report results in a monthly cadence
Your monthly performance reports should turn analytics into operational actions. Include:
- IAQ outcomes by floor and occupancy band.
- Energy deltas normalized by weather and occupancy where possible.
- Heat recovery performance changes and any recovery-related control adjustments made during the month.
- Maintenance alerts and follow-up outcomes.
If you can’t explain why the number moved, you don’t yet have a reliable measurement framework.
A focused checklist you can reuse
Baseline data quality
Confirm monitoring coverage, time alignment, and control-event timestamps.
Outcome metrics, not just energy
Pair IAQ targets with operational health to prevent “savings at any cost.”
Attribution and seasonality
Separate pre, run, and stabilize windows, and report zone-level drivers.
Next reading
If you’re mapping out the full deployment workflow, these companion topics help you operationalize the framework:
Note: This article explains a measurement framework for building operations. It is not engineering advice and does not replace your local code requirements or a qualified professional’s review.