AI Down Under

Mackenzie

I interview AI experts in Australia.

Episodes

  1. 22h ago

    He Builds AI Agents For $15B Revenue Companies | JC Gosselin (Enterprise AI & Tech @ Terem)

    Most companies buy a pile of AI licences, watch usage peak in week 3, and by month 6 nothing has actually changed. Jean-Christophe (JC) Gosselin is Principal of Enterprise AI and Technology at Terem, an Australian tech consultancy. He has spent his career shipping data and machine learning inside large organisations, including loyalty and freight analytics in aviation, and these days running agentic engineering uplifts inside some of Australia's largest enterprises and government agencies. He is also French, an aviation geek, and the funniest man on Australian enterprise AI LinkedIn. In this conversation he walks through the 6 levels of agentic engineering in plain English, where Australian companies honestly sit on that ladder, what a 4 week uplift actually looks like inside a client, and why making your engineers 3 times faster can leave you no better off. CHAPTERS 0:00 The funniest man in Australian enterprise AI 1:21 Battle scars from real clients 2:47 The 6 levels of agentic engineering, in plain English 9:34 How companies actually move up the levels 13:42 Find the bottleneck before you automate anything 14:44 Restructuring a team without cutting it 16:49 What to measure, and the 3 hours per person per task 19:54 Pick one thing, get it right, measure it properly 20:54 The 2 roles a project team turns into 23:28 Inside a 4 week uplift, week by week 25:40 The stack, and picking the right model for the job 28:58 What those 4 weeks actually look like for the client 32:49 The engineer a vendor sends to lock you in 34:52 How his own day changed, and AI brain fry 39:44 Automate the 80% they hate, protect the 20% they love 41:35 Where the jobs go next, and why he expects rehiring 44:15 Building an iOS app at night knowing nothing about iOS 46:08 Start small, take one thing all the way to production 47:30 Who else is doing good work in Australia 49:30 How to reach him CONTACT JC JC's LinkedIn: https://www.linkedin.com/in/jean-christophe-gosselin Terem: https://terem.tech CONTACT MACKENZIE (YOUR HOST) My name's Mackenzie. Ex-KPMG consultant, built coaching and recruitment businesses to just under $2m in revenue, now all in on (and obsessed with) AI. I help Australian CEOs and founders hire AI, ML, data and software engineers, from first hire through to CTO, backed by a 40k+ LinkedIn network. I build with AI daily! If you're hiring, or working out where AI fits in your business, message me on LinkedIn: https://www.linkedin.com/in/macthompson Subscribe for more conversations with the people actually building AI in Australia!

  2. 2d ago

    The AI CEO Who Backed Startups Worth $3.5B+ | Michael Batko (CEO, Hourglass AI Consulting Services)

    Most companies build AI backwards. They start with agents, the agents break, and they think AI "doesn't work". Michael Batko is the co-founder and CEO of Hourglass AI consulting services, a Sydney and Melbourne AI implementation firm that makes Australian companies AI native. Before this he ran Startmate, Australia's most active startup accelerator, for 8 years. He raised $50M and backed 230 startups worth $3.5B, including Heaps Normal, Lyrebird Health and Amber Electric. He is a 2x founder with two exits, one of them his Saturday morning run club, which he sold with a single LinkedIn post. In this conversation he lays out the order it all has to happen in, foundation then infrastructure then agents, what every piece actually costs in dollars, why 95% of his clients arrive by referral, and the build he tells clients to stop paying him for. CHAPTERS 0:00 The best connected man in Australian startups 1:24 Why he went all in on Hourglass AI 3:06 How Hourglass AI was born, and how fast 5:25 The values card exercise that made them co-founders 6:45 The moment he knew Finlay was the one 8:24 What makes Hourglass AI different 10:28 The one number they run the company on 11:13 The enquiry form he replaced with an AI roadmap 15:08 Where their clients come from 17:02 The 3 layers for implementing AI agents 22:33 Why build a thin AI layer instead of a year long data cleanup 25:03 The tech stack, and why it is boringly simple 27:26 Your data, your servers, and who gets to see what 29:21 Inside the company brain they built for themselves 31:16 Why the foundation is the hard part 33:13 The build he tells clients to stop paying him for 33:48 Nobody wants half price heart surgery 36:27 What it costs, priced line by line 40:09 The service he has never sold, but might 43:58 The businesses this works best in, and the size sweet spot 45:32 The operators he watches, and the PE roll up trend CONTACT MICHAEL Michael's LinkedIn: https://www.linkedin.com/in/batkomichael Hourglass AI: https://thehourglass.ai CONTACT MACKENZIE (YOUR HOST) My name's Mackenzie. Ex-KPMG consultant, built coaching and recruitment businesses to just under $2m in revenue, now all in on (and obsessed with) AI. I help Australian CEOs and founders hire AI, ML, data and software engineers, from first hire through to CTO, backed by a 40k+ LinkedIn network. I build with AI daily! If you're hiring, or working out where AI fits in your business, message me on LinkedIn: https://www.linkedin.com/in/macthompson Subscribe for more conversations with the people actually building AI in Australia!

