The compliance operator for VET assessment. It rules on two axes — is the evidence the student's own, and does the task assess the work the way it's actually done now — then commits the change end-to-end: the assessment, the Word doc, and the LMS page.
Warrant researches the unit. Ruling rules on the task — and acts on it. The researcher finds where AI has reshaped the work; the operator enforces it, modernises what's obsolete, and ships the fix.
📺 Walkthrough · ~6 min · the full task-to-committed-change run
The usual question is can a student outsource this to AI? — authenticity. But a task can be perfectly authentic and still obsolete: it assesses a pre-AI workflow when the real job is now AI-augmented. An RTO needs both calls. Ruling makes both, on every task, and routes to one decision.
The outsourcing test. Could a student run this through AI, unsupervised, and have it not show?
The modernisation test — acting on Warrant's research. Or is it a pre-AI workflow the industry has left behind?
Ruling never researches the currency question itself — it consumes Warrant's Resist/Integrate split and acts on it. No split, no web search. That firewall is what keeps it an operator, not a researcher.
Authentic and current. → one line naming the resistance mechanism.
A named fix preserves coverage — for authenticity, to modernise, or both. → the exact problem + the exact fix.
The method is structurally wrong for the evidence type. → why rebuild beats revise + the replacement.
A human must decide what the operator structurally cannot. → what + why.
A critical input is missing. → the assumption, the flip — and it still decides.
Asked to execute, Ruling applies the fix to the assessment Markdown (the source of truth), regenerates the Word doc, and regenerates the LMS page — leaving a tracked, reversible change pack (before/after + changelog + diffs). No silent overwrite; an RTO can see and undo every edit.
The task asked students to produce a memo by hand. Warrant's split marks document production INTEGRATE — the real competency is now directing and quality-checking AI output. v1 Ruling would have passed this (it's authentic). The currency axis caught it and modernised it to the judgment triad (Step 3A) — then, because that triad is still outsourceable as unsupervised take-home, Ruling added a verification anchor of its own (Step 3B): it consumed Warrant's verdict, not its method. Committed across all three formats.
1 · THE ASSESSMENT TASK (md + Word doc)
Step 3: Prepare the welcome memo Open the New Starter Welcome Memo Template (RLG-GRP-TMP-002) in your word processor. Prepare a short welcome memo for the three new starters, following the template and Ridgeline's house style. Save the file when you are done.
Step 3A — direct the tool, document judgment Use an AI tool to draft the memo. Then: (a) record your prompt; (b) identify ≥2 problems in the AI draft; (c) correct each and explain why it matters. [Template T2a] Step 3B — verify it's yours (unaided) Submitting 3A unlocks an unseen change to the brief. You record a 2–3 min, one-take, unaided response adapting the memo. [Template T2b] The anchor compensates 3A's take-home outsourcing → authenticity holds at LOW.
2 · THE LEARNER GUIDE, SECTION 2.1 (the taught content) — so what's taught matches what's assessed
2.1 The common business applications [word processor, spreadsheet, email, presentation, database, PDF, cloud storage — what each is for] These are the applications used for organising electronic information in most Australian workplaces. The types are the same even when the brand differs.
... (same application list) ... Working with AI in business software Most of these applications now include an AI assistant — Copilot in Word and Excel, Gemini in Google Workspace. Producing a routine document now means directing an AI tool to draft it, then checking and correcting the result. The skill is your judgment over the output. This is why Assessment 2 asks you to draft the memo with an AI tool and then evaluate and correct what it produces.
3 · THE REGENERATED CANVAS PAGE, RENDERED — the green-outlined block is what Ruling added
Rendered from the regenerated Page07_Section2-1.html in the change pack. A live Canvas URL drops in here once published.
The change pack — authorising ruling, changelog, before/after, diff cards — is committed in the repo at worked-units/BSBTEC201/_ruling-change-packs/.
On CHCDIV001 (Work with diverse people) the split marks every element RESIST — culturally responsive practice isn't AI-augmented. Ruling leaves the AI-free tasks COMPLIANT and does not bolt "use an AI tool" onto interpersonal work. An operator that can't leave a current task alone is as broken as one that can't catch an obsolete one.
Set mode (task/package · advise/execute). Read the Warrant split. Assume what's missing.
Name the task type (T1–T9).
Outsourcing (authenticity) and currency (modernisation), scored together.
Self-undermining instructions? Does the AI policy match the tasks?
Anchored? (package: coverage · chain · ≥1 secure task)
Arbitrate both axes → one decision, one action, every rule ID cited.
Apply md → docx → LMS. Verify each. Write the change pack. Reversible.
ruling/ to a Claude project's knowledge.JUDGE_GUIDE.md — each states the decision it must reach and the rule that forces it.