Easing the cost on the education sector with AI powered grading.
Problem
Education providers are under sustained cost and capacity pressure. Grading is one of the biggest hidden costs in online and blended learning; Up Education was spending six figures a year on grading for a single institution, a figure that scales to millions across the group. Tutors and contracted graders were manually working through large volumes of student exams, which was slow, costly and arduous for time poor teachers.
Solution
We developed a custom AI grading tool that is now used across 50% of Swinbourne Open Education. It reads and interprets student answers and generates a summary of each answer along with a confidence rating for the pass mark. As it's a legal requirement that every question is approved by a human, we deliberately built in friction: the tutor has to confirm, or revise the AI's grading summary before approving the grade. The tool integrates directly with Moodle (Up’s Learning Management System) via a custom grading module, so it operates inside the same secure platform staff already know and use.
Result
The impact on grading speed and cost has been significant. Grading can now be done roughly 80% faster. During the pilot phase, one tutor graded all of their students in 2 working days, a task that would normally have taken her 10 working days with contractor support. At scale across a ten-college group, this represents a meaningful productivity and cost saving. Just as importantly, there's been no evidence of hallucinations with the models in use, and the human-in-the-loop review step means the system remains fully auditable by design.
Perhaps most importantly, our custom AI grading tool is giving teachers more time to engage with students on a human level and focus on quality teaching. Up Education now has a scalable, proven foundation for further automation across other academic processes that will help them further combat cost pressures on the sector.
Technologies
