Capture got fast. Turning what comes back into linework, breaklines, surfaces and classified deliverables did not, and that step is where production capacity actually gets spent.Two talks on closing that gap. The first covers how two engineering firms rebuilt their extraction workflow around machine learning for survey and design deliverables, including how operators stay in control of the output and what it took to scale the approach across project types. The second looks at AI classification for drone lidar, covering ground, vegetation and utility feature detection, and why point density, trajectory accuracy and collection method still determine how well any of it works. Both are clear that automation does not rescue bad data.
