Kia ora! Welcome to New Zealand’s weekly roundup of AI news and education.
Back home after meeting a dozen prospects and customers across the ditch.
The main observed gap between Australia and New Zealand on AI isn't who's further ahead. It's what each market thinks it needs to invest in.
Check out 10 key insights on how AI differs right now between Australia and New Zealand, based on 19 conversations.
Happy reading ✌️
Did someone forward you this? Sign up!
academyEX is the EXceptionally proud sponsor of this week's newsletter.
🇳🇿 New Zealand News
NZ is "AI illiterate" and the economic cost is compounding. Mark Laurence, CEO and Founder of AI consultancy Ten Past Tomorrow, told RNZ's Midday Report that NZ lacks the foundational AI understanding needed to compete. Singapore has invested over $1 billion; NZ allocated $70 million through NZIAT. Harvard Business School research shows workers using AI complete tasks 20% faster and at 40% higher quality. Lawrence's three asks for politicians: make AI a mainstream election issue, fund AI literacy at scale, and leverage NZ's 85% renewable energy as a data centre competitive advantage.
RNZ Midday Report.
Our take: No adds from us. We’re big fans of Mark and TPT. The fact that this message is now on mainstream radio (not just business media) signals the conversation is moving from industry bubble to public discourse.
78% of NZ CEOs say AI has had little or no impact on their business. PwC's annual CEO survey found NZ lags 13 points behind the global average (65%). PwC NZ CEO Andrew Holmes described an "execution gap" between investment and results. Companies succeeding are "quite deliberate around investing in those use cases which are going to have real, scalable benefits".
4 min listen.
Our take: NZ CEOs are buying AI and not getting results. 78% is not a technology problem. It's a deployment problem. The 22% seeing impact are being deliberate about use cases. NZ businesses are investing in AI, confident about strategy, and struggling to convert any of it into measurable results. The execution gap is the story of 2026.
Every NZ emergency department now runs an AI scribe. Health Minister Simeon Brown confirmed all EDs have access to Heidi Health's documentation tool. Documentation time dropped from 17 minutes to 4 minutes per patient. 80% of Middlemore ED staff reported improved productivity. Doctors are seeing one extra patient per shift nationally. Brown described NZ as "among the fastest" health systems to move from pilot to nationwide frontline ED AI use.
4 min read.
Our take: No new regulation. No multi-year procurement cycle. NZ's small health system shipped nationwide while NHS UK pilots across 25+ trusts and the US VA hasn't completed its rollout. Being small isn't a disadvantage here; it's the entire competitive edge.
NZ's Privacy Commissioner joined 60 nations calling for AI deepfake safeguards. The joint statement targets non-consensual intimate imagery and child protection. Agencies from the UK, Australia, Canada, and France co-signed. The statement carries no legal force in NZ but signals regulatory direction. Netsafe reports daily deepfake complaints. School principals flag it as a growing problem among students.
4 min read.
Our take: Sixty governments saying "fix this" without specifying how puts the engineering problem squarely on the companies building the models. The generation tools are a year ahead of the detection tools.
59% of NZ workers think AI benefits companies, not them (Feb 11). Randstad's 2026 Workmonitor surveyed 27,000+ workers globally. NZ employers are 100% confident in business growth. 55% say AI already boosted productivity. NZ's scepticism rate sits 12 points above the global average. Meanwhile, 65% want more employer investment in AI skills.
3 min read.
Our take: Workers don't trust AI but want AI training. Not contradictory. They can see the technology coming whether they like it or not. They'd rather be prepared than replaced.
NZ's AI habit has a carbon cost that doesn't show up here. Auckland University's Prof Mark Gahegan mapped AI's energy consumption. GPT-4 training alone used 52-62 million kWh. NZ uses the services; Australia and the US host the data centres burning the power.
4 min read.
Our take: Same logic that made fast fashion possible: if the factory is overseas, the footprint doesn't feel like ours. Every NZ business running AI tools generates emissions that never appear in a sustainability report. NZ's "clean and green" branding has an AI-shaped accounting gap.
