Not necessarily luddites, but lack having a development team available for potential streamlining. Also the Government Department not setting the industry up to allow such streamlining (with a strong asterisk noting that such streamlining may actually be worse for learning outcomes, no matter how much better it would be for freeing up a teacher's time).

This person actually uses ChatGPT to generate report comments "around" the core themes that need to be pointed out in the report. She's said that it's saved a lot of time that would be spent on the "fluff" that wraps around the central message. I think this is one of the perfect use cases for an LLM; "fluff" generation, not the crucial message.

"Machine grading" would be something that the "machine" would have to be tuned for, which is Government Department level jurisdiction, and pretty well outside the skill set of teachers in general (I would think).