The proof
We show our work.
Most construction software sells you a promise. Kanopi's engine was tested across 400+ bids and 4,000+ line items, rates, and regions before it was offered to anyone. Here are the dated, measured results, on real projects, including our own building.
The testing base
400+ bids.
4,000+ line items.
The engine was tested across 400+ bids and 4,000+ line items, rates, and regions. Then we went further: a controlled experiment. 11 complete bids re-priced blind, misses published.
The blind test: the estimator received only the original intake plan set, produced a complete takeoff and estimate without seeing the bid, and only then compared against what was actually sent. Projects spanned ADUs, custom homes, additions, restaurants, commercial shells, a 179,000 SF office interior, a 23,000 SF youth facility, and a 121-unit hotel, at plan stages from concept sketch to permit-issued sets. Method safeguards: a logged blind seal, a visual pass on every plan sheet, explicit denominators on every rate, per-unit math throughout.
What measurement got right
- 121-unit hotel: building SF within 1 percent, 122 individual quantities exact, window count within 4 percent.
- Wall systems: linear footage and coated area within about 2 percent; a foundation extension matched the shipped bid to the dollar.
- Vendor and material pass-through quotes reproduced exactly in every bid that carried them.
- One blind run independently caught a square-footage error our own first bid revision had made and a later revision corrected.
What the totals did
- Uncalibrated bottom-up totals (standard rate library, no per-client tuning) landed a median 35 percent from the shipped number: best 2.5 percent, worst 152 percent.
- Four of eleven fell inside the margin-of-error band declared for their plan stage.
- Every out-of-band miss traced to a pricing convention: scope tier, who supplies material, the FF&E boundary, the markup stack, or comp-priced bids. Zero misses traced to takeoff quantities.
That split is the finding that matters. Two estimators can measure the same building identically and land 50 percent apart, because bidding style is a set of conventions. So no live bid ships until calibration closes the drift on your own history: you give us two or three bids you won, we reproduce one blind, and we reconcile line by line on your rates, your markups, your scope boundaries, until the number lands inside the stage band. A total that lands by compensating errors does not pass.
Read the numbers honestly: n=11 applies to the blind backcast study specifically, run on our own bid history, scored against bids that themselves carry stated margins of error by plan stage. The broader 400+ bid, 4,000+ line-item base is the testing ground the engine grew up on. The blind study demonstrates measurement accuracy and a working calibration method. It does not claim a universal accuracy percentage, and we will not publish one until calibrated live-bid results exist per client.
Kanopi Design, on our own building
Our own building. Drawn by AI, corrected by AI.
We are producing the full, permit-ready Revit model for our own adaptive-reuse project in Midtown Phoenix, a 1971 concrete shell becoming 22 condominiums. The agent team built the structural model and, with no human in the loop, caught and fixed real dimensional errors.
The model
Seven levels, 13 grids, 30 columns, a 22-unit program. A full structural shell, built from the survey and the drawings.
Self-corrected: the bay
The agents caught a north-south bay that was off by 9.5 feet against the field measurement, and fixed it with no human flag.
Self-corrected: the setback
They caught the upper-floor setback the first pass got wrong, and re-cut the floors to the correct envelope.
Projected savings versus commissioning an outside architect for the set: about $250,000. A licensed architect reviews and stamps the final documents.
One model
The design feeds the bid.
Because Kanopi designs and prices from the same model, the Revit geometry flows straight into the estimate, BIM-direct. No re-drawing, no second takeoff, no gap between the plans and the price. It is the same loop, start to finish.
Who built this
Built by an operator who builds.
Kanopi was not built by a software company guessing at construction. It was built by an operator running real projects, used on his own buildings first, and only then offered to others. The proof on this page is the same work we do for clients.