AI in November

AI in engineering and New Zealand

Sign up for our next AI webinar Hear how Mott MacDonald uses AI in engineering design to build dynamic engineering apps which remove barriers to efficient calculations. They'll be sharing insights from their involvement in the Canterbury Multi Use Area project, showing how AI can speed up routine design tasks whilst keeping engineers firmly in control of decisions.

AI Forum AEC webinar series The AI Forum's AEC working group continues to bring together industry voices on practical adoption challenges. Check out their AEC focused webinar series under ‘our events’.

Māori leadership in emerging tech Kaupapa Māori and data sovereignty are shaping how Aotearoa approaches AI and quantum technology, from consent and guardianship through to governance structures. What's particularly encouraging is seeing practice move beyond policy statements into procurement settings, research protocols and capability pathways that grow Māori technical leadership. This isn't just theoretical work - it's changing how projects get commissioned and delivered.

No-code AI for small businesses Wellington-based Raygun launched Autohive earlier this year, a no-code AI automation platform targeting small and medium businesses. It lets non-developers spin up agents to handle everyday workflows and customer operations, lowering the barrier to practical adoption.

Ethics

Montreal AI Ethics Institute State of AI Ethics Report Volume 7 includes new case studies and frameworks for responsible deployment. Particularly relevant for engineers are the sections on accountability in automated decision systems.

Comparing AI lab transparency This comparison examines how OpenAI and Anthropic approach disclosure and governance differently. Understanding these approaches helps inform organisational AI policies.

AI tools and software advancements

Three years from GPT-3 to Gemini 3 Ethan Mollick provides a concise tour of how capabilities have shifted and what actually matters for practice. The key takeaway: design for verification and human-in-the-loop checks, not just benchmark peaks. Models are better, but they're not infallible.

DeepSeek releases two open models Chinese AI lab DeepSeek has released two high-performance open models. For teams wanting locally hosted models with strong capabilities, these provide serious alternatives to API-based services.

OpenAI declares ‘code red’ as Google catches up in AI race Internal reports suggest OpenAI is feeling competitive pressure as Google's AI capabilities narrow the gap. For users, this competition is driving faster innovation across both platforms.

Meet new Amazon Nova AI models that help build highly reliable AI agents Amazon has released Nova models designed specifically for building reliable AI agents. These are optimised for consistent, predictable behaviour rather than just impressive one-off demonstrations.

Introducing Mistral 3 | Mistral AI European AI lab Mistral AI has launched their third-generation model, continuing to provide strong open-weight alternatives for developers wanting more control over their AI infrastructure.

AI global

Interim International AI safety report Governments and labs are starting to converge on what good AI safety looks like – evaluations, red-teaming and incident reporting are becoming standard expectations. The challenge is that sector standards still lag behind the rhetoric, leaving individual organisations to figure out their own approaches.

Nvidia invests in Synopsys This US$2 billion investment signals that AI is being baked into the EDA toolchain used by hardware engineers. Expect AI-assisted place-and-route, verification and timing closure inside standard flows. For engineers working with these tools, plan now for traceability of AI suggestions to keep change control and accountability intact.

AI in science labs Robotics and AI are genuinely changing chemistry and materials R&D. Autonomous platforms can plan, run and learn from experiments, compressing development timelines dramatically. The catch is ensuring rigorous quality controls to maintain reproducibility and reliable datasets. Fast results are only useful if they're trustworthy.

One-minute deep research Opera Neon has launched an experimental mode that compiles fast, source-backed research briefs in about a minute. Rather than just aggregating search results, it synthesises information across multiple sources and provides citations. For engineers doing quick literature reviews or feasibility checks, this could significantly cut down initial research time. The key advantage is getting a structured overview with traceable sources, rather than wading through individual search results yourself.

Green innovations November saw several sustainability breakthroughs worth noting. Electricity-conducting concrete is moving from lab to field trials, offering potential for heated roadways and embedded sensors in infrastructure. Meanwhile, AI-optimised microgrids are demonstrating how machine learning can balance distributed energy sources more efficiently than traditional control systems. For civil and electrical engineers, these innovations suggest new possibilities for climate-resilient infrastructure design.

AEC software roundup The latest AEC Magazine covers significant updates across architecture, engineering and construction software platforms. Several vendors are integrating AI-powered clash detection and automated quantity take-offs into their core offerings. What's notable is the shift from AI as an add-on feature to becoming embedded in standard workflows. If you're managing BIM projects, the magazine's comparison of AI features across major platforms is particularly useful.

New Zealand spectrum update Radio Spectrum Management has extended the consultation period for 24-30 GHz spectrum allocation. This frequency range is crucial for next-generation connected devices and edge-AI applications, including autonomous vehicles and industrial IoT sensors. For engineers working on smart infrastructure or industrial automation projects, the outcome of this consultation will affect what technologies you can deploy and when. The extended deadline means there's still time to provide input if your work depends on these frequencies.

AWS Trainium Accelerator Amazon has released details on Trainium, their custom-designed AI training chip infrastructure. This is significant because it offers an alternative to Nvidia's dominant position in AI hardware, potentially lowering costs for organisations training their own models. The chips are optimised specifically for training large language models and computer vision systems. If your organisation is considering training models in-house rather than relying on API services, Trainium-based instances might offer better economics than traditional GPU clusters.

Robotic roundup

Robotics had a big month, with progress spanning from microscopic medical devices to factory floors.

Microrobots find their way Autonomous micromachines from ETH Zurich can navigate complex environments for targeted tasks. These aren't remotely controlled - they're genuinely autonomous at a microscopic scale.

A home robot that clears tables Memo, a new domestic robot, can clear tables and load dishwashers. Whilst we're still years from affordable home robots, this glimpse at domestic manipulation shows the technology is advancing beyond research labs.

BMW factory robotics BMW's flexible automation systems are contributing to mass-production milestones, showing how adaptable robotics can work alongside human workers in complex manufacturing environments.

Endurance record for humanoids Chinese researchers set a world record with a humanoid robot walking 105 kilometres non-stop. Long-distance walking benchmarks like this demonstrate improving energy efficiency and control systems.

Sand-grain sized drug-delivery bot Micro-scale robotics are opening new frontiers in medicine, with tiny devices that can navigate the human body to deliver drugs precisely where needed.

Lessons from a failed humanoid A Russian humanoid robot project failed spectacularly, offering valuable lessons about the importance of robust testing and realistic expectations. Not every robotic venture succeeds, and understanding failures helps the field progress.

Robot delivered takeaways Uber Eats is expanding pilot programmes for robot deliveries, raising practical questions about human-robot interaction in public spaces. How do pedestrians react? What happens in bad weather? These real-world trials are providing answers.

Interesting reading

Neurodiversity 101:The Fusion of Brains and Machines Professor Amanda Kirby explores what brain-computer interfaces and AI mean for neuro-inclusion. As these technologies advance, how do we ensure they support rather than exclude neurodivergent people?

AI's Path Ahead: Reinforcement Learning Environments IEEE Spectrum explores how reinforcement learning environments are shaping the next phase of AI development. Rather than just making models bigger, the focus is shifting to teaching AI through interactive simulations where machines learn by doing, failing and adapting in realistic scenarios.

Give it a go

Complete Guide to Nano Banana Pro Google AI Studio shares 10 tips for professional asset production using their latest tools. Worth experimenting with if you're creating technical illustrations or marketing materials.

We welcome member case studies, tips and resources that would benefit the wider engineering community. If there is a topic you want covered next month, let us know. You can look out for our next newsletter early February 2026.

Get in touch