Gradium is BCU's staff-led marking platform. You read the work and grade it against the rubric. The AI then reviews your decision independently — flagging gaps, drafting developmental feedback, and surfacing where your read diverges from its own. You stay in control of the academic judgement; the AI is a quality-assurance and feedback tool, not a marker.
Sign in to continueTwo pages, one button between them, and a clear audit trail at the end. Every step is captured for internal verification, external moderation, and regulatory defence.
Pick a band per criterion against the rubric. Write your narrative. Compose a summary and overall comment. The AI hasn't seen your work yet — your judgement is yours.
One click runs an independent re-grade against the rubric. The AI reports where it agrees, where it diverges, surfaces evidence from the submission, and drafts developmental feedback. Accept, mould, or override per criterion.
When you mark complete, the grade is committed and routed to the moderator. Every band, narrative, AI verdict and edit is preserved with attribution and timestamps — defensible to IV, EV, OfS and QAA.
Most AI marking tools position the AI as the marker and the academic as the quality check. Gradium inverts that. Same end product — a moderated mark plus developmental feedback — but the human's read is committed first, and the AI never gives the grade.
Your band, your score. The AI's verdict is always a second opinion on a decision you've already made — a senior moderator's read, not a replacement.
Developmental feedback is drafted after you and the AI have settled on the mark, so the wording matches the agreed band. No more feedback that contradicts the score.
Easier ethics committee approval, easier "what if the AI gets it wrong" answer, easier conversation with academics who feel ownership of the marking pen.
Human judgement first, AI second — the order in the audit log matches the order in the room. Every state change, agent run, and edit is captured for IV, EV, OfS and QAA.