Every model that understands a jobsite has to learn from somewhere, and the built environment is where that data lives. This panel looks at what is real today: where spatial and physical AI are already producing results on projects, what training data those models actually require, and who is capturing it. From there the conversation opens up. What becomes possible as these models mature? What should firms be doing now so the data they collect is worth something in three years? And where does the person holding the scanner fit in a world where machines can interpret what they capture? A grounded look at the present, and an honest one at what comes next.