Ask anyone at a coaching institute where their week goes and they will say teaching. Look at the week properly and teaching is the part that was never the problem. The load is everything wrapped around it: drafting the next test paper, marking the last one, writing notes, rewriting those notes when the syllabus shifts, and entering the same student twice because the admission register and the batch sheet do not speak to each other.
Evaluation is the heaviest single block. A faculty member with 120 answer sheets in front of them loses a weekend, and the part that actually helps a student — why this answer lost four marks and what to do differently — is the first thing dropped when the pile is that tall. Students get a number and no reason for it.
The numbers that moved.
All figures modelled — see the note above.
The list we will not build.
Most people selling AI leave this part out. It is the part that decides whether the whole thing survives its first year.
Partial credit is a judgement about what a student understood, not a string match. A wrong mark is discovered on results day, when it is far too late to fix. Teachers mark these; AI does not touch them.
Low volume, high consequence, and a wrong call changes a child's year. There is no efficiency argument strong enough here.