Practical AI capability guide
Which routine administration tasks AI can genuinely help with, which ones must stay human under POPIA and institutional policy, and what to try first.
The route is intentionally practical: diagnose the gap, unlock a useful asset, build one artefact, then move into structured learning, feedback, and portfolio evidence.
Start with work that repeats often enough to justify a reusable prompt, workflow, or review checklist.
University administration runs on documents. Student correspondence, meeting minutes, policy summaries, faculty reports, the same six forms filled forty times during registration. Most of it is routine, none of it is optional, and it all arrives at once.
This page covers which of those tasks AI can genuinely help with, which ones it should not touch, and what to try first. It assumes you already know your job and are short of time.
If you work in a South African institution, the question is not only what AI can do. It is what you are permitted to do with it.
Two constraints shape everything else:
Student personal information. POPIA obligations apply to the student data you handle daily. Putting identifiable student information into a tool your institution has not sanctioned is a governance question, not a personal preference. Check your institutional position before you paste anything.
Institutional policy. Most institutions have an AI policy, or are writing one. It may lag what staff are already doing. Knowing where your institution stands tells you which of the tasks below are open to you now.
Anything you cannot do with de-identified or non-personal information is a conversation with your institution, not a workaround.
These are drafting and summarising tasks. In each case AI produces a first version and you decide whether it is right.
Standing letters, standard responses to common queries, acknowledgements and reminders. Draft from a template and a short instruction, then check it. Useful when the same message goes out repeatedly with small variations.
Long minutes compressed into decisions taken and actions assigned. Faster than rereading them, and it makes the actions visible to people who did not attend.
Turning a policy document into guidance a student can act on. Administration staff do this translation constantly, often under time pressure, and it is work AI genuinely reduces.
Given figures you already have, a first draft of the narrative around them. The figures stay yours. The prose is the part being drafted.
Checking a document for inconsistent dates, names, references and terminology before it reaches someone senior. This is proofreading with fresh attention, not judgement.
The line matters more than the task list, and knowing where it sits is most of the skill.
Anything touching identifiable student information in an unsanctioned tool. This is the first line, and it is not negotiable by you alone.
Decisions affecting a student's status. Admission, exclusion, progression, appeals, funding. AI can draft the letter conveying a decision. It cannot make the decision, and the reasoning must be yours.
Anything you would have to defend. If a document could be challenged, you need to have made every judgement in it yourself.
Final approval. A draft is a draft. Nothing goes out unread.
Pick one recurring document you write from a template. Something routine, with no personal information in it, going to a known audience.
Write the instruction as you would brief a new colleague: who it is for, what it must say, what tone, how long. Read what comes back critically. It will be roughly right in structure and wrong in the specifics only you know.
Edit it. Notice how long that took against writing from scratch. That comparison, on your own work, tells you more than any general claim about AI can.
Then do the same task again next week. The second attempt is where you learn what to ask for.
Someone competent with AI in this role can take a routine drafting task, brief it clearly, judge the output critically, and know without hesitating which tasks are not candidates at all.
That is a working skill, not a qualification. It is built by doing it on real work, and it is closer than most people assume.
If you are unsure whether you are behind, the readiness quiz asks what you have actually done rather than how confident you feel, and points you at a starting place.
Suggested route
Find your institution's position on AI tools and student data before anything else. Start with tasks involving no personal information.
Something you already write from a template, going to a known audience, with no personal information in it. Brief it as you would brief a new colleague, then edit what comes back.
Notice how long editing took against writing from scratch. Do the same task again next week. The second attempt is where the skill forms.
Lead magnet
The AI Skills Quiz sits behind login so the platform can connect your quiz result, saved prompts, weekly challenge work, and upgrade path.
Questions
Not without checking first. POPIA obligations apply to student personal information, and your institution decides which tools are sanctioned. Start with tasks that use no personal information, and treat anything involving student data as a question for your institution rather than a personal decision.
Many are still writing them. Absence of a policy is not permission. Keep to tasks with no personal information, and ask whoever owns information governance where the institution stands before going further.
Routine drafting where you already work from a template and no personal information is involved. Standing correspondence, minutes into actions, and plain-language rewrites of policy are the usual first three.
The tasks described here are drafting and summarising, where a person briefs the work and judges the result. Decisions, judgement and accountability stay with you. We have no basis for predicting anything beyond that, and would not.
myAISkills pathways are about capability you can use at work. We do not claim accredited qualifications or NQF alignment. Check the current pathway details for what is included.