How educators make the course honest about AI in one working session
“Don’t use AI inappropriately” is a rule nobody can enforce and every student can lawyer around — sitting in a syllabus written as if it’s 2019, next to a flagship assignment that grades the one thing AI produces best. This guide walks the three-step ladder that fixes it: decide your stance once, retrofit the syllabus without rebuilding it, and harden the assignment you’re most worried about.
The whole session starts with one message:
“Here’s my syllabus and the assignment I’m worried about. My stance: AI allowed for specific tasks, with disclosure. Make it enforceable.”
Why this isn’t just a ChatGPT prompt
It refuses vagueness — which is the actual problem. The policy interview pushes back until consequences are something a student can predict, and every rule survives the test “could I point to this in a conversation with a student?” Generic prompts hand you plausible-sounding mush.
It diagnoses component by component, not vibes. The assignment analysis rates each part — what AI can do to it right now, and how fast. In our documented run, that granularity found a vulnerability the essay-panic think-pieces miss entirely.
It changes only what’s broken. The syllabus audit marks each section KEEP, TWEAK, REWRITE, or ADD, and shows you findings before touching anything. Your voice, your structure, and your institution’s locked language stay intact.
It produces the chair-facing paper trail. Every change lands in a prioritized log — original text, revised text, one-sentence defensible reason. Written so someone who never saw the original understands each change.
Your stance is respected absolutely — including a full ban. Ban, allow-with-disclosure, require: all valid. The tools make whatever you choose specific and enforceable; they never argue you into more AI.
The first step is free. The Classroom AI Policy Writer is free tier — the ladder starts at zero cost.
The steps
1.Step 1 — decide once: the policy (free)Free
Run the Classroom AI Policy Writer. It interviews you — where AI helps in your course, where it undermines learning, what happens on a first vs. repeated violation, what tone fits you — and outputs a one-page policy plus a student FAQ that answers the questions they’ll actually ask (“does Grammarly count?”). If rules differ per assignment, it builds a GREEN/YELLOW/RED tag system.
Say this
“Find the Classroom AI Policy Writer on Amplifiers and run it for my course.”
2.Step 2 — retrofit the syllabusPro
Run Syllabus Refresh and paste your syllabus. It audits section by section, shows you the findings, and — after your go-ahead — rewrites only what needs it: AI-completable outcomes, unenforceable integrity language, the missing “how AI fits” section. You get the retrofit with changes marked and the change log for your chair.
Say this
“Run Syllabus Refresh on this: [paste syllabus].”
3.Step 3 — harden the assignmentPro
Run the Assignment Strength Builder on your worried-about assignment. It maps where AI can shortcut each component and how fast, then rebuilds the assignment and rubric within your real constraints — class size, no TA, no spare class time. At least half the new rubric grades process and judgment.
Say this
“Run the Assignment Strength Builder on this assignment: [paste].”
4.Step 4 — the part that’s still teaching
Chair sign-off (the change log is built for that conversation), LMS upload, and ten first-day minutes walking the class through the policy. The credibility in the room is yours.
What it looked like when we ran it
We ran the full ladder on a typical undergrad research-methods course and published the outputs. The sharpest finding wasn’t about essays:
The real vulnerability was data fabrication, not prose
The classic survey assignment — collect 30 responses, write a report. An AI can’t field a survey, but it can fabricate 30 plausible respondents in one prompt, and nothing in the original assignment made fabrication harder than collection.
The redesign closed it without busywork
Raw response export plus a recruitment log, a decision log for design choices, and one genuinely un-fakeable component: students mark three claims an AI-drafted analysis of their own data got wrong — and how they know.
The syllabus retrofit touched less than a third of the document
Most sections came back KEEP or TWEAK. The two rewrites that mattered: an outcome AI can trivially complete, and an integrity section that never mentioned AI.
The educator and course materials in the sample are synthetic and disclosed as such; the diagnoses, redesigns, and policy text are the tools’ real output from the run.
Common questions
I want to ban AI entirely. Does this still work?
Yes. A ban is a valid policy — the tools just make it specific and enforceable. They respect your stance absolutely and only refuse vagueness.
Do I need to be an AI person?
No. You bring the subject expertise and the stance; the tools handle drafting, enforceability language, and the shortcut analysis. Several are built for teachers who don’t use AI themselves yet.
What will my department chair see?
A change log: every modification with original text, revised text, and a one-sentence defensible reason, prioritized as must / should / could.
What does it cost to start?
The policy step is free tier. The syllabus retrofit and assignment redesign are Pro.
Run the ladder before the semester starts
The policy writer is free; Syllabus Refresh and the Assignment Strength Builder are in Amplifiers Pro.
Try Amplifiers Pro todayWorks inside Claude · setup takes minutes · cancel anytime