TECH TEAMS TUESDAY · A Building Tech Teams publication
How Australia hires, builds, and leads tech teams in the age of AI.
01 · THE PATTERN
I spent Friday morning in a workshop with an industrial manufacturer, they have seven factories, hundreds of people, and we were working through where AI fits in how they design and build. The proposed solutions are still being workshopped, but their case is very similar to a lot of other companies having the same conversation at the moment around “we need AI” or “what can we do with AI”. More often than not the answer really is not AI, it is automation. Capture the paperwork digitally. Route the feedback to the right person. Make the data accessible. Track the rework. An LLM shows up in a small percentage of it.
This same week, a brief landed on our desk for a forward-deployed engineer, the new hot job title where this engineer is being asked to embed AI (or automation) into an organisation. Lenny released his annual survey and one of his summaries says “Trying to use AI for everything is how you end up overwhelmed and conflicted, not empowered”. And Anthropic published a number that should have been front-page news here: Australia is now the biggest Claude user in the world, per capita, out of 121 countries. More than six times what our population predicts.
So here is my take for the week, and it is the thread through everything below. Strip the AI branding off most of the success stories and you find automation. For the majority, stop with trying to be AI everything, fully agentic, AI native. The companies getting real productivity and real ROI are mostly automating tasks, with a model in the loop for a minority of the work. That is not a lesser version of the AI story. It is the actual AI story I am seeing and hearing across industries.
02 · THE MOVES
1. The fear is not being replaced, it is being squeezed
Lenny Rachitsky’s annual survey of 5,920 tech workers found “One half feels amplified by AI—more capable, more confident, more excited than they’ve been in their entire career. The other half feels shaken by it—less sure of their value and whether there’s still a place for them”. (Lenny’s Newsletter, 7 July)
The two numbers that stood out to me why it came to fears: only 22 percent of workers feared losing their jobs to AI, but 51 percent fear being expected to do more for the same pay. Overlay that with the 55.7 percent affected by burnout and you have the theme: workers are worried about being asked to do more with AI and becoming burned out.
Personally I understand the dilemma. Since I’ve gone down the rabbit hole with AI I’m now capable of significantly more, but I’m also spending more and more time on the computer talking with Claude Cowork, running multiple terminals, asking more of myself. Late nights, overwhelm of new things to do. I think this burnout theme may become more prevalent.
For leaders: your team’s AI anxiety is not a technology problem, it is a workload problem you control. If AI made everyone faster and the company roadmap just grew to absorb it, you’re making the gains and the employees have carried the cost, and they know it.
For tech professionals: the data shows the professionals who have used AI to amplify them make a common first move: they found two or three tasks that measurably changed their output and got very good at those. They used AI to make their lives easier, and the people I’ve spoken to who are enjoying the benefits of AI are the ones automating the mundane tasks while getting to do more of the things they enjoy, or the higher value tasks.
2. Australia is quietly the heaviest AI user on earth. Look at what we actually do with it
Anthropic’s Economic Index put Australia first out of 121 countries for Claude usage, population adjusted. (SmartCompany, 7 July, on the Anthropic report)
The detail is actually more interesting than the headline. Australians use Claude less for coding than the global average and more for the ordinary machinery of work: business operations, workplace writing, admin, management tasks. And 54.5 percent of Australian usage is what Anthropic classifies as augmentation, human in the loop, above the global average. We are not the world’s number one AI engineering country. We are the world’s biggest workflow-automation-with-a-human-attached country. As a recruiter who reads position descriptions all day, that matches exactly what I see: AI is being folded into existing jobs, not creating AI-specific new ones.
It was the same story as my Friday workshop. Improving these processes does not necessarily require traditional engineering. It requires people who can use the new tooling to document and automate workflows, remove admin, knock over the manageable repeatable tasks.
For leaders: your people are already automating their work, the only question is whether you can see it. A third of workers admit hiding their AI use from their employer (Employment Hero, July). The fix is not more surveillance, it is permission: the companies where leaders use AI openly and have clear AI policies are the ones who reduce shadow usage.
For tech professionals: lean into the trend. Coding is only around one in seven Australian Claude requests, the rest is the everyday work. Whether you are in product, design or project management, the tools are available to make you more efficient, and the power users are the ones moving ahead. The person who documents and automates the workflow ends up owning the workflow.
3. The companies investing heavily in AI are growing headcount faster
Revelio Labs matched Ramp corporate card and bill-pay records to workforce data across more than 21,000 US companies. Instead of asking companies whether they use AI, it watched who actually pays for it, then watched what happened to their headcount. (Revelio Labs, 30 June, with Ramp)
The finding: companies making the deepest AI investments grew employment by roughly 10 percent more over the two years after adoption than comparable companies that had not adopted yet. Light adopters showed no measurable change at all. The researchers are honest about cause and effect, committed companies were already growing faster, so nobody should read this as a definite “buy AI, get headcount”. But the theme here matches what I see when recruiting for teams. The forward-deployed engineer brief we received this week exists because a company committed to rolling out AI. The companies dabbling with a licence and a pilot are not making significant changes.
Rather than re-architecting your entire organisation, make a significant commitment of time and budget to automation and AI, move the needle, provide the ROI, then take on more. That’s the theme I’m seeing, and the theme of the Revelio data with companies that are growing.
For leaders: the majority of projects delivering ROI right now are workflow automation with a model doing a minority of the lifting, and when you scope them that way they ship. The flashy “we’re going truly agentic” is for a very small number of companies. Most would benefit from doing something, proving value, then doing more.
