AI in September

AI New Zealand

Engineering New Zealand AI webinars Our latest webinar, a panel Q&A, brought together academic leaders, teaching staff and a recent graduate to explore how engineering education is responding to the challenge of AI, and what this means for students and employers.

AI Education New Zealand A local hub offering bite sized, free courses to build practical AI capability across Aotearoa. Intro modules cover agentic AI, prompting, risks and ethics, with a forthcoming micro-course on Māori data sovereignty for AI led by Te Taka Keegan.

Albatross research with AI conservation tools Te Herenga Waka – Victoria University of Wellington researchers are applying AI to monitor endangered southern royal albatross on Campbell Island, aiming to scale seabird conservation while reducing costly field expeditions. This is a useful example of AI augmenting environmental science and long-term species monitoring.

ASB–NZ Product Accelerator links businesses to AI talent ASB has partnered with the New Zealand Product Accelerator to place Master’s students and interns in AI, data science and business analytics with ASB business customers, aiming to lift productivity through practical, short-term projects.

New Zealand Institute for Advanced Technology launches major AI investment The New Zealand Institute for Advanced Technology has announced a major programme to accelerate artificial intelligence research and commercialisation, signalling new funding pathways and partnerships for applied AI across universities and industry.

Meet KAT Hybrid New Zealand company Kara Technologies has introduced KAT Hybrid, a human-in-the-loop sign language translation service that combines its AI engine with review by a Certified Deaf Interpreter to deliver fluent, culturally accurate American Sign Language for recorded or scheduled content.

Claude in Aotearoa Anthropic’s September economic index places New Zealand in the global top five for per-capita Claude usage (around 4.05× our working-age population share), alongside Israel, Singapore, Australia and South Korea. The signal for local organisations is to lean into capability building, governance and measurable pilots that emphasise human-AI augmentation over unchecked automation.

Highlights from the recent AI Summit This newsletter from Caelan Huntress captures key takeaways from the recent AI Forum summit in Aotearoa. Sessions stressed governed, high-ROI use cases, the heavier compute needs of reasoning models and upskilling so AI lands inside regulated workflows.

AI global

RICS releases global AI standard for surveyors The Royal Institution of Chartered Surveyors has issued its first global standard on responsible AI use in surveying, effective March 2026. It requires training, tool registers, due diligence, client disclosure and explainability checks. For New Zealand engineers working with surveyors, it signals growing expectations that AI-enabled outputs will be documented, tested and transparent in joint projects.

Podcast: for the agent era, work charts beat org charts A short Spotify episode arguing that as AI agents take on more tasks, organisations should map work as flows (who/what does what, with which tools and data) rather than just reporting lines. Handy framing for engineering leaders planning pilot projects or governance for agentic tools.

Autonomy moves into the browser Agentic browsing is emerging with Copilot Mode in Edge. Assistants that can read pages, compare options and automate routine web tasks under supervision.

AirPods as in-ear translators An opinion piece argues Apple’s new live translation in AirPods could be a profound shift for everyday communication, noting the current requirements and limited language support at launch.

Tesla FSD 13.2.9 in Auckland inner city Short local drive test that shows how Full Self-Driving (supervised) handles inner-city Auckland: lane selection, parked-car squeezes, roundabouts and roadworks. Useful reference for members tracking how the latest 13.2.9 behaves on New Zealand geometry and road rules, and what interventions are still needed.

Robotic roundup

Factory-in-a-box robots TechCrunch profiles MicroFactory, a San Francisco startup building a benchtop “robot factory” about the size of a dog crate, with two arms that learn by watching humans rather than needing heavy integration. For New Zealand manufacturers, this points to lower-cost automation for short runs and rapid prototyping that could fit on a workbench instead of a factory line.