One NZ shipped a satellite messaging product in 45 days using AI-assisted code. The project normally takes 4-7 months. The telco runs 30+ AI agents across churn prediction, automated campaigns, call analytics, and a prepaid AI concierge. CEO Jason Paris: "AI first, but human when it matters". An internal AI School provides mandatory AI training for all staff.
4 min read.
Our take: Not a hackathon prototype; a production satellite messaging service in 45 days. While the PwC and Randstad surveys show NZ CEOs struggling with AI execution, One NZ is publishing the before-and-after numbers. Every competitor now has to answer: what's taking so long? (Note: sponsored content.)
✨A few people have been asking…
It’s Mike here, I run The AI Corner.
I’m not just into writing about AI. I help businesses grow without adding headcount by shipping AI systems that cut costs and increase throughput.
🎙️ The AI Corner Podcast
This week's guest is Derrick Edward, co-founder at Harnex AI & AuraScope. Hear:
Why small New Zealand businesses risk becoming invisible to AI search, and what to do about it.
How Derrick convinced resistant software engineers that AI makes them more valuable, not redundant.
Why there's never been a better time in human history to build something, and how obsession is the only real competitive advantage.
📚️ AI for Business
Helping leaders and teams adapt, learn, and scale with AI.
1️⃣ Elon Musk, Jeff Bezos and Sam Altman walk into a bar… Not quite, but surely that made you look. AI video just crossed the uncanny valley, and the above characters featured in a recent AI generated video clip. The characters look right, sound right, and hold their opinions. Frontier AI video generation has moved past "impressive demo" to "indistinguishable from produced content", and the implications for media, trust, and creative industries land all at once.
1-min watch.
2️⃣ Most businesses confuse three different stages of AI: Productivity AI makes individuals faster. Engineered AI redesigns workflows. Compound intelligence routes signal across functions automatically. Most organisations think about stage one, maybe stage two, and have no concept of stage three.
6-min read.
3️⃣ Two viral economic AI forecast articles with completely opposite conclusions, and our take: One models the S&P crashing 38% and the mortgage market cracking. The other models household purchasing power rising 18% and 7.2 million new businesses forming in a single year. The difference comes down to one assumption about how humans respond when jobs get disrupted.
6-min read.
4️⃣ Two types of AI strategy, most companies only run one: Lane 1 (productivity AI) makes individuals faster with copilots and writing assistants. Lane 2 (engineered AI) changes the system itself with agents, decision engines, and collapsing coordination costs. The problem is businesses treating Lane 1 as the entire strategy.
7-min read.
🌍 Tech Updates From Global
The selected top headlines from each major AI tech company.
OpenAI
Closed $110B funding round (Amazon $50B, Nvidia $30B, SoftBank $30B) at $730B valuation with 900M weekly active users. Three of tech's most strategic players now financially locked to OpenAI's success.
AWS partnership expanded $100B over eight years — AWS becomes exclusive cloud for OpenAI Frontier. Enterprises wanting OpenAI's best models must run them through AWS.
Frontier Alliances with Accenture, BCG, Capgemini and McKinsey to embed AI teams in enterprise clients. Consulting firms become OpenAI's distribution channel into thousands of enterprises.
ChatGPT ads rolling out to US free-tier users at $60 CPM, $200K minimum. First time OpenAI monetises the free tier; sharpens Anthropic's "no ads ever" positioning.
Codex coding agent hit 1.6M weekly users with 3GW dedicated Nvidia inference capacity. Coding agents have crossed into mainstream adoption.
Cash burn forecast revised to $665B through 2030, $25B burn in 2026, no profitability until 2030. More fundraising rounds inevitable.
Anthropic
Refused Pentagon demands for unrestricted military Claude access despite Defence Production Act threats. First major AI lab to publicly hold the line against military demands.
Claude Cowork plugins launched for Excel, PowerPoint, Google Drive, Gmail with custom plugin design per business unit. Claude moves from chat to embedded agent inside tools teams already use.
