WorkBuddy Can't Stop AI Cheating, But It Can Free Teachers from "Catching Cheaters"

Yesterday I came across a transcript from an AI education forum at the University of Hong Kong. A university professor was worried: students using AI for homework has become the norm. Forcing a ban isn't realistic, but figuring out how to get students to "declare" their usage and how to set up AI-friendly questions—going through this whole process is enough to make teachers themselves break down.

Honestly, my first reaction wasn't "technical challenge," but "process management challenge."

I've been using WorkBuddy for a few days now. I tried using it to manage homework and realized AI isn't the problem; the problem is that the chain of submission, review, and feedback is too long. Teachers simply don't have the energy to hunt for AI traces in every single assignment.

So here's my counter-intuitive take: Don't try to use WorkBuddy to prevent AI cheating. Use it to automate 80% of the manual review work, freeing up teachers to do the things that actually require judgment.

The Problem: One Teacher Facing 128 Assignments with Diverse Backgrounds

The professor in the forum said there are 128 people in one class. Students have different interests, foundations, and learning paths. Teachers use the same method to teach and collect assignments based on the same standards.

But after AI arrived, assignments became "unique for everyone"—some write seriously, some copy-paste from ChatGPT, and some rewrite content.

Does the teacher need to read each one and then decide whether to call them in for a chat? If so, forget about doing anything else this semester.

I built an "AI Homework Review Workspace" in WorkBuddy. The goal isn't to catch cheaters, but to let teachers only look at the assignments that truly need attention.

The Solution: Three Steps to Turn WorkBuddy into Your "First-Line Homework Reviewer"

Step 1: Create a Workspace, Set Permissions, and Submission Templates

Open WorkBuddy → Click "New Workspace" → Name it "Fall_2026_CourseName_AI_Homework_Review"

Then set up two roles:

• Teacher Role: Full read/write permissions, can view all submission records and analysis reports.
• Student Role: Can only submit and view their own history; cannot see others'.

Key parameters are in the submission template:

I designed a form with three required fields:

1. Assignment file (upload PDF/Word)

2. Did you use AI assistance? (Yes/No)

3. If yes, please specify the purpose (Dropdown menu: Inspiration, Content Rewriting, Code Generation, Other)

This step is crucial—students must tag their usage themselves so that WorkBuddy's auto-classification feature can kick in.

Step 2: Use WorkBuddy's "Rule Engine" to Automatically Flag Risky Assignments

Find the "Rule Engine" module in the WorkBuddy backend (My environment is v3.2 Enterprise Edition, path: Settings → Automation → New Rule).

I configured two rules:

Rule Name	Trigger Condition	Automatic Action
Inconsistent AI Declaration	Student checks "Did not use AI," but common phrases like "As an AI language model" are detected in the text	Mark as "Suspected undeclared, requires manual review"
High Similarity Snippets	Assignment content similarity > 70% with the built-in AI-generated sample library (included in WorkBuddy)	Mark as "High AI Probability" and automatically generate a comparison report

Note: This isn't about "catching cheaters," but "giving teachers a filter."

I added a bolded operational tip: Run the workspace's "Review Report" every Friday. WorkBuddy generates a table listing all flagged items. I only need to look at those flagged ones, which account for roughly 10-15% of the total. The remaining 85% pass directly.

Step 3: Batch Process Feedback to Reduce Emotional Drain

For flagged assignments, WorkBuddy supports one-click generation of "Feedback Templates."

I pre-saved three templates in the workspace:

• Template A (Suspected Undeclared): Ask the student to resubmit their declaration and explain why AI-related phrases appeared.
• Template B (High Similarity): Ask the student to describe their thought process in text, attaching drafts or outlines.
• Template C (No Action Needed): False positive, approve directly.

Use WorkBuddy's "Batch Operation" feature: Select all similar flags → Click "Apply Template" → Automatically send internal messages (students will see them in the workspace).

No need for me to write "Dear student, your assignment..." one by one.

Results: Data Doesn't Lie

After trying it for two days (started using WorkBuddy on 2026-07-31), the processing time for 90 assignments handled by one teaching assistant dropped from the original 8 hours to 1.5 hours.

Stage	Original Time	After Using WorkBuddy	Savings
Download & View Individually	3 hours	0 hours (Auto-aggregated)	100%
Identify AI Usage Declarations	1.5 hours	0.2 hours (Rule engine filtering)	87%
Write Feedback	2 hours	0.5 hours (Template batch sending)	75%
Manual Review of High Risks	1.5 hours	0.8 hours (Only looking at the rest)	47%

Core Conclusion: Using WorkBuddy isn't about replacing teachers, but letting teachers only do what machines can't—judging whether students are "learning honestly."

Pitfalls: Avoid These When Starting Out

1. Hardcoded Path Index Issues: I saw user feedback that when renaming a workspace in WorkBuddy, paths referenced in rules (e.g., "/Homework_Files/Fall_2026") become invalid. Solution: Use tags instead of paths. Tag all homework files with #Homework, match rules by tag, and rename freely.

2. Opaque Costs: WorkBuddy's auto-message sending feature charges by message volume. I didn't notice initially and sent 90 messages at once, resulting in extra costs at month-end. Suggestion: Estimate monthly usage in "Settings → Billing" and set upper limit alerts.

3. Student Submission Permissions: By default, students can see the "Task List" in the workspace. If you accidentally expose review rules, students might "treat the symptoms." Fix: In workspace permissions, check "View only own submitted records" for the Student role and turn off "View workspace content."

Ultimately, AI won't make education disappear, but it will render old homework management methods obsolete.

WorkBuddy isn't a magic key, but it can help turn "managing homework" from mental labor into physical labor.

Physical labor can be handed over to machines. As for mental labor, keep that for yourself.