Work record / Stories

Building the estimating algorithm a decade before AI could do it

Around 2014 I encoded trade pricing by hand so the firm stopped waiting on other people's bids.

Around 2014 I looked at where preconstruction time went, and the answer was waiting. Every trade prices its work off a repeatable metric, whether room count, linear feet or square feet, yet the standard process was to send drawings out and sit for three weeks until someone else's numbers came back. That wait was the largest source of delay we had.

Experience already tells you roughly what a linear foot of a given trade costs, so waiting weeks for confirmation made no sense to me. I encoded those pricing metrics myself, and the firm produced complete estimates in house. We stopped waiting on trade bids for our baseline numbers.

That in-house algorithm is where Kanopi comes from. Kanopi is the estimating engine, AI-driven, that I built over a decade later, and the engine itself is twelve years old. Only the AI is new. The tool people assume I invented in 2026 is a twelve-year-old idea that technology finally caught up with. I never waited on software for a problem I understood well enough to work by hand first.

Story details

All stories · Work record