Claude engagement: 34.7 min average daily session — highest among all AI chatbots. Lower share but deepest engagement suggests strong retention.
Claude Code Security launched: AI vulnerability scanner found 500+ zero-days in open-source code. Extending from coding assistant to security auditor — higher-value use case.
Chinese AI labs ran 16M+ exchanges through 24,000 fake accounts to distill Claude's capabilities. Model distillation attacks now a real competitive threat.
Gemini 3.1 Pro scored 77.1% on ARC-AGI-2 and 94.3% on GPQA Diamond, topping OpenAI and Anthropic. No single lab holds a durable reasoning lead.
Nano Banana 2 image generation (4K, SynthID watermarking) now default across Gemini, Lens, Search and Flow. AI image gen embedded into products billions already use.
Gemini surpassed 750M MAU, narrowing gap with ChatGPT's 810M. Distribution advantage closing the gap.
Confirmed no ads planned for Gemini in 2026. Positions alongside Claude as ad-free, isolating ChatGPT.
Meta
$60B multi-year AMD partnership for 6GW AI data centre infrastructure using MI450 chips, including 10% AMD stake. Meta building the credible Nvidia alternative the industry needs.
Manus AI autonomous agent integrated into Ads Manager. First scaled example of autonomous AI managing real business budgets.
NVIDIA
Record Q4 revenue $68.1B (up 73% YoY), full-year $215.9B. Demand still exceeds supply.
AI chip startup boom: MatX, Axelera, SambaNova raised $1.1B+ in a single week. Investors betting the market is big enough for Nvidia alternatives.
Amazon
Invested $50B into OpenAI's round, expanded AWS partnership by $100B over eight years. Amazon hedges: own models plus OpenAI locked into AWS infrastructure.
Kiro AI coding tool caused 13-hour AWS outage by autonomously deleting and recreating an environment. First high-profile AI agent production outage.
Apple
Siri revamp delayed again — pushed to potentially iOS 27 (September). Every month behind is another month users build habits elsewhere.
Foundation Models now based on Google Gemini, reportedly 1.2T parameters (8x jump). Apple outsourced its AI brain to its biggest competitor.
AI leadership restructured again — third major reshuffle in 18 months. Still searching for the right team.
Three AI wearable devices in development: AirPods with cameras, Smart Glasses, AI pendant. Betting AI needs new hardware form factors.
Microsoft
Copilot Tasks previewed: background AI agent completing recurring work without supervision. Shift from "AI that answers" to "AI that works overnight."
MCP-based agents getting interactive widgets in Copilot Chat. Dynamics 365 ERP MCP Server launched. MCP becoming the standard protocol for enterprise AI agents.
UK Government selected Microsoft for deepfake detection framework (350+ participants including INTERPOL). Positioning as the trust layer for AI content.
Stock down 12% on OpenAI dependency — nearly half of RPO tied to OpenAI commitments. Wall Street questioning single-partner dependence.
xAI
Pentagon deal for Grok in classified systems. Anthropic refuses military access; xAI embraces it.
Six of twelve cofounders departed, five in the past year. Research continuity questions mount.
Grok market share surged to 15.2% from 1.6% a year ago. X/Twitter converting social users into AI users fast.
Mistral
Multi-year Accenture partnership for enterprise AI deployment. Mirrors OpenAI's consulting firm distribution strategy.
Ericsson partnership for telecom-specific AI. Niche OpenAI and Anthropic haven't prioritised.
Samsung
Perplexity integrated into Galaxy S26 as default AI agent with "Hey Plex" voice activation. Samsung chose Perplexity over Google — Android OEMs breaking from Google's ecosystem.
Perplexity
Perplexity Computer launched: autonomous agent orchestrating 19 AI models, $200/month for Max subscribers. Evolving from search to agent platform.
Samsung deal gives default AI assistant status across Samsung's global device base. Zero to Samsung's hardware distribution overnight.
👋 Mike & Erin