Giving learners feedback on assignments with AI help

Personal feedback on assignments is one of the most valuable things a course or coaching membership can offer. It is also one of the first things to slip when you get busy. Submissions pile up, replies get shorter, and members who waited a week for a comment receive two lines that could apply to anyone.

AI can help you give better feedback faster, as long as you are clear about its role. It drafts; you decide. Used with a good rubric, an assistant can point to specific passages, name strengths and suggest priorities, leaving you to correct, adjust and add the human insight your members are paying for.

Decide what AI is allowed to do

Before you set anything up, write down the boundaries. A sensible starting point for most memberships:

• AI may draft comments against your rubric and highlight passages worth discussing.
• AI may not assign a final grade, pass or fail anyone, or send anything to a learner directly.
• You read every submission yourself and approve every word that goes out.

Tell learners how you work. A short line in your course guide, such as “I use an AI assistant to help organize my feedback, and I read and edit every comment myself,” is honest and rarely raises concern. Our piece on telling members when you use AI covers how to phrase this.

Protect learners' work and privacy

Assignments are personal. They can include a member's business details, their writing, their mistakes and sometimes their life story. Treat them with care:

• Remove names, email addresses and other identifying details before pasting anything in.
• Use a paid business plan rather than a free consumer account, and check that your content is not used for training.
• Read the vendor's terms on how submitted content is stored and handled.

For the practical differences between plans, see free or paid AI plans, and for the wider picture, protecting member privacy when you use AI tools. Privacy rules vary by country, so check what applies to you with a qualified professional.

Build the rubric the AI will use

A rubric is a scoring guide that lists the criteria you judge work on and describes what each level of quality looks like. It is the single biggest factor in whether AI feedback is useful or generic. Without one, the assistant falls back on bland praise and vague advice.

Priya runs a grant-writing course for staff at small nonprofits. Her final assignment is a two-page funding proposal. Her rubric has four criteria: a clear statement of need, outcomes that can be measured, a budget that matches the plan, and a persuasive opening. For each, she describes what “strong,” “developing” and “needs work” look like in a sentence or two. If you have not built one before, rubrics and assessments beyond multiple choice explains the basics.

The feedback prompt

With a rubric in place, the prompt does most of the heavy lifting. ChatGPT, Claude, Gemini and Microsoft Copilot all handle this kind of structured task; try the same anonymized submission in two of them and compare how specific the comments are.

You are helping me draft feedback on a learner's assignment. I will review and edit everything before the learner sees it. Here is the assignment brief: [paste brief]. Here is my rubric: [paste rubric]. Here is the anonymized submission: [paste submission].

For each rubric criterion, suggest the level that best fits and quote the passage that supports your view. Then write: one genuine strength, with a quote; the two changes that would most improve the work, each with a concrete suggestion; and one next step for the learner. Write in a warm, direct tone, addressing the learner as “you.” Do not give an overall grade. If you are unsure about a criterion, say so rather than guessing.

Asking for quotes is the key detail. It forces the draft to point at the actual work, and it lets you check quickly whether the AI read the submission accurately or invented something that sounds plausible.

Your review pass

Read the submission yourself before you read the AI's draft. If you read the draft first, it can quietly shape your opinion. Form your own view, then compare. As you review, look for:

• Quotes that do not exist or are taken out of context.
• Generic praise that could apply to any submission.
• Missed points that you, as the expert, noticed immediately.
• The wrong priorities, such as polishing style when the real problem is the plan.
• Tone that is too harsh for a beginner or too soft for someone preparing for real stakes.

Then add at least one line only you could write: a connection to something the learner said in a live session, a story from your own experience, or encouragement based on how far they have come. That line is what makes the feedback feel like it came from a person who cares.

Keep feedback consistent over time

Save your rubric and prompt together and reuse them unchanged for a whole cohort, so every learner is judged by the same standard. Keep two or three sample submissions with the feedback you finally sent. When you update the prompt, run it on these samples first and check that the drafts still match your judgment. If they drift, adjust the prompt before using it on real learners.

Getting started

1. Write down what AI may and may not do in your feedback process.
2. Add a one-line note about your approach to your course guide.
3. Build or tidy a rubric for one assignment, with a description for each level.
4. Set up a business plan with training turned off and anonymize submissions.
5. Run the feedback prompt on one past submission and compare it with feedback you already gave.
6. Read, correct and personalize every draft before it reaches a learner.