Presented by Ben Felcht, Upswell
CE Pending

Every drawing set ships with errors. What if AI could help architects catch them before they become RFIs and change orders? This session introduces RedlineBench, an open benchmark that scores AI models against a real construction document set with known issues across 7 review categories. Through a live demo, attendees will see what AI catches, what it misses, and where it confidently invents problems that don't exist. Together we'll discuss what these results mean for QA workflows and professional liability, and how architects can take an active role in shaping the standards AI is measured against.