AI-assisted development is now the fifth most required skill in Australian technology job ads. Building with language models is a different skill, asked of different people. Counted from employers' own ads, dated 29 October 2026.
November 7, 2026
Contents
Nearly every technology job ad now mentions AI somewhere. That tells you almost nothing.
What is worth knowing is which AI skills employers actually require, as opposed to mention in passing, and who they require them from. We count that, because we read the ads rather than survey anyone.
The answer, at the end of October, is that there are two AI skills and not one. They are asked of different people, and the gap between them is wider than I expected.
Start with where AI actually sits. Across every technology job we read, the skill employers require most often is not technical at all. It is stakeholder communication, at 48% of employers, more than twice the next skill on the list (23%).
Then come the ones you would expect, each at under half that: software engineering fundamentals at 23%, cloud infrastructure at 22%, product thinking at 17%.
And then, fifth, AI-assisted development at 16%, named in 207 ads from 93 employers. Using AI to do the work now outranks continuous delivery at 15% and problem solving at 14%.
For a requirement that was barely in an Australian job ad two years ago, that is a fast climb. But I want to stay on the first number, because in twenty years of hiring it is the one I have seen decide the most outcomes.
In my experience work is about relationships, because it is done with other people. That has been true for a very long time and AI has not changed it.
Working with founders taught me how much of it rests on trust, and how often the default is not to trust. The effort to build it usually has to come from the other side of the table. Communication is the building block for that, whether we like it or not, and when it fails it fails spectacularly.
So when three times more employers ask for someone who can explain their work than ask for any particular technology, that is not softness. That is the market pricing the thing that actually breaks.
The other AI skill in the data is building products out of language models, required by 9% of employers (against 16% for AI-assisted development) and named in 195 ads from 82 employers.
Those two skills sound like the same thing. In the ads they are not, and the people being asked for them are different.
Using AI to do your own work is strongest in engineering. Software engineers 22%, DevOps and platform engineers 22%, product managers 11%, IT support 9%, security 7% (16% across all technology jobs).
Building products with language models turns that around. Product managers 19%, IT architects 18%, software engineers 7%, security 6%, DevOps and platform 6% (9% across all technology jobs).
A product manager is nearly three times more likely than a software engineer to be asked to build with language models, at 19% against 7%.
Engineers are being asked to use the tools. Product managers and architects are being asked to decide what gets built out of them. That is not the split most people assume, and it is not the one I expected either.
The same name tops both lists, and it is a bank.
Commonwealth Bank names AI-assisted development in 20 ads, ahead of Macquarie Group, Mitti and Xero on 12 each, then Quantium on 7 and Heidi Health on 6.
On building with language models it is Commonwealth Bank again on 20, Macquarie Group on 18, then Accenture Australia on 9, Deloitte Australia on 8, and Airwallex and Mantel Group on 6 each.
The first list has product companies in it. The second is banks and consultancies almost all the way down. Read into that what you like, but the organisations with the most regulated data are the ones advertising hardest for people who can build with language models.
There is a debate running through the industry right now that no job ad states plainly, and you should know it is there before you walk into an interview.
Non-technical executives look at AI and see cost reduction, productivity and speed of throughput. Engineers are increasingly rallying around the quality of what comes out: how to use these tools effectively and safely, how to get high quality results rather than spaghetti code, and how to ring fence the business so ideas still ship faster and it still gets the sugar hit AI promises.
Both positions are reasonable. The ad usually only carries the first one.
Underneath that, I see three groups of engineers sitting in quite different places.
Some are in mourning for the craft. That is genuinely unfortunate, because some of them are the best engineers any of us would have the pleasure of working with.
Some have gone headfirst into AI and are charging along quickly.
And there is a group caught in between, rightly pushing back on the tools in plenty of use cases, but sometimes to their own cost, because they are not exploring the places where these tools work extremely well. That last group is the one I would worry about, because from the outside their caution reads as refusal.
If you are an engineer, the ask is that you use the tools well in the work you already do. One in five employers hiring software engineers now requires it (22%, against 16% across all technology jobs), which makes it part of the baseline rather than a specialisation.
If you are a product manager, the ask is changing shape underneath you. Nearly one in five employers hiring product managers wants someone who can build with language models, against one in eleven across technology jobs generally. That is a shift inside the role, not a new job title.
If you are in support or security, the honest answer is that this has barely reached you. IT support 9%, security 7%, against 16% across all technology jobs. There is no evidence in our data that every technology job now requires AI skills, and anyone telling you otherwise is not counting.
The full skill list for each role sits on our pages for software engineers, product managers and DevOps and platform engineers.
We count what an ad requires, not every tool it mentions in passing, so these numbers run lower than ones quoted elsewhere, and the parts about trust and the industry argument are my own read from twenty years of hiring, not the data (how we count).
Hughman, job ads read from 842 Australian employers' own careers sites. Jobs open as at 29 October 2026, skills read to 30 October 2026. Share figures count each employer once, however many ads it posts.
2026 © Hughman.io. All rights reserved.
Reach Out: anthony@hughman.io
