Summary
In this live-recorded Season 4 premiere of The Blueprint Tour, captured at the IIBEC Metro New York Chapter’s Building Smarter with AI conference, hosts Kenneth Shultz and Carter Huddleston sit down with Jonathan Ehrlich, CEO of T2D2, to unpack one of the biggest questions in the industry: Can AI truly replace engineering judgment?
Jonathan shares how T2D2 leverages computer vision, drone-based reality capture, and AI-powered reporting to automate building enclosure inspections, while also explaining why AI in AEC is fundamentally different from ChatGPT-style text generation. The conversation dives deep into image classification, instance segmentation, bounding boxes vs. crack-level detection, and the reality of training niche datasets in engineering environments.
Together, they explore the “engineer in the loop” model, why AI reviewing drawings isn’t quite there yet, the difference between hype and practical deployment, and how predictive maintenance, from vibration sensors to drone autonomy, fits into the real-world workflow of architects and engineers.
They also examine edge devices, NVIDIA platforms, 2D vs. 3D analysis, LiDAR, photogrammetry, and why data capture strategy may matter more than the model itself.
Technical, candid, and grounded in field experience, this episode separates AI ambition from engineering reality.
“It’s going to get you to 70, 80, maybe 90%, but you’re always going to have to fill in the extra 10-20%.”
Keywords
AI in construction, AEC technology, computer vision, building enclosure inspection, engineering automation, predictive maintenance, drone inspections, T2D2, Jonathan Ehrlich, Kenneth Shultz, Carter Huddleston, PermitZIP, IIBEC Metro New York, NVIDIA AI, edge computing, reality capture, LiDAR, photogrammetry, machine learning in engineering, AI reviewing drawings, engineer in the loop
Takeaways
AI in AEC functions best as a copilot, not a replacement for engineering judgment.
Image detection in niche engineering applications is far more complex than general object recognition.
Training high-quality datasets is one of the biggest bottlenecks in industry-specific AI.
AI-powered reporting can reach 70–90% completion, but engineers still close the gap.
Predictive maintenance requires structured data, not just sensors and optimism.
Reality capture strategy (drones, LiDAR, imagery) directly impacts AI effectiveness.
Edge AI plays a larger role in autonomous navigation than in inspection analysis, today.
“One request I see all the time is: ‘I’d like AI to review my drawings.’ It’s not really there yet.”
Chapters
Welcome from IIBEC: Building Smarter with AI
Meet Jonathan Ehrlich and the Origin of T2D2
Computer Vision in Building Enclosure Inspections
Bounding Boxes vs. Instance Segmentation
Engineer in the Loop: Why AI Needs Oversight
AI Reviewing Drawings, Where It Stands Today
Predictive Maintenance and Sensor Data Reality
Edge Devices, NVIDIA, and AI Deployment
2D vs. 3D Analysis: Point Clouds, LiDAR, and Drones
Data Capture Strategy and Workflow Integration
Final Thoughts: AI as Copilot, Not Replacement
“It’s not just finding cracks, it’s tracing them precisely enough to quantify them.”
