The useful question is not “should we adopt AI.” It is “should this workflow use AI, now.”
A college can be ready for a first-reply draft on web enquiries and completely unready for AI in assessment. Treat each process on its own. If you cannot name the process in one sentence, you are not ready to automate it.
This is a readiness check, not legal advice, and it does not replace your Standards, privacy obligations, or ASQA’s guidance on responsible AI use in VET.
Work through the list. Count the yes answers. Then use the score at the end.
1. Is There a Real Leak, Not a Curiosity?
Yes if: a process is slow, repetitive, and already failing in a way you can see. Unanswered enquiries. Files sitting on missing documents. After-hours voicemail. A weekly report that takes half a day to build.
No if: someone saw a demo and wants the college to “look modern,” or there is no owner for the current process.
If you cannot point to the leak, do not adopt yet. Fix the offer, the form, or the roster first.
2. Can You Describe the Path on One Page?
Yes if: you can map enquiry → next step → who owns it → what “done” looks like. You know where a human waits, chases, or re-types.
No if: the process lives in people’s heads, inboxes, and side-spreadsheets, and two staff would draw two different maps.
AI will copy the mess. Write the path down first.
3. Is This a Draft Job, or a Decide Job?
Yes (safe to consider) if: the system would acknowledge, remind, route, summarise, or prepare a first draft for staff.
No if: the system would decide competence, funding eligibility, “compliance,” visa advice, fee balances, or enrolment status as a fact.
Decide jobs stay with qualified people. If the workflow is a decide job, do not put AI in the decision. Practice tools and admin drafts are a different conversation.
4. Do You Have Facts the System Is Allowed to Use?
Yes if: you have a short, approved set of answers: hours, campuses, how to apply, usual documents, who to email. Someone owns updates when those facts change.
No if: staff would need the model to invent start dates, fees, or policy.
A public model will fill gaps. That is a risk, not a feature.
5. Who Approves What Goes Out?
Yes if: you can name the person (or role) who reviews outbound messages, and you know what may send automatically, if anything. Default is draft and hold.
No if: “the bot will just handle it” and nobody owns exceptions.
If you cannot say who is accountable when the output is wrong, stop.
6. Where Does the Official Record Live?
Yes if: your student management system stays the source of student truth. New tools sit beside it. You are not pasting identity documents or funding files into a consumer chatbot.
No if: the plan is a second unofficial student record inside a chat history.
Access should match who already handles that work.
7. Can Every Relevant Student or Applicant Still Get Through?
Yes if: there is a human path when the tool fails, and you have thought about digital confidence, disability access, language, and accents.
No if: the only way to enquire, practise, or submit is through a system some of your cohort cannot use fairly.
A workflow that locks people out is not ready, even if it is fast for everyone else.
8. Is the First Pilot Small Enough to Reverse?
Yes if: you can trial one channel, one intake, or one campus, measure a few numbers, and switch it off without breaking the college.
No if: the plan is “AI everywhere” in one go, including assessment evidence you cannot later defend.
Name the success measures before you start: time to first useful reply, stalled files, staff hours on chasing. Not “we launched AI.”
9. Are Staff Willing to Use a Named Workflow?
Yes if: reception and admissions know who drafts, who approves, and when to say “I’ll confirm that with a person.”
No if: trainers and admin will quietly keep using personal tools on the side, or will ignore the new path.
An ungoverned shadow process is worse than no AI.
10. Have You Checked the Obvious “Not yet” List?
Do not adopt this workflow yet if any of these are true:
- You do not know which tools are already in use in the college
- Nobody in leadership owns AI
- The website or course information is wrong, so faster replies would spread the wrong facts
- You are trying to replace your student management system
- You want the system to mark assessments or confirm competence
Those are sequencing problems. Solve them, then come back to the checklist.
How to Read Your Answers
Mostly yes on 1–9, and clear on 10: pick one leak and scope a small pilot, with human approval and a written job description for the system.
Mixed: adopt only the slice that scored yes (often acknowledgements and reminders). Leave decide jobs and messy paths alone.
Mostly no: you are not late. You are not ready. Clean the process, the facts, and the ownership. Then run the list again.
If you want a second pair of eyes on a specific workflow, bring the one-page map and this checklist. That is enough to decide go, wait, or not yet, without buying a platform first.
