<p>86% is the classification accuracy of our two AI models reading your Revit model, not a promise that every price is dead-on.</p> <p>Here&#39;s what&#39;s actually happening. When you export a model, the AI reads geometry and metadata to sort line items into the right BOQ categories, framing, MEP, finishes, and so on. 86% of those classifications land correctly without you touching them. The rest get flagged so you can confirm or correct in a couple clicks instead of building the whole takeoff by hand.</p> <p>Pricing accuracy is a separate layer. Once a line item is classified, BidLight pulls current numbers from our own roughly $30,000 pricing database plus Craftsman, 1Build, and RSMeans. That&#39;s what keeps the estimate grounded in real, current costs, and why it updates automatically when the design changes instead of going stale.</p> <p>What that 86% is worth to you in practice: you&#39;re reviewing a small slice of line items instead of building every one from scratch. That&#39;s the time savings, not a claim that you can skip review entirely. Treat it like a strong first pass from a good junior estimator: fast, mostly right, and still worth your eyes before it goes into a bid you have to defend.</p>