Regrade assignments and update grades in Brightspace
First time I’ve reached this level of confidence in AI grading: The TA had already graded the assignment and had done a good job. Task: The author used OpenAI 'dot' to regrade assignments and identify missed deductions, then verified AI-generated answers using Claude. The workflow involved revising grades and feedback directly in Brightspace. The human role was limited to initiating the request and reviewing the output, as the author noted TAs might no longer grade.; tools: Claude Code. Human role: The original post does not say how a person stayed involved.; limitations: The source does not detail specific integration constraints, API costs, or error handling for Brightspace updates. Reliability is secondary, based on a single user's experience. Human approval is critical before publishing grades, as the author implies a shift away from TA grading without specifying safeguards for student disputes..
Tools
Published on Agent Work Atlas:
Overview
First time I’ve reached this level of confidence in AI grading: The TA had already graded the assignment and had done a good job.
Workflow
The author used OpenAI 'dot' to regrade assignments and identify missed deductions, then verified AI-generated answers using Claude. The workflow involved revising grades and feedback directly in Brightspace. The human role was limited to initiating the request and reviewing the output, as the author noted TAs might no longer grade.
Reported result
The author reported reaching a high level of confidence in the AI grading process. However, no specific metrics, error rates, or time savings were provided. The outcome is anecdotal, based on the author's subjective assessment of the AI's ability to find missed points and verify answers.
Human role
The original post does not say how a person stayed involved.
Limitations
The source does not detail specific integration constraints, API costs, or error handling for Brightspace updates. Reliability is secondary, based on a single user's experience. Human approval is critical before publishing grades, as the author implies a shift away from TA grading without specifying safeguards for student disputes.
Try it
Act as an academic operations assistant. I have a set of student assignments in [replaceable input: file path or LMS export] and existing TA grades in [replaceable input: Brightspace gradebook export]. Use OpenAI to regrade the assignments, identifying any missed deductions or feedback gaps. Then, use Claude to verify the AI-generated answers for accuracy. Prepare a revised gradebook and feedback summary. Do not update Brightspace directly. First, process one sample assignment and show me the proposed grade and feedback for my approval. Only after I confirm the sample is correct, proceed to batch process the remaining assignments. If any assignment has ambiguous criteria, flag it for human review instead of guessing. Output the final results as a CSV file with columns: Student ID, Original Grade, Proposed Grade, Feedback Summary, and Verification Status.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.