Kia ora! Welcome to New Zealand’s weekly roundup of AI news, events, jobs and education.
You may have seen the news about the All Blacks this week.
The Rugby World Cup draw was released with New Zealand first up playing Australia in what'll be the first time these trans-Tasman rivals have faced off in the pool stage.
Why are we talking about rugby in an AI newsletter for New Zealand?
Because ESPN used AI to run a simulation to predict the entire tournament, and that prediction held that…
The All Blacks wouldn't make it past the quarter-finals, losing to South Africa 24-22 in a tight battle.
The Springboks would win a 5th title and 3rd on the trot, beating Ireland 24-18 in the final.
The Springboks might be incredibly strong, don't get me wrong, but there might be some serious questions we should be asking about the quality of the prompting used to run that simulation.
On a separate note, last week we announced our first in-person event coming up. Stoked to share it sold out within a matter of days. Thank you to everyone who grabbed a ticket, genuinely appreciate the support. Looking forward to meeting a bunch of you in person.
Keep your eyes peeled for future events. If you've got ideas for what you'd like to see or topics you want covered, let me know.
This week’s highlights:
Police cut case backlog 80% with AI
GP practices deploy digital workers
One NZ slashes diagnosis time by 80%
NZ leaders: 95% see no AI returns yet
Custom GPTs fail without deep research
PainPal: AI safety companion for tradies
AI builds interactive dashboards in 30sec
Female AI assistants dominate NZ launches
805% AI traffic surge proves agent commerce
Dave Howden: why legal drops from $5K to $200
OpenAI declares "code red" on ChatGPT quality
Happy reading and listening ✌️
Did someone forward you this? Sign up!
MIKE’S MUSINGS
Beyond Agents: Build Interfaces With AI That Match Your Tasks
Agents and workflows handle execution brilliantly, but they often give you the wrong interface for what matters most: editing assumptions, toggling scenarios, reviewing outputs, and making decisions with the information they produce.
You can now describe what you need and have the perfect interface materialise in 30 seconds, then watch it disappear when you're done. This isn't about vibe coding new apps; it's about fundamentally changing how you complete work by using natural language to create tools that eliminate repetitive tasks.
HEADLINE STORY
📊 AI Foundations Roundtable: NZ leaders reveal AI adoption gap
📣 Word On The Street: Cognizant and NZ Herald roundtable found 80% of organisations adopt generative AI but 95% see no measurable returns yet. MIT Media Lab research confirms most are still experimenting to see what works.
🔎 Zooming In:
Kat Carter, Markets Lead, Cognizant NZ: Identified that agentic AI (where autonomous agents act like workers) is the most complex application requiring complete organisational rethinking from people and culture perspectives.
David Johnson, CIO, UDC Finance: Warned that early adopters are investing at the bleeding edge right now and will need to wait for market maturity before seeing returns on their AI investments.
Chris Bradley, Head of Enterprise Architecture, 2Degrees: Ran an "AI Month" that went beyond technical implementation to explore ethical questions around agentic AI and embed an AI-literate culture across the entire business.
Emerald Cagney, AI Enablement Lead, ANZ Bank: Takes a cautious approach by asking employees about their AI concerns first, then unpacking what the organisation needs to do to address those worries and build trust.
Frances Valintine, AI Futurist, AcademyEX: Found 200 NZ business leaders rated their organisations just 3 out of 10 on AI readiness because senior leaders still haven't communicated benefits in language beyond "IT-speak".
Ana Ivanovic-Tongue, Executive Council, AI Forum: Emphasised that better communication on AI's benefits is the key to driving buy-in across the nation, making education and upskilling essential.
Peter Kennedy, CIO, NZ Post: Stated bluntly that AI is becoming absolutely pervasive in products and services whether people like it or not, with different implications for large corporates versus small businesses.
🏘️ Our Take: Not much to add here. Well covered. Without addressing underlying AI literacy and mindset barriers today, and improving foundational organisational processes and data challenges first, organisations risk accelerating towards creating more problems through attempting to build an AI-forward entity.