  3. 3d ago

    Evgueni Iakhontov - Partner, Revity (AI Consultancy)

    Australia is 2 to 3 years behind other countries when it comes to adopting AI at scale. That assessment comes from Evgueni Iakhontov, a Partner at Revity, a Melbourne-based technology consultancy helping organisations turn AI from experimentation into practical business value. Evgueni has more than 20 years of experience leading technology teams and delivering large-scale transformation programs across Australia. Today, he works with organisations ranging from small businesses to major enterprises, helping them identify where AI can make a real difference, implement it safely and ensure the investment delivers a measurable return. CHAPTERS 0:00 3 decades in Australian tech, Sausage Software to Revity 1:42 Who Australia is behind and what the gap looks like 5:00 Why the big end of town can't test and learn 6:29 The business functions AI still hasn't touched 8:12 Voice bots at scale, what's already normal overseas 10:29 Why the big pilots fail, every scenario instead of one 13:45 7 months iterating a beauty salon voice bot 14:44 The API advantage enterprises don't realise they have 19:11 Why generative AI should never calculate payroll 22:16 The 4 layers a company must change to catch up 25:33 Your next job, managing and quality checking AI agents 28:03 Necessity, scale and the real cost of frontier models 30:35 Inside the salon voice bot experiment 33:56 Admit it's AI and more callers ask for a human 36:19 How to launch a voice agent without losing customers 40:49 The fastest ROI is inside the business 41:51 Recruitment rebuilt end to end with AI 45:42 Don't automate the workflow, redesign it 48:03 Agentic product development, not just coding 53:04 AI is an ongoing program, not a project CONTACT EVGUENI Evgueni's LinkedIn: https://www.linkedin.com/in/evgueni-iakhontov CONTACT MACKENZIE (YOUR HOST) My name's Mackenzie. Ex-KPMG consultant, built coaching and recruitment businesses to just under $2m in revenue, now all in on (and obsessed with) AI. I help Australian CEOs and founders hire AI, ML, data and software engineers, from first hire through to CTO, backed by a 40k+ LinkedIn network. I build with AI daily. If you're hiring, or working out where AI fits in your business, message me on LinkedIn: https://www.linkedin.com/in/macthompson Subscribe for more conversations with the people actually building AI in Australia!

