Artificial intelligence has moved past the pilot stage in education. In 2025 the question schools ask is no longer “should we use AI”, but “where exactly does it save teachers time and help students learn”.

Three places AI makes a visible difference

Across the NineEdu deployments we have run, three problem areas pay off quickly and measurably:

• Marking and feedback on written work. AI drafts the comments, the teacher reviews them. Marking time drops sharply while students still get detailed feedback.
• Early warning for at-risk students. Attendance, grades and classroom interaction combined surface a downward trend before it becomes a problem.
• Personalised revision paths. Each student gets exercises aimed at their actual weak spots instead of one worksheet for the whole class.

What it takes to not fail

Most AI projects in education stall at the demo for the same reason: the data is not clean. Before thinking about models, a school needs student records, timetables and results digitised into one consistent system.

The second factor is teacher trust. AI should show up as an assistant that proposes, with a human approving at the end. When teachers can see that the decision stays theirs, real-world adoption is far higher.

Where to start

Pick one narrow, measurable process and run it for a term with a single year group. Get before-and-after numbers, then expand. On paper it looks slower, but it almost always finishes - whereas rolling out school-wide from day one usually bogs down in data cleanup.