Fairly significant, apparently - you have to put up a scaffold alongside the wall for SAM to work. (And presumably, you need firm, open ground to put the thing on.) And they can't do corners or complex features, so when you hit those you need to either stop the machine or move it to a different part of the project. Headlines say they're 6x faster than normal masons, but that's during a given span of operation; over the course of a day, with overhead, it's apparently more like 3x.

That said, the machines are in field use, since masons are scarce and expensive. You're right about the big jobs, too; they're apparently pretty much limited to sites like universities and hospitals which have lots of big, featureless walls. The Tech Review article about it is much more informative than the NYT one here. [1]

I think this sort of automation worry seriously over-generalizes from factory robots. Robots are spectacular at doing repetitive tasks with nicely-arranged inputs and fixed workspaces. Hence, Amazon warehouses and Ford factory floors. But (as the SAM people admit) outdoor, new-worksite tasks are going to involve way more mixed human/robot tasks.

[1] https://www.technologyreview.com/s/540916/robots-lay-three-t...