The AI Talent Squeeze: Why You Can't Find an AI Engineer (and How to Assess One When You Do)

The numbers explain part of it. Indeed Hiring Lab found 6.2 per cent of Australian job ads on Indeed mentioned AI in February 2026, up from 3.3 per cent a year earlier. Its September analysis found the number of distinct job titles carrying "AI" more than tripled in a year, to 71 in the March quarter.

More ads, more AI titles, and more CVs rewritten to match.

The demand is real. A lot of the supply is relabelled.

Why the shortlist feels so thin

Plenty of people have built a demo. The scarce person has taken a model into production, kept it running, and owned it when it went wrong. That's a much smaller pool, and it's the one everyone is fishing in.

The Hays Salary Guide FY26/27 found 82 per cent of employers had experienced skills shortages over the past year, and that's across the whole market, not just tech. For mid-size organisations it bites harder. You're competing with bigger brands for the same people, usually with a smaller team to sell the role.

So the problem isn't volume. It's signal.

Why the CV won't tell you

Hays also asked hirers how they evidence AI capability. Sixty-nine per cent point to portfolios or practical examples. Forty-three per cent say none of the credentials on Hays' list is enough.

I'd agree with both. A certificate tells you someone finished a course. A CV tells you which tools they've heard of. Neither tells you whether they've handled a model that worked beautifully in testing and fell over with real users.

In a field this young, years of experience and brand-name logos are noisy signals too, which I've written about in hiring for a field that's barely three years old.

How to test for real production experience

Ask for the story of one system, end to end. What was the problem, what did they build, who used it, and what is it doing now? People who've shipped talk about users, data and trade-offs. People who haven't talk about frameworks.

Go straight to what broke. Ask about the last time something failed in production and how they found out. Strong candidates answer quickly and specifically, because they've been the one fixing it.

Ask how they knew it worked. Evaluation, monitoring, cost, response times. If there's no answer beyond "the demo went well", that's your answer.

Use a short practical task, not a take-home marathon. An hour working through a real problem from your business, ideally with you in the room, tells you more than a week-long assignment. It also respects the fact that the best candidates have other offers.

Be clear on which engineer you need. An AI engineer, a machine learning engineer and a generative AI engineer overlap, but they're not the same hire. Brief the wrong one and you'll screen the right person out.

Get the offer right before you start

Once you've found someone who has genuinely shipped, you won't be the only one who's noticed. Benchmark before you advertise, not after. I've set out current ranges in my AI engineer salary guide, and it's worth checking before the role goes live.

Then move quickly. A slow, five-round process signals indecision and loses candidates to faster competitors.

A long list of applicants isn't a talent pool. It's a sorting problem.

If you're working through an AI engineer brief at the moment, I'm happy to talk it through. Get in touch here.

‍