💼 Business & Industry
Digital workers to ease GP pressure. Four major NZ primary care groups, including ProCare (I interviewed Group CEO Bindi Norwell last week), have deployed robotic process automation to handle repetitive tasks such as cardiovascular assessments and ACC claims. Every workflow receives clinical governance approval before it runs unsupervised to ensure patient safety.
2-min read.
Our take: Automation done transparently helps practices serve communities better without worsening health inequalities. GPs can focus on actual patient care instead of paperwork and time savings will unlock revenue and redeploy staff to higher-value work.
Police slash case backlog with automation. NZ Police cut case processing from two weeks to four hours using Appian's AI-powered platform. Backlogs dropped from 4,000 to under 50 cases during most shifts, freeing staff for higher-value work.
3-min read.
Our take: The platform surfaces correct case assignment rules by offence code and location, then matches work to trained staff. Public enquiries get faster, more accurate responses. This biggest concern from the public will be what guardrails are in-place to avoid bias entering the AI’s decision making compass.
One NZ cuts network diagnosis by 80%. Network Concierge AI agent on AWS Bedrock pulls performance data, alarms and incidents into one summary for customer service reps. Analysis time dropped 80%, helping customers get back online faster during outages.
2-min read.
Our take: Losing network connectivity, even briefly, disrupts daily life. This technology will eventually predict issues before they occur and automate repairs before major disruptions. AI handling repetitive data gathering while people focus on complex issues adds another strong string to One NZ’s bow as they pursue an “AI-first, human when it matters most” model on their path to becoming the world’s most AI-enabled telco.
Construction workers get AI safety companion. PainPal provides personalised recommendations tailored to worker's trade and tasks. Workers ask questions like "Why does my shoulder feel tight?" and get expert guidance instantly, whether on site or in the van.
3-min read.
Our take: Most musculoskeletal issues are preventable with timely intervention. This tool acts as a constant, trusted companion empowering workers to make safer choices.
⚖️ Government & Legal
Kiwis paid out in settled Anthropic book piracy case. Kiwi authors Rose Carlile and others will receive US$3,000 per book after Anthropic agreed to a USD$1.5B settlement. The AI startup scraped pirated books from Russian websites instead of licensing them properly.
5-min listen.
Our take: This sets a pricing floor for training data: USD$3K per book is remarkably cheap for billionaire AI companies. Expect licensing organisations to demand far more.
⚖️ Society & Education
Kiwi Female AI assistant trend under scrutiny. NZ Post, Skinny, ASB, Noel Leeming and Eden Park all launched female AI spokespersons despite having limitless design options. Even Beef and Lamb's AI farm dog is female (accidentally, they claim).
4-min read.
Our take: The tech doesn't require personification at all, it's entirely synthetic. At least Clippy was gender-neutral.
📚️ AI for Business
Helping leaders and teams adapt, learn, and scale with AI.
1️⃣ Your Custom GPTs Are Failing Because You're Skipping the Hard Work. How to build Custom AI Coaches: Building AI coaches in ChatGPT only works if you feed them deep research, robust context files, and run systematic testing. Most people skip straight to prompts, get mediocre outputs, then abandon the entire approach.
2-min read.
2️⃣ The Real AI Race Isn't OpenAI vs Google. It's Chat Interface vs Infrastructure: OpenAI owns the brand moment with ChatGPT, but they're boxed inside a chatbot extending slowly outward. Google already owns browser, search, email, docs, and Android, positioning as the default platform where AI assistants will live.
1-min read.
3️⃣ Google's Beautify This Slide Feature Just Killed Presentation Formatting: Nano Banana converts messy text into polished slides inside of Google Slides in seconds with clean hierarchy and visual structure. No Canva, no dragging boxes, just instant storytelling-ready design whilst you focus on delivery and persuasion.
1-min demo.
4️⃣ Gemini's Dynamic View Builds Custom Interfaces for Every Question You Ask: Ask about property markets and get an interactive dashboard. Plan a Rome trip and get a visual itinerary with filters. Google's generative UI doesn't return text anymore, it codes bespoke experiences from scratch for your exact query.
1-min demo.