  4. Jul 25

    Joff Outlaw - Managing Director (Australia), MakerTech

    His team's tool for New Zealand's Department of Conservation turned 1,000s of hours of paperwork into minutes. The uncomfortable part is what he sees everywhere else. Clients keep bringing him AI ideas, and often when you pull back the process underneath is completely broken. AI is a logic beast, and it chokes on messy human workflows. Joff Outlaw is Managing Director for Australia at MakerTech, a design and engineering consultancy of just under 30 people building custom AI, digital products and composable platforms across ANZ. Everything is designed, built and supported onshore. Before this he spent a decade inside the big firms. Most recently he ran Designit, Wipro's global design firm with 100s of designers worldwide, as Managing Director for Australia and New Zealand, while a Partner in Wipro's consulting business. Earlier he worked at Accenture and PwC. He also wrote Busy Idiots with psychologist Brad Marshall, a Wiley published book about how modern work keeps us busy with things that don't matter. IN THIS EPISODE The government paperwork that took 1,000s of hours, and the custom build that now does it in minutesWhy Copilot got lost in the complexity, and the paralegal with unlimited focus they built insteadThe bank that cut 45 contact centre roles for an AI voice bot, then wound it back a month laterWhy an AI tool saving someone 15 minutes often means nothing for the businessThe dangerous, dear or dull test for deciding what to automateThe rebuild behind 10,000 conservation plans, plus approval onto the Australian government's AI panelComposable technology explained with Lego blocksHow a team of just under 30 keeps beating Deloitte Digital in pitchesThe RFP typo his business partner wanted to keep to prove a human wrote itThe $2B cow collar company Peter Thiel just backed, and the radiology AI he rates most in Australia and New Zealand CHAPTERS 0:00 The man telling us to slow down sells speed 1:49 A third of New Zealand runs through one agency 3:04 Copilot got lost, so they built custom 4:33 A paralegal with unlimited focus 4:56 Ten years in corporate, then the jump 7:43 What MakerTech does differently 11:39 What Busy Idiots teaches about AI 13:26 "The process is just completely broken" 14:46 The bank that cut 45 jobs, then wound it back 16:04 AI saved you 15 minutes. So what? 18:36 Dangerous, dear or dull 20:30 Humans know more than they can tell 22:07 Approved to build AI for the Aussie government 23:17 The tool behind 10,000 conservation plans 29:34 What composable actually means 34:05 The Lego blocks explanation 35:35 The most valuable lesson of his career 36:47 "Transparency is kindness" 39:42 Beating Deloitte Digital with a team of three 40:32 "Leave the typos in" 41:56 The AI companies nobody's talking about 45:28 His offer to every graduate listening CONTACT JOFF Joff's LinkedIn: https://www.linkedin.com/in/joff-outlaw-970a3b11 His book Busy Idiots: https://a.co/d/0fuuyDV0 CONTACT MACKENZIE Ex-KPMG consultant, built coaching and recruitment businesses to just under $2m in total revenue, now all in on AI. I help CEOs and founders hire AI, ML, data and software engineers, from first hire through to CTO. 40k+ LinkedIn network. DM me: https://www.linkedin.com/in/macthompson

  5. Jul 23

    Simon Kriss - CEO, Sovereign Australia AI

    Simon Kriss is building Australia's own AI. His argument is simple. Every model Australian businesses and governments use today is rented from overseas, and somebody else holds the off switch. Simon spent more than 35 years in the corporate world, held the chief AI officer role, and has keynoted at the UN in Geneva. He mentors boards and executives on AI adoption, and he is now CEO of Sovereign Australia AI, raising the backing to build a large language model from the ground up right here. He explains why fine tuning an open weight model doesn't get Australia there. You still don't know what data went in, what biases came with it, or whether it starts spewing hate speech in six months. He calls that "domestic, not sovereign". A true foundation model means knowing every data source, paying Australian copyright holders (a public minimum commitment of 10 million dollars, and he thinks the honest number is closer to 30), and around 256 of NVIDIA's latest Blackwell chips, which he says would make it bigger than any supercomputer in Australia today. Then he gets into the part most people have never heard of. The US Cloud Act. He says the American government can compel any American company to hand over the contents of any server it owns or leases, anywhere in the world, and the provider is not allowed to tell the client anyone is looking. On his telling, Australian data sitting in an Australian data centre doesn't help if the server is American owned. His words, it might as well be sitting in Texas. He also walks through what building it here would involve. Copyright holders get paid properly, every data source gets meta tagged so anyone can ask for their content to be pulled out at the next retrain, and there is the question of how to include the indigenous voice without appropriating a storytelling culture. ALSO IN HERE - The six layers of sovereignty (most people stop at data centres) - The accidental Bogan AI they built from Reddit data - December's privacy law change that turns black box AI into a legal problem - Why his auntie saying "not too bad" confuses an American model - The war scenario where America turns Australia off CHAPTERS 0:00 One of Australia's leading voices on AI 0:52 What Sovereign Australia AI is building 1:22 Somebody else holds the off switch 2:51 Talk fests, and why the money hasn't come 5:40 China builds practical AI. America builds God in a box. 7:47 The six layers of sovereignty 10:01 Why "not too bad" confuses an American model 12:54 Anti woke orders and Australian values 16:18 Why fine tuning is only "domestic, not sovereign" 23:51 Five years out, and the trust problem 30:26 The US Cloud Act, explained 33:09 "It might as well be sitting in Texas" 35:31 Staying at the model layer, and "we built Wi-Fi" 38:39 The talent case, and the DeepSeek connection 42:17 The Friday evening a model got switched off 43:21 Investors can't see the burning bridge 44:21 How to spot fake "Australian owned" 47:52 The UK playbook he wants to copy 51:15 Fencing wire, snow skiing, and having a go 53:17 Digital colonisation and the war scenario ABOUT SIMON Simon Kriss is the CEO of Sovereign Australia AI, which is making the case for Australia to build and own its own AI models rather than rent them from overseas. He is a keynote speaker who has taken the stage at the UN in Geneva, an author, a former chief AI officer, and a mentor to boards and executives on AI adoption, after more than 35 years in the corporate world. Simon on LinkedIn: https://www.linkedin.com/in/simonkriss

