What careers should our kids go into in 2027 — and is university still relevant?

The most expensive advice in a household right now is panic dressed as strategy.

A 17-year-old is told university is a trap. A 54-year-old is told to “learn AI” as if that were a job title. Both are being sold fluency. Neither is being sold authority.

Fluency is not authority. A model that can write a cover letter does not make you a communications lead. A degree that can be recited does not make your child employable. The question is the same at both ends of the table: what still compounds when the tool gets cheaper?

The 2027 labour market is not “AI or nothing”

Three things are true at once.

First, generative tools have collapsed the cost of first drafts — copy, code, layouts, lesson plans, legal memos. That is not a forecast. That is last year’s invoice.

Second, the bottleneck moved. Organisations still need people who can set the brief, own the risk, and say no. NIST’s AI Risk Management Framework exists because “the model did it” is not a defence. The OECD AI Principles, updated in 2024, still put accountability on human actors, not on the autocomplete.

Third, Australia already had a skills problem before ChatGPT. Care work, the trades, cyber, nursing, teaching, and applied engineering were tight when the hype cycle was still arguing about chatbots. AI did not abolish those jobs. It made the ones that combine judgement with a domain more valuable, and the ones that were only output more fragile.

Careers that still compound

I would not bet a child’s twenties — or your next five years — on a single job title. I would bet on four kinds of work.

1. Domain plus machine, not machine instead of domain.
Nursing, allied health, teaching, electrical, plumbing, construction supervision, agriculture, logistics. The tools will sit in the pocket. The licence, the duty of care, and the 2 a.m. decision stay human. If your child has hands and nerve, do not talk them out of a trade because an uncle saw a demo.

2. Governed technical work.
Cybersecurity, data stewardship, AI operations with audit trails, systems that may not silently fail. Not “prompt engineer.” That title is already a punchline. The hire is the person who can bound what the system is allowed to do, and show the evidence when it did it.

3. Translation jobs.
Product owners, clinical informatics, operations leads, communications people who can brief a model and a board. Over-50 restart lives here. You already know a domain. The missing skill is not Python. It is being able to specify work so a machine cannot wander, and to kill output that sounds right and is wrong.

4. Ownership jobs.
Small firms, practices, agencies, farms, clinics. People who carry P&L, customers, and reputation. AI makes a one-person shop look bigger. It also makes a sloppy shop look sloppy at scale. If you restart after 50, this is often the honest path: not “junior developer reporting to a 28-year-old,” but “I run the work, the machine runs the volume.”

What I would not sell: generic digital marketing intern, unlicensed “AI consultant,” or a bootcamp that promises a salary without a domain. Those are caption factories for CVs.

Is university still relevant?

Yes — as a filter, a network, and a deep-skill workshop. No — as a three-year postponement of contact with reality.

University still earns its keep when the field has a body of knowledge that takes years (medicine, engineering, law, research science, serious design, accounting with actual standards). It still earns its keep when the credential is the licence. It fails when it is bought as a lifestyle and the graduate cannot point to one shippable thing.

For a 17-year-old in 2027 I would ask four questions before HECS:

1. Does this course produce a licence, a studio, a lab, or a clinic — or only essays?
2. Will they leave with a portfolio a stranger can judge in ten minutes?
3. Is there a trade, cadetship, or diploma that gets them to paid work faster with less debt?
4. Can they explain the so what — to a hiring manager, not to a careers fair?

If the answers are thin, university is a very expensive gap year.

TAFE, apprenticeships, and stacked micro-credentials are not the consolation prize. In a lot of households they are the better instrument. The prestige tax is real, and it is paid by the child.

How to steer without steering into a wall

Parents want a list. Children need a stance.

Tell them: the model will be better at first drafts every year. That is not their competition. Their competition is the classmate who can hold a standard when the draft is fluent and false.

Tell them: pick a domain you would still respect at 40. Add the machine on top. Do not pick the machine and hope a domain appears.

Tell yourself: restart is not intern cosplay. It is taking a life of judgement into a stack that finally has labour. You do not need a new personality. You need a permission slip: what you will own, what the machine may draft, and what never ships without your name.

So What Law applies at the kitchen table. If you cannot answer so what about a course, a pivot, or a LinkedIn binge, it does not ship.

This week’s two moves

For you. Write a one-page Restart brief: domain you already own, work you will not do, evidence you can show in 30 days. Not a vision board. A permission slip.

For them. Sit with the child and list three 2027 paths: one licensed, one trade or applied, one degree that produces a portfolio. Price each in time and debt. Pick from a comparison, not from fear.

The new world is not asking you to become young. It is asking you to stay responsible while the tools get loud.

Gail