Kia ora! Welcome to New Zealand’s weekly roundup of AI news, events, jobs and education.
We’ve redesigned our website to make it a true home for everything we’re building. You’ll now find:
A growing library of our own content (expanding quickly!)
An archive of past newsletter issues
A showcase of our YouTube videos
We’re also building a trusted list: the New Zealand AI Directory. A curated, trusted platform to discover AI tools and companies in NZ (and beyond), compare options, and stay on top of the best vendors: saving time, staying updated, and making smarter choices.
Get in touch to be featured.
This week’s highlights
🛋️ AI On The Couch: What AI tools to use?
📊 77% of Kiwi firms now report AI cost savings
💸 Setup costs drop under $5k for most NZ businesses
👔 Spark warns execs scaling AI the wrong way
🎨 Creatives slam “microwave dinner” AI quality
🌾 Farmers risk giving away gold-mine data for free
🎓 90% of kids know AI, but few spot bias risks
📑 Lincoln Uni cancels 120 exams over AI cheating
🍌 Google drops “Nano Banana” image editor
Happy reading and listening ✌️
Did someone forward you this? Sign up!
HEADLINE STORY
📊 77% of NZ Businesses report operational cost savings from AI (71% in March)
📣 Word On The Street: NZ businesses report 91% efficiency gains while AI setup costs plummet to under $5,000 for most firms.
🔍 Zooming In:
Setup costs: 75% of NZ firms now keep AI setup under $5,000, while 58% spend less than $1,000, but barriers remain for bigger custom builds.
Financial impacts: More than 25% of firms now report $50k+ annual benefits, while only 4% spend over $50k a year on operations, but that’s down from nearly 30% last year.
Workforce impacts: 14% of organisations now report job losses from AI (up from 7%), while 45% report less need to hire, but 55% still see AI creating new career opportunities.
🏘️ Our Take: The AI honeymoon phase is coming to an end in New Zealand. Everyone’s got their hands on AI. The magic has worn off. We’ve moved past job-loss anxiety, and “shadow AI” is no longer going to cut it. The next phase is here: when CEOs and CFOs start asking hard questions about ROI, and laggards risk being left behind.
The catch is that cohesion is missing. Most organisations still run fragmented pilots. Teams dabble in isolation. Double-handling and messy processes persist (just slightly faster). What matters now is how deeply AI is embedded into workflows, decisions, and operations.
That requires AI operators and champions embedded inside finance, customer service, compliance, marketing, and operations. Not a separate AI department that becomes a bottleneck.
What this means:
The next productivity leap won’t come from adding another tool, it comes from redesigning work, integrating systems, and embedding AI into decision-making.
Firms that don’t formalise AI operator roles and rebuild around AI will stall, while competitors become AI-emergent (and some AI-native).
Government has set up ‘light-scaffolding’ with its national strategy, but the heavy lifting will happen inside firms through rewiring processes and building trust with staff and customers.
In short: New Zealand is moving from AI-aware to AI-adoption (we’re not yet close to AI-emergent). The constraint isn’t cost or access anymore. It’s whether leaders can design, champion, and scale new workflows that compound value month after month.
🐝 Save hours of work starting today.
Try Autohive free and build AI agents in minutes that tackle your most repetitive tasks - secure, private, and ready when you are.
💼 Business & Industry
Most Kiwi executives are asking "Am I behind the AI curve?" - the wrong question entirely. Few organisations are scaling AI effectively despite widespread experimentation, according to Spark's data. The real opportunity lies in careful business alignment and robust measurement frameworks, not racing to deploy the latest models.
4-min read.
Our take: FOMO has driven poor AI decisions to date, but the winners will emerge from methodical experimentation with clear business alignment. Smart companies will view stalled pilots as valuable data, rather than wasted investment.
Kiwi creatives just declared war on "microwave dinner" AI quality. AKQA's inaugural Q&A Club revealed that AI is making mediocre work obsolete while turning great creativity into a luxury export. Creative director Jae Morrison's brutal assessment: AI output is "frozen in the middle, burnt on the ends, f***ing plastic".
5-min read.
Our take: The conversation has moved beyond job replacement anxiety to strategic deployment. Smart creatives treat AI like photographers treated painting: rather than a threat, but a new medium with unique possibilities. When craftspeople with 10+ years of experience can show everyone the difference in quality, that intuition becomes the most valuable input in any AI workflow.
Kiwi farmers are potentially handing over their gold mine data for free AI advice - and tech companies are cashing in. Every app feed, pasture prediction, and stock number creates valuable datasets that agribusinesses and banks want to buy. Most farmers focus on getting work done, not reading terms-of-service agreements that hand over data ownership forever.
3-min read.
Our take: We're witnessing the emergence of a new farmer archetype: the hybrid decision-maker who combines algorithmic insights with hard-earned intuition. The winners won't be those with the fanciest AI tools, but those who develop the critical thinking skills to know when to trust the algorithm and when to trust their gut.
👨🏻🎓 Education & Society
90% of Kiwi primary students know about AI, but only 19% understand its biases. Over half of students use AI tools at home without adult guidance, while 68% of teachers worry about over-reliance on AI undermining foundational skills. Cultural accuracy is particularly concerning with 60% of teachers found AI's te reo Māori outputs unreliable.
66-pages.
Our take: We're creating a generation fluent in AI but illiterate about its limitations. Without digital literacy education now, we risk students who can't distinguish AI fiction from fact.
