Grading is the single largest invisible part of teaching. AI is finally good enough to draft rubric-based feedback, but only if you keep the teacher in the loop. Here's the workflow that actually works.
1. Upload the rubric first, not the essays
The biggest mistake teachers make with AI grading is asking for 'feedback on this essay.' Without a rubric, the AI writes generic vibes. With one, every comment is anchored to a criterion you actually teach to.
2. Batch by criterion, not by student
Read all 28 thesis statements in one pass. Then all 28 evidence sections. You'll calibrate faster, and the AI's drafts will land closer to your standard from the start.
3. Approve, don't read line-by-line
TeacherAI shows the AI's score next to the sentence it came from. Skim, approve, override the outliers. The bottom 10% and top 10% are where your time goes. The middle 80% is approve-and-move-on.
4. Send feedback the same day
Same-day feedback doubles its impact. Faster grading isn't really about saving your weekend (though it does). It's about closing the feedback loop while the writing is still alive in students' heads.
AI didn't replace my judgement. It replaced the 4 hours between my judgement and my students reading it.