5️⃣ 805% Surge in AI Traffic to Retail Sites Proves Agentic Commerce Is Already Here: Black Friday exposed the shift: AI agents drove an 805% increase in U.S. retail traffic. If your products aren't discoverable and transactable inside AI tools, you're watching market share erode whilst competitors build for the invisible marketplace.
1-min read.
6️⃣ Claude Just Solved the Problem of Finding Past AI Conversations: Endless scrolling through ChatGPT history is over. Claude's conversation search links you directly to specific chats, turning your entire conversation history into an instantly accessible, searchable knowledge base.
1-min read.
7️⃣ Google Handed 450 Million Users the Ability to Build AI Agents Without Code: Workspace Studio lets anyone create agents in minutes that orchestrate work across Gmail, Drive, Sheets, and external tools like Salesforce. This isn't workflow automation, it's infrastructure for delegating reasoning work that adapts.
2-min demo.
🎙️ The AI Corner Podcast
This week's guest is Dave Howden, CEO and co-founder of Superhuman AI. Hear:
Why protecting 400 jobs in Council might hurt 400,000 people trying to build houses, and the fiduciary math leaders avoid.
How legal conveyancing drops from $5,000 to $200 when you remove humans from the loop, and why no incumbent survives that gap.
Why Google's quietly winning enterprise AI whilst OpenAI wins headlines, and what CTOs must become to survive the shift.
🌍 Tech Updates From Global
The selected top headlines from each major AI tech company.
OpenAI
Declared "code red" emergency to improve ChatGPT quality after competitive pressure from Google's Gemini 3, delaying advertising and other product launches.
Acquired Neptune research platform for $6.5M to deepen model behaviour visibility and strengthen experiment tracking capabilities.
Launched OpenAI for Australia, signalling continued international expansion despite refocusing resources on core product improvements domestically.
Teased new reasoning model release for 9 December intended to reclaim technical leadership position in the AI model race.
Anthropic
Secured $200M multi-year partnership with Snowflake, making Claude models available across 12,600 enterprise customers through AWS, Google Cloud, and Azure.
Reached $1B usage milestone for Claude Code and acquired Bun JavaScript runtime to strengthen developer tooling capabilities.
Weighing massive IPO at valuations above $300B, which would make it one of the largest public offerings ever contemplated.
Claude Opus 4.5 launched with significant improvements in coding, agent capabilities, and token efficiency for enterprise applications.
Microsoft
Slashed AI agent sales growth targets after salespeople missed quotas and customers resisted "unproven agents", marking first major enterprise adoption setback.
Extended OpenAI partnership through 2030 or AGI achievement, retaining 27% stake and exclusive Azure API access under restructured agreement.
Announced Microsoft 365 restructuring expanding AI, security, and management capabilities with pricing updates for 2026 and new Copilot app.
Amazon
Unveiled four major AI product families at re:Invent including Nova 2 models, three autonomous agents, and Trainium3 claiming 4x faster inference.
Launched AI Factories enabling on-premises enterprise AI infrastructure deployment, accelerating customer buildouts by "months or years" according to AWS.
Described "walled garden" vision where customers build agents that don't break when base models upgrade, challenging Microsoft's agent strategy.
Perplexity
Secured investment from Cristiano Ronaldo and entered global sponsorship agreement, marking AI search's first major celebrity endorsement for mainstream adoption.
Launched PayPal Instant Buy enabling direct purchases through Perplexity's shopping interface with merchants including Abercrombie & Fitch and Adorama.
Rolled out AI Assistants with Memory adding persistent context capabilities to compete with ChatGPT's conversational continuity features.
📅 AI Events in New Zealand
14 AI events happening across the country this week.
This week’s featured event:
Autohive Intermediate AI Cooking Class, this Friday in Auckland. This is a hands-on midday session for those ready to level up their AI agent skills.
📅 Promote your event with us. Reply to let us know.
💼 AI Roles Around Aotearoa
Picklist of 🌶️ HOT 🌶️ new roles in AI this week
AI Advisory Manager, Deloitte: Wellington
💼 Promote your job with us. Reply to let us know.
🤦 ️ AI Funny Of The Week
We all love AI, but it’s certainly far from perfect 🤔 …
👋 Mike & Erin