AI Estimating System for a Countertop Fabricator
Estimators were re-keying dimensions off drawings into spreadsheet quote workbooks by hand. This system ingests the drawing directly, extracts the layout, and returns a priced estimate with a confidence score, so the estimator reviews a draft instead of building one from scratch.
Overview
An AI estimating pipeline that turns countertop shop drawings into priced quotes. It supports three modes — direct AI estimation, a traditional geometry pipeline, and a side-by-side comparison mode that runs both so the estimator can see where they disagree.
The Challenge
Quoting required an estimator to read a drawing, measure it, and hand-enter every dimension into an Excel quote workbook. Drawings arrive in wildly inconsistent formats: professional shop drawings, builder plans, retailer layouts, and hand sketches on paper.
The Solution
Drawings are classified by type, then routed through an extraction pipeline that combines OCR with vision-model analysis. Each estimate carries a confidence score; low-confidence jobs retry against a dynamic threshold with exponential backoff, and anything still uncertain is queued for human review in an admin dashboard that tracks per-category accuracy and failed jobs.
Results
More Projects
See more of the work below
Want Something Like This?
Let's talk about your project and how we can help bring it to life.