  6. Jul 22

    Dr Brian Lowe - Chief AI Officer in Australia

    Brian Lowe has held the title of Chief AI Officer—one of the few in Australia to do so—and brings a clear message for business leaders concerned they are falling behind on AI: you are not missing out. In fact, up to 95% of AI initiatives never reach production, and much of what is presented as progress is often little more than hype. Brian began his career with a PhD in Information Systems and has since built deep experience across enterprise AI and data. His roles have included Assistant Director at the Bureau of Meteorology, CTO of an AI startup, Head of Data Architecture at one of Australia’s big four banks, and Director at a major university, where he helped the institution navigate the early emergence of AI. Most recently, as a Chief AI Officer, he advised organisations across government, utilities, and global retail. He is now focused on building a sovereign Australian AI model—trained on Australian data, hosted domestically, and governed by Australian law. Brian on LinkedIn: https://www.linkedin.com/in/brianslowe In this episode, Brian explains what a Chief AI Officer actually does day to day, and much of it involves cutting through noise and challenging misplaced optimism. He has seen conferences showcase questionable case studies, consultancies promote solutions that fail to deliver, and executive sponsors persist with struggling initiatives to avoid admitting failure. His approach to navigating this landscape draws on an unexpected source: a lesson from a Simpsons episode about a monorail. Then he takes apart agentic AI. Much of what is sold as agentic is just automation, a technology that has been around for 20 years, rebadged with a bit of AI sugar on top. The properly agentic systems are useful but they misfire, and his test for where you can trust them is simple. Would you hand your online banking to an assistant who gets it wrong 3% of the time? He also explains why the US government accidentally handed Australia the business case for its own AI model, why students who learn with AI do worse, why everyone running a team of AI agents has become a manager whether they like it or not, and why AI has made hiring good people harder than it has ever been. ALSO IN HERE - Why beer ended up in the nappy aisle, and what it says about your data - The "bad debtors" his team discovered that you actually want to keep - The AI he instructed to abuse him, and why it improves his thinking - Why the best AI implementation he has ever seen won a Nobel Prize - Why a $70k graduate might beat your AI subscription CHAPTERS 0:00 Who is Brian Lowe 1:15 He is building an Australian AI model 2:00 The two events that made the case for it 5:33 "95% of an AI project is data" 7:36 Start from scratch, or fork and fine tune 10:15 Most business problems do not need model training 11:27 The Industrial Revolution question 14:57 A four day week, if we get this right 16:45 What a Chief AI Officer actually does 18:50 "There's a lot of b******t out there" 23:10 How to spot snake oil before you pay for it 25:22 The monorail episode 28:01 95% of AI projects never make it 30:18 "You are not missing out" 33:52 Rebadged RPA with AI sugar on top 35:03 Beer in the nappy aisle 40:09 What agentic AI really is 43:31 The online banking test 45:00 The best AI implementation he has ever seen 49:18 The infinitely patient tutor 53:06 Students who learn with AI do worse 55:46 The AI he built to abuse him 56:52 Everyone is becoming a manager 1:02:08 Why teams cap out at seven 1:04:54 Why AI made hiring harder 1:09:44 The $70k graduate vs the API bill 1:15:08 How AI hacked your brain 1:17:44 "You cannot substitute people for AIs"

About

I interview AI experts in Australia.