The Report Card Problem
Australian teachers spend 4.7 hours per week on administrative tasks — the fourth highest in the OECD. A massive chunk of that is report writing.
I've watched teachers spend entire weekends writing individualised comments. Thirty students, two subjects each, 60 comments that need to sound personal, accurate, and parent-ready. The maths is brutal.
So when AI tools became accessible, the first thing every teacher tried was "write me a report comment for a Year 5 student." And the first thing every teacher got back was unusable. Generic praise. No specifics. The kind of comment that could apply to any student in any school in any country.
The problem isn't the AI. It's the input. Most people give AI almost nothing to work with, then blame the output.
Here's the workflow I built and taught to over 50 teachers. It produces 10 individualised, parent-ready comments in about 20 minutes.
Why Most AI Comments Fail
Before the how-to, you need to understand what goes wrong. Most teachers open ChatGPT or Claude and type something like:
Write a report comment for a student in Year 4 Maths who is doing well.
The AI responds with something like: "[Student] has shown great enthusiasm in mathematics this term. They demonstrate a solid understanding of key concepts and participate actively in class discussions."
That could describe any student. It's technically correct and completely useless. A parent reads that and learns nothing about their child.
The AI didn't fail. You gave it nothing specific to work with. No observation notes, no assessment data, no curriculum standard, no tone guide. The AI filled the gaps with generic filler because that's all it could do.
The RCFCC Comment Prompt
This is the exact prompt structure I teach. It uses the RCFCC framework — Role, Context, Format, Constraints, Call to Action. Copy it, modify the details in brackets, and paste it into Claude or ChatGPT.
ROLE: You are an experienced classroom teacher writing end-of-term
report comments for parents. Your tone is warm, professional, and
specific. You write like a teacher who knows this student personally.
CONTEXT:
- School: [Your school name and type, e.g., "international primary school"]
- Subject: [e.g., "Year 4 Mathematics"]
- Curriculum: [e.g., "Australian Curriculum v9" or "IB PYP"]
- Term: [e.g., "Term 1, 2026"]
- Student name: [First name]
- Observation notes: [Paste your raw notes — bullet points are fine]
- Assessment data: [e.g., "Scored 78% on unit test. Struggles with
fractions but strong in multiplication."]
FORMAT:
- 80-120 words per comment
- Start with a strength (specific, not generic)
- Include one area for growth with a concrete next step
- End with an encouraging forward-looking statement
- Use the student's first name naturally (not in every sentence)
CONSTRAINTS:
- Never use phrases like "a pleasure to teach" or "always tries their best"
- Do not invent achievements or behaviours not mentioned in the notes
- Do not use deficit language ("struggles," "fails to," "unable to")
- Use growth-oriented language for areas of development
- Match Australian English spelling conventions
CALL TO ACTION:
Write one report comment for [Student Name] based on the notes above.
Step-by-Step: Your First Batch of Comments
Here's the exact workflow, written for someone who has never used an AI tool.
Step 1: Gather your raw notes.
Before you open any AI tool, pull together your observation notes for each student. These can be messy — bullet points, shorthand, whatever you normally write. The key is specificity. "Good at maths" is useless. "Confidently solves 3-digit multiplication but freezes when fractions appear" gives the AI something real to work with.
If you don't have notes, open your gradebook or assessment records. Pull out two things per student: one strength (with evidence) and one area for growth (with evidence).
Step 2: Open Claude or ChatGPT.
Go to claude.ai and sign in (or create a free account). You can also use chatgpt.com. Both work for this. I prefer Claude for writing quality, but use whatever you have.
Step 3: Paste the RCFCC prompt.
Copy the prompt template above. Paste it into the chat box. Don't hit Enter yet.
Step 4: Fill in the brackets.
Replace every bracketed section with your actual information. The most important field is Observation notes — this is where quality of input determines quality of output. Paste your real notes for the first student. Even rough notes like "Emma — strong reader, needs work on inference questions, loves group work, sometimes rushes written tasks" will produce a dramatically better comment than giving no notes at all.
Step 5: Hit Enter and read the output.
The AI generates a comment. Read it carefully. Does it sound like something you'd actually write? Does it reference specific things from your notes? If it mentions something you didn't include in your notes, that's a red flag — the AI invented it.
Step 6: Generate the next comment.
Here's the efficiency trick. Don't start a new conversation. In the same chat, type:
Next student:
- Name: [Next student's name]
- Notes: [Their observation notes]
- Assessment: [Their data]
The AI already has your role, format, and constraints from the first message. It applies them automatically. Each additional comment takes about 30 seconds to generate.
Step 7: Repeat for all students.
Work through your class list. For a class of 30, generating all comments takes about 15 minutes once you have the prompt set up and your notes ready.
The 3-Step Verification Audit
This is the part most people skip. It's the most important part.
No AI-generated comment should reach a parent without passing through three checks. I teach this as a non-negotiable rule.
Step 1 — FACT CHECK. Read the comment against your notes. Did the AI add anything you didn't provide? Check student names, subject references, assessment details, and any specific claims. If the comment says "Emma excelled in the geometry unit" and your notes don't mention geometry, delete that sentence. The AI made it up.
Step 2 — TONE CHECK. Read the comment out loud. Does it sound like you? AI tends to drift toward corporate language — phrases like "demonstrates proficiency" or "exhibits a strong commitment to" don't sound like a teacher. Rewrite anything that sounds like a LinkedIn post instead of a report card.
Step 3 — BIAS CHECK. Look for assumptions about family structure ("parents" vs "family"), socioeconomic assumptions, cultural assumptions, or deficit language. AI models carry biases from training data. A comment that says "with more support at home, [student] could improve" makes assumptions you may not intend. Rewrite to focus on what happens in the classroom.
This audit takes about 2 minutes per comment. For 30 students, that's an hour of review. Total time: 15 minutes generating + 60 minutes reviewing = 75 minutes for 30 individualised comments. Compare that to a weekend.
Making It Even Faster Next Term
Once you've done this once, you have a reusable system.
Save your prompt. Copy the RCFCC prompt with your school-specific details into a Google Doc. Title it "[Your Name] Report Comment Prompt — [Year/Subject]." Next term, you open the doc, update the term, and you're ready.
Build a Claude Project. If you use Claude Pro ($20/month), you can create a Project. Upload your curriculum documents, school style guide, and previous report samples. Set the RCFCC prompt as the project instructions. Now every conversation in that project starts with full context — you just paste student notes and get comments that match your school's voice.
Create a template spreadsheet. Set up a simple spreadsheet with columns for student name, strength notes, growth notes, and assessment data. Fill it in throughout the term instead of scrambling at report time. When reports are due, copy each row into your AI prompt. The investment is 30 seconds per student per week. The payoff is a 20-minute report writing session instead of a 20-hour one.
What This Doesn't Replace
AI writes the first draft. You write the final version.
The verification audit isn't optional. It's the difference between a tool that helps you and a tool that creates problems. A single hallucinated detail in a report comment — a subject the student doesn't take, an achievement they didn't earn, a name that's wrong — destroys trust with the parent and with your school leadership.
The AI doesn't know your students. It knows what you told it about your students. The quality of every comment is bounded by the quality of your observation notes. That's not a limitation of the tool. That's just good teaching practice showing up in a different format.
I teach this workflow in depth — along with the verification habits, the prompt library, and the full RCFCC framework — in my AI Leadership course. If you want the complete system for integrating AI into your professional practice without cutting corners on quality, that's where to start.
