AI-Assisted Formative Feedback on Student Writing
Cut feedback turnaround by 40–50% while improving specificity — AI drafts, faculty review and approve.
AI-Assisted Formative Feedback on Student Writing
BeginnerTeaching Problem
Providing detailed, individualized feedback on written finance assignments—case analyses, equity research reports, financial plans, investment memos—is the most time-intensive aspect of teaching. In sections of 40 or more, turnaround stretches to two or three weeks, long past the window when feedback drives improvement. Faculty face a painful tradeoff between depth and speed, and students in large sections often receive generic comments that fail to address their specific analytical weaknesses.
How AI Addresses the Problem
AI generates a structured first-pass critique aligned to the instructor’s rubric. Faculty then review, edit, and personalize the AI’s draft feedback—adding the human judgment, encouragement, and domain-specific nuance that AI misses. This workflow cuts turnaround time by roughly 40–50% while maintaining (and often improving) feedback specificity. Students can also use AI for self-directed pre-submission revision, turning in stronger drafts that require less corrective feedback.
Deployment Details
- Courses
- All finance courses with written components (UG and MBA): equity research, case analyses, financial plans, investment memos, executive summaries
- Tools
- ChatGPT, Claude (Projects feature ideal for pre-loaded rubrics and course context), Grammarly, LMS-integrated writing tools
Student Workflow (Self-Directed Revision)
- Complete a full first draft of the assignment independently (stock pitch, case analysis, financial plan).
- Paste the draft into AI along with the instructor-provided rubric. Use the structured feedback prompt below.
- Review AI feedback critically: Which suggestions reflect genuine weaknesses? Which are generic or miss the financial point?
- Revise the draft, accepting some suggestions and rejecting others with clear reasoning.
- Submit: original draft, AI feedback transcript, revised draft, and a 5-sentence reflection memo explaining what you changed, what you rejected, and why.
Sample Prompts
Equity Research Report Feedback
You are a finance professor evaluating an equity research report. Use the following rubric to evaluate this submission. For EACH rubric category, provide: (1) a score estimate on the rubric scale, (2) one specific strength with a direct quote from the submission, (3) one specific weakness with a direct quote, and (4) one concrete, actionable revision suggestion. Be rigorous but constructive. Do not rewrite the student's work-guide them to improve it themselves.
Rubric categories: (A) Investment Thesis Clarity [1-5], (B) Valuation Methodology and Assumptions [1-5], (C) Risk Analysis Completeness [1-5], (D) Use of Evidence and Data [1-5], (E) Writing Quality and Professional Tone [1-5].
[Paste rubric details and student submission below]Financial Plan Review
Review this personal financial plan narrative for completeness and quality. Check for coverage of ALL of the following: (1) clearly stated short-term and long-term goals with dollar amounts and timelines, (2) current income, expenses, and cash flow analysis, (3) debt inventory with interest rates and a prioritized payoff strategy, (4) emergency fund adequacy, (5) risk tolerance assessment and insurance needs, (6) tax considerations relevant to the client's situation, (7) retirement projections with stated assumptions. For each area, rate coverage as Strong / Adequate / Missing and provide one specific suggestion for improvement.Assessment Approach
Grade the revision process, not just the final product. Include a process dimension (15–20% of total grade) evaluating the quality of the student’s reflection on AI feedback: Did they exercise judgment in accepting and rejecting suggestions? Did they identify feedback the AI got wrong? Spot-check AI feedback quality periodically to ensure rubric prompts are generating useful critiques.
Human Credibility Touchpoints
OECD evidence shows that AI and human feedback are not pedagogically interchangeable—students trust human feedback more and respond to it differently. Protect the learning value of this workflow with three touchpoints: (1) the instructor adds at least one comment that only a human who knows the student could make (referencing their improvement arc, career goals, or a class discussion), (2) the student writes one sentence explaining what they would have missed without the AI’s critique, and (3) the instructor spot-checks whether the AI-generated feedback contains any finance-domain errors before returning it to students.
Skills Developed
Finance Concepts
Financial writing craft (research reports, memos, plans), ratio interpretation narrative, valuation storytelling, professional communication for multiple audiences
Analytical Skills
Self-assessment, critical evaluation of external feedback, iterative improvement methodology, distinguishing substantive from cosmetic revision
Professional Skills
Professional written communication, audience adaptation (board vs. lending committee vs. client), editing discipline, revision as a professional practice
AI Literacy Skills
Crafting effective feedback prompts, critically evaluating AI suggestions against domain knowledge, understanding what AI feedback misses
Evidence
Greene (AACSB, 2025) documents structured feedback workflows in finance using the Skills/Replacement/Complement framework. Bowen & Watson (2024, 2nd ed. 2025) provide detailed guidance on the write-first-then-AI-then-reflect pedagogical cycle. Abeysekera (2024) tested ChatGPT on financial accounting assessments, finding GPT-4 scored at the 90th percentile for introductory courses — meaning AI feedback quality benchmarks well against expert human review for foundational content. Broader higher education literature consistently identifies AI feedback as a high-value, low-risk entry point for faculty adoption.
Risks and Safeguards
Depersonalization
Over-reliance on AI feedback may weaken the mentor-student relationship. Mitigation: always add at least one personal, human-only comment per submission.
Finance Domain Blind Spots
AI misses finance-specific errors that sound plausible on the surface. Examples: a terminal growth rate of 4% flagged as "reasonable" without noting it exceeds long-run nominal GDP growth; an equity risk premium of 8% accepted without questioning whether it reflects current market conditions; or a WACC calculated correctly but applied to a project with a very different risk profile than the firm. Mitigation: faculty must review all AI-generated feedback before returning it, particularly the valuation and methodology sections.
Gaming
Students may use AI to generate entire documents from scratch, defeating the learning purpose. Mitigation: require annotated rough drafts, in-class writing samples for comparison, and the reflection memo.
Privacy
Do not paste identifiable student data into non-institutionally-approved tools. Use Claude Projects or Custom GPTs where data stays within the session.
Try it this week
First-Week Implementation Pilot
Pick one assignment you are grading this week. Paste one student’s submission and your rubric into ChatGPT or Claude. Use this prompt: ‘Evaluate this submission using the attached rubric. For each category, provide a score, one strength with a quote, one weakness with a quote, and one actionable suggestion.’ Compare the AI’s draft to what you would have written. Edit it, add your personal touch, and return it. Track your time—most faculty report saving 15–25 minutes per submission on first use.