Built for A&E and engineering firms
Screening work doesn't bill. Stop doing it by hand.
Your most expensive resource is consumed by preliminary assessments on projects that never get funded. Energy Pilot takes that work to a defensible number in hours and hands your engineers the calibrated model with nothing to rebuild — so the same bench carries more work that actually bills.
The utilization problem
Billable hours lost to early-stage screening.
When a client calls about a retrofit, your engineers spend 40–80 hours on preliminary assessment before anyone knows whether the project will be funded. Most of that work doesn't require detailed engineering — it requires a credible baseline and directional savings, which is exactly what a trained model produces in minutes.
The screening still happens. Your engineers just stop being the ones doing it by hand.
The technical edge
Credible baselines from two months of bills.
You know the problem: a client sends five months of bills and asks for a baseline. Change-point regression needs twelve. Below eight, error degrades exponentially — and at three months the output isn't worth presenting.
Traditional change-point regression
Five-parameter models require 12+ months of continuous data for a reliable baseline. Prediction error degrades exponentially below eight months, and at three months results are essentially unusable.
Energy Pilot AI · ML approach
Quantile regression models trained on 10M+ calibrated simulations. Transfer learning compensates for sparse billing data, and calibrated uncertainty bands narrow as more arrives.
Every forecast ships with an 80% confidence range. Estimated months are labelled as estimates, never quietly filled in.
The handoff
Preliminary work your engineers can actually build on.
The reason screening is expensive isn't the analysis — it's that none of it survives. The spreadsheet gets abandoned, the field notes get retyped, and design development starts from a blank model. Here the preliminary work is the foundation of the detailed work.
On site
The walkthrough stops being a transcription exercise.
Photograph a nameplate and the make, model and serial are read on the spot, then matched against our equipment database for full specifications. What used to be three days of writing up field notes is finished before the drive back.
Built for rigor
The technical credibility your engineers will ask for.
Built by a team with building energy science backgrounds, validated against ASHRAE standards rather than marketing benchmarks. Your engineers should interrogate this section — it's written to survive that.
Calibrated EnergyPlus and DOE-2 runs across building types, climate zones and measures, underpinning the ML surrogate.
Median error against calibrated simulations, inside ASHRAE Guideline 14 tolerances for monthly calibration.
Hourly forecasts capturing seasonal shift, weekly pattern, time-of-use rate structure and demand charge exposure.
10th–90th percentile ranges on every output, narrowing as data improves. Uncertainty is reported, not hidden.
HVAC equipment and controls, lighting, envelope, renewables, storage and demand flexibility.
ML forecasts, equipment engineering calculations and calibrated physics — selected by what the deal needs and what data exists.
The offer
Run it on a project you're already screening.
Pick a building in your pipeline with incomplete utility data — the kind that's been waiting for someone to find a week for it. We'll produce the baseline, the measure analysis and a sample deliverable, and let your engineers judge the quality themselves.