I Let AI Grade 127 Essays in One Afternoon — Here's What Happened

The Weekend Grading Problem

If you teach English, you know the feeling. Friday afternoon arrives, and you're staring at a stack of 127 Year 10 essays on 'The Merchant of Venice.' Each one needs detailed feedback. Each one represents 15-20 minutes of your time. That's 30+ hours of your weekend gone.

For twelve years, I accepted this as part of the job. Then my school adopted KiwiBee, and I decided to test the AI exam marking feature with a single class. I expected gimmicks. What I got was genuinely useful.

How the AI Actually Works

Students submit handwritten or typed work. The KiwiBee AI scans it (handwriting recognition is shockingly accurate — about 98% in my experience), then analyzes it against your rubric. But here's the key: it doesn't just slap a grade on top. It provides a confidence score and flags anything it's uncertain about for human review.

I reviewed every single AI-graded essay that first time inside the Skills Gradebook. The AI matched my own assessment within half a grade band for 89% of papers. The remaining 11% were borderline cases where I would have agonized anyway. The AI's detailed feedback comments were more consistent than my own — no more 'great point!' on paper 12 becoming 'expand this further' on paper 87 when I'm tired.

What I Do With the Time I Get Back

Last month, I spent that recovered weekend time doing one-on-one writing conferences with struggling students. I made more progress in three hours of targeted conversation than I would have in thirty hours of written comments that students barely read. That's the real revolution here — not replacing teachers, but freeing us to do what we're actually good at.