For tech professionals: when you are weighing up an employer, the useful question is not whether they use AI, everyone says yes. It is how deep the commitment goes: what they have actually rebuilt around it, what they spend, who owns it. The heavy adopters are where the hiring and growth is.
03 · FROM THE INDEX
One read a week from the AI Jobs Index, the live tracker I run at aijobsindex.com.au.
This week the automation story turned up on the AI Jobs Index’s radar: the Finance Sector Union went public with CBA proposing to cut 176 technology and engineering roles, and the bank’s stated reasons led with “workflow automation”.
This same week, CBA shared that, following the introduction of its AI orchestration agent, 84.6 percent of self-service messaging interactions were resolved end to end within the messaging channel, and the platform is planned to expand into other areas of the bank. Cuts attributed to workflow automation on one news day, automation doing measurable work on another news day. That is the whole newsletter thesis in one company. (iTnews, 10 July)
The Index has logged 3,610 Australian jobs cut in announcements where the company itself named AI, and CBA’s 176 will be assessed for the ledger this week. My gut feel is that it won’t qualify based on the data we have, given it’s not a CBA statement attributing job cuts to AI, but that may change with more information.
04 · TEAM MOVE OF THE WEEK
Indara appointed its first ever CTO. Until now, technology sat across the whole leadership team.
Milinda Wijesuriya steps into the newly created role as part of a wider executive restructure under CEO Emilio Romeo. Indara runs about 4,700 towers and poles behind Australia’s mobile networks, and since forming in 2022 it has run technology as a shared responsibility across multiple business functions. No one person owned it. Now someone does. (iTnews, 6 July)
There’s a pattern forming in this section. Issue #1’s team move was CBA splitting AI leadership into two senior seats, an Officer for strategy and adoption, a Scientist for research, secruity and responsible AI. Now an infrastructure company that got by for four years with technology as a shared responsibility has made it one person’s primary responsibility. Technology responsibility is moving out of the shared-across-many-leaders model and into dedicated, specialised seats.
The detail worth noticing: Wijesuriya's twenty-plus years are all one industry. Nokia radio networks, a decade of NBN Co delivery, then the sales side at Indara. It's the argument Nathan makes on this morning's podcast episode: you can't take 20 years of experience, grab a dev, and say now you know telco. Indara's first CTO is a bet on industry depth, and the roadmap job going to the person who knows the industry, not the deepest specialist, tells you what they think the hard part is.
05 · WHAT I’M BUILDING
A confession that proves this issue’s point. The chief of staff I have been building, an agent that runs my mornings, is mostly not AI, it’s automation.
Reading my calendar, my inbox, my Slack and my task list every weekday at 7am: automation. Filing tasks with who asked and when: automation. Counting the days since I committed to something and putting the number in front of me every morning: automation, and confronting (it turns out I am excellent at committing to things on a Friday and rediscovering them the following Friday). The AI earns its spot as a layer on top: deciding what actually matters today, catching contradictions and executing on the tasks it can. Last week two of my scheduled tasks reported different versions of the same jobs number and the chief of staff flagged it before that number went into code.
That ratio, a lot of automation carrying a little judgment, is exactly what the Anthropic usage data was describing. My current take is the tools were the easy part. Deciding what deserves my hours is the actual job, and that call is still mine every morning. The machine just makes sure I make it with the full picture in front of me.
06 · THIS WEEK ON THE PODCAST
Episode 3 is out this morning: Nathan Hill on why you upskill the network engineer, not hire the dev.
Nathan runs AWS’s telco business in Australia after more than 20 years inside the telco industry, and his day job is rolling AI into some of the country’s largest enterprises. He said the thesis of this newsletter in the podcast: automate the half of the process you can do safely now, and bank half the value.
What we get into:
• Where enterprise AI actually goes wrong: buying a standard product and expecting deep customisation, or building a platform for a problem you could buy off the shelf.
• The proof-of-concept trap. “We’ve done AI” is not an outcome if nothing has a path to production.
• The handover moment: the partner leaves, the chatbot breaks, and nobody in the building can fix it. Who owns it then?
• The cool kids problem. A centralised AI team presenting at town halls, isolated from the business, is how you buy consulting engagements forever.
• His big call: ten years ago maybe 5 to 10 percent of telco network engineers knew software. Within five years, 90 percent. You upskill the domain expert. You do not teach a dev 20 years of telco.
It opens with a hypothetical I have been asking myself and others for a few weeks now. I put it to him: if I used to be called a principal software engineer, was paid $200,000 a year, and now manage a team of agents that has made me 10 times more productive, what should I be paid? The debate is had online and different companies are treating it differently. Nobody has a settled answer yet, Nathan included, which is why I keep thinking on it.
All ten takeaways will be on LinkedIn later this week.
Listen: the episode page · Apple · Spotify
07 · THE ASK
Edition 1 of the AI Jobs Index lands Friday 24 July, the day after the ABS jobs numbers. So here is my question: what is one thing about AI and the Australian job market you want answered with actual data? Hit reply and tell me. If the Index can answer it, it goes in the edition, and I will credit the question if you are happy for me to.
Cheers,
James
Tech Teams Tuesday is the weekly read from Building Tech Teams. Written by James MacDonald, MD of NTP Talent. The AI Jobs Index is a Building Tech Teams publication. Powered by NTP Talent.