Lincoln University just nuked 120 students' exams after suspecting AI cheating - and it's working. One-third of postgraduate finance students showed "unexpected proficiency" in coding assignments, triggering mass re-sits where students must verbally defend their code. The lecturer's brutal rule: "If you wrote the code yourself, you can explain it".
4-min read.
Our take: Universities face an impossible choice: embrace AI as a learning tool or maintain traditional assessment integrity. The arms race between AI tools and academic integrity just shifted to real-time human verification. If we're questioning the process, yet students are only assessed on the outputs, perhaps our assessment approach requires a rethink?
🎙️ The AI Corner podcast
This week’s guest is Matteo Castiello, the founder of Insurgence.
Matteo walks us through his journey from a planned career break to spearheading enterprise generative AI adoption across Australasia. Discover how he shifted from consulting with design sprints to building agent‑driven workflows that deliver real productivity gains.
This episode was recorded in mid-July, but the principles still ring true!
📚️ Levelling Up With AI
Practical AI for Everyday Work
1️⃣ AI On The Couch: Too many AI tools, too little time. Erin’s question (“Which ones should I actually use?”) nails it.
7-min watch.
2️⃣ Should it be a workflow or an AI agent? That’s the fork most teams don’t map. Understanding the spectrum between chatbots, copilots, and agents is key to designing effective systems.
4-min read.
3️⃣ Beyond ChatGPT: we’ve been testing Relay.app, Autohive, Glean, n8n, and Google AgentSpace. Here’s what we learned about where these platforms actually deliver business transformation.
4-min read.
The Speed & Scale of AI
4️⃣ Google’s “Nano Banana” 🍌 is the leap in AI image editing nobody saw coming. Natural language + precision editing = a new jagged frontier.
8-min watch.
🌍 The News From Global
Teen suicide lawsuit targets ChatGPT after family alleges the AI chatbot provided suicide methods and discouraged seeking help from family. Legal action claims OpenAI rushed GPT-4o's May 2023 launch, creating design flaws that prioritised engagement over user safety for vulnerable minors.
5-min read.
Tech Updates You Should Know
OpenAI: Published a case study with Retro Biosciences on a custom AI model that redesigns proteins for cellular rejuvenation; involved in a joint safety and alignment stress test with Anthropic on their respective AI models; launching 'gpt-realtime' and Realtime API updates for production voice agents; Codex upgraded to GPT-5 with tighter GitHub integration; will introduce parental controls and emergency features following a lawsuit over a teen's suicide; revealed GPT-6 will focus on personalised, non-exploitative interactions; maintaining top spot in consumer AI usage rankings despite legal and safety scrutiny.
Google: Released ‘Nano Banana’ (Gemini 2.5 Flash Image model) with high-fidelity image editing and multilingual upgrades; rolled out AI-powered NotebookLM tools in 80 languages; introduced Gemini for Home to replace Assistant starting October; expanding AI model availability to 180 countries and incorporating AI into Photos, Translate, and video tools (avatars and editing); reportedly in talks with Apple for Siri integration, boosting investor confidence; facing criticism over AI Overviews affecting publishers and a restaurant incident; maintains #2 consumer app rank behind ChatGPT.
Anthropic: Launched Claude for Chrome with sidecar chat and context integration; piloting Claude Code for browser automation, but facing prompt injection and cyber misuse vulnerabilities; announced policy to store user data for five years unless users opt out by Sept 28; reached proposed class settlement with authors over copyrighted training data.
Microsoft: Launched in-house models MAI-Voice-1 and MAI-1-preview focused on speech and text generation efficiency; integrating these models into Copilot while expanding Copilot to Samsung’s 2025 smart TVs and monitors; introduced VibeVoice-1.5B open-source model supporting advanced multi-voice synthesis.
Meta: Partnered with Midjourney to license its aesthetic tech for image and video generation; investing $14.8B for a 49% stake in Scale AI while facing a talent exodus from its Superintelligence lab; freezing hiring amid leadership reshuffles chatbot misuse with minors drew AG scrutiny.
xAI: Released Grok Code Fast 1 for efficient coding tasks and open-sourced Grok 2.5 under restricted license; plans Grok 3 launch in six months; filed a sweeping federal antitrust lawsuit against Apple and OpenAI alleging anti-competitive practices in the iOS ecosystem.
NVIDIA: Requested suspension of H20 chip production for China amid regulatory tensions.
DeepSeek: Facing user decline with >40% drop in web traffic.
Other: Andreessen Horowitz and Greg Brockman spearheading a $100M campaign to shape AI regulation ahead of elections; Stanford study links generative AI to a 16% drop in entry-level job openings.
📅 AI Events in New Zealand
16 AI events happening across Aotearoa this week, with online sessions taking the lion’s share.
This week’s featured event:
Rethinking Roles - AI, Ethics and the Future of Communications Teams, Auckland: PRINZ tackles AI implementation frameworks for comms leaders at a 7:30 breakfast session.
📅 Promote your event with us. Reply to let us know.
💼 AI Roles Around Aotearoa
Picklist of 🌶️ HOT 🌶️ new roles in AI this week.
Digital Apps & AI Consultant, Arinco: Auckland
Machine Learning Engineer, Halter: Auckland
💼 Promote your job with us. Reply to let us know.
🤦 ️ AI Fail Of The Week
We all love AI, but it’s certainly far from perfect 🤔 …
Somehow Mike (is it him?!) might have met Obama at a McDonald’s diner (and his legs might have disappeared into the booth…).
👋 Mike & Erin