Diving robot targets harbour waste Germany’s TUM has unveiled an AI-guided diving robot that spots seabed litter with sonar and cameras, then lifts debris up to 250 kilograms using a four-finger gripper, demonstrated in the port of Marseille. For New Zealand councils, port companies and aquaculture operators, this points to a practical option for post-storm clean-ups, marina maintenance and net/gear retrieval in murky conditions where diving conditions are slow or unsafe.

Bomb disposal robots add haptic ‘touch’ The United Kingdom has begun fielding L3Harris T4 bomb disposal robots, a compact, high-agility platform with high-definition cameras and haptic feedback that lets operators ‘feel’ what the manipulator is doing, useful for delicate renders and cramped urban spaces.

AI tools and software advancements

Image generation and editing Google’s Gemini image tools show improved identity and scene consistency for concepting design options and stakeholder visuals.

Autodesk unleashes neural CAD Autodesk is introducing foundation models trained on professional design data and integrating them into Fusion and Forma, with bold automation claims and a re-think of geometry creation workflows. It’s worth tracking for impacts on design iteration, standards compliance and auditability.

Forma Building Design targets BIM Autodesk’s new Forma-based offering aims at LOD 200/300, AI-assisted modelling and integrated analysis, while tightening links between Revit and the Forma cloud. Expect more connected, cloud-first workflows across concept and detailed design.

Perplexity adds AI-powered email assistant The startup’s new Email Assistant lets users schedule meetings, organize emails and draft replies using AI. The tool is limited to the $200-per-month (max plan) and currently supports only Gmail and Outlook. The AI doesn’t train on user emails, but it can mimic their writing style when drafting responses.

Anthropic’s Claude is coming into everyday Office workflows. Microsoft 365 Copilot now supports Claude models (including Claude Sonnet 4) across Copilot experiences such as Researcher and Copilot Studio, giving Word, Excel, Outlook and PowerPoint users a multi-model option alongside existing OpenAI models. Rollout is underway, with model selection visible in supported Copilot entry points.

Security spotlight

The lethal trifecta for AI agents Simon Willison names the security anti-pattern of combining private data, untrusted content and external tool access. The takeaway is clear: don’t let agents that can read the open web also see secrets or call powerful tools – instead, split responsibilities, add allow-lists and treat content as hostile by default.

AI browsers could leave users penniless Malwarebytes explains how indirect prompt injection planted in ordinary web pages can hijack “AI browsers” and trigger harmful actions. Useful talking points for engineers include isolate agentic features, gate high-risk actions behind explicit user approval, and strip or sandbox third-party content.

OpenAI admits AI hallucinations are mathematically inevitable Computerworld reports OpenAI’s new study argues large language models will sometimes produce plausible but false answers even with perfect data, due to statistical and computational limits. The paper also says common benchmarks reward guessing over uncertainty, which worsens the issue, and notes some advanced reasoning models hallucinated more often. The authors propose explicit confidence targets rather than promising elimination.

Ethics

Companionship chatbots aren’t just a dependency risk Montreal AI Ethics Institute highlights a quieter harm: social comparison that can erode real-world relationships and wellbeing.

Give it a go

Claude for spreadsheets
Claude’s new create and edit files features let you upload Excel files, ask for pivot tables, charts, summaries and even export tidy workbooks and slide decks, handy for quick exploration before you formalise analysis in your standard tools. Practical guides and courses are popping up if you want a fast start.

Image editing gets a big upgrade
Google’s Nano Banana (Gemini image generator/editor) is earning praise for identity-consistent edits and better instruction following. Your first two images are free – make sure you spend time refining your prompt first.

Check your AI readiness with one of these free tools
Microsoft AI readiness wizard – Free tool that covers strategy, data, people and governance, and gives next-step guidance that is vendor-agnostic enough for SMEs. Microsoft Adoption

Changeable NZ – AI readiness and maturity assessment tool that is local, lightweight and tailored to New Zealand organisations. It is good for getting the team talking about gaps without consultant-speak.

Get in touch

As always, we encourage members to share AI insights, case studies or resources that could benefit the wider engineering community, or let us know what resources you would like to see us provide.