Physical AI is here - and it’s already changing the world

Sunday, 16 November 2025

The type of artificial intelligence we are yet to see take shape in New Zealand and this part of the world is physical AI - advanced AI that lives in machines and robots, powering them to move, act and integrate themselves into the real world.

Some - less futuristic - use cases of physical AI include self-driving cars, warehouse robots and drones, but far more advanced versions that could look human-like may soon be among us, and, experts say, in the not too distant future.

Physical AI is expected to be deployed and heavily used in industries such as construction and building, manufacturing and logistics - sectors that already have Internet of Things sensors at use within their operations and factories.

Amazon uses robots in their warehouses and fulfilment centres to stack shelves and pick and pack online orders across the United States, Europe and some locations in Australia.

In China, uses include standalone robots and wearable robots that augment humans at work. For example, a backpack with prosthetic robotic arms and so more physical arms to help pick up more items from conveyor belts.

Physical AI is also present among biotechnology and neural implants and will increasingly become more so, as a way to provide greater accessibility, offering the ability to control everything from computer games to robots packing shelves with the mind.

In Japan, some people with disabilities are already using neuro implants to control robots so they can have jobs working in grocery stores that may not of otherwise been able to have.

It is predicted that by next year, people will be able to have a neuro implant inserted (by a robot) for the price of an Apple Watch.

Datacom director of artificial intelligence Lou Compagnone describes physical AI as “living intelligence”. She says it is not just using robots in isolation, but rather “robots with senses in their physical environment, so that these robots are able to sense what's around them and actually learn through observation”.

That is already playing out overseas in countries such as the United States, China and Japan, she says.

When it may debut in New Zealand, she is unsure, but says it will take a huge amount of investment here.

“If we look around different places in the world, we're seeing different horizons of AI that haven't quite hit us. It's easy when you're not surrounded by that to go, ‘Oh, it's progressing, but it's not very fast’, but that’s just because we're not surrounded by that level of technological advancement.”

Generative AI (and so agentic AI) is the most common type of AI used in New Zealand currently, but “the more that you start looking out for what's coming; it gets more interesting, but also weird”, says Compagnone.

She says physical AI is all about finding problems to solve in unconventional ways.

Compagnone says many organisations are currently stuck in the adoption phase of AI, and so not yet adapting or realising its potential.

“People are adopting what I call personal productivity tools with AI. They're rushing to use tools like [Microsoft] Copilot and ChatGPT, those more foundational models, and they're expecting to see big gains from it, but actually it's only sort of creating incremental changes in the organisation.

“Mature organisations are starting to think about what they could shift from and to as a result of AI. They're thinking about it more as an organisational transformation, and starting to identify what I call functional productivity opportunities; areas where they might not have efficiencies and they can see a great opportunity for AI, and then identifying good use cases where they they could either have an assistant. so something to augment the way that someone does something or optimise.”

Datacom’s State of AI index reports 46% of New Zealand firms are just exploring AI, piloting projects, 33% are implementing it to use day-to-day, and only 8% are transforming because of it.

Microsoft estimates 38% of working age adults in New Zealand are using AI.

But AI is impacting workforces around the world. In the US, for example, there are less entry-level jobs advertised, believed to be the result of widespread ans more advanced adoption.

“[In New Zealand] we're still thinking about it as a tool rather than as an opportunity to transform,” says Compagnone.

“If you think of AI as being owned by your IT department or tech services area, it's just going to be a tool that you have available to deploy across the organisation. The big shift comes when you move away from that and start thinking about it in different business units and as part of an organisational strategy.”

Spark data and AI lead Anshuman Banerjee thinks of physical AI as advanced AI - agentic AI - and robotics coming together to help each other out.

What that looks like is someone “can talk to these human art robots in natural language”, say what task they want done, and then it physically does the task, such as clean the dishes, for example.

AI robots to help around the house, is already being developed and investigated by a number of companies, he says.

“Robotics and large language models are coming together, and we'll see some of these things happening in the near future,” says Banerjee.

Companies working on developing robots that can ask questions and ask tasks to be done with prototypes in test modes include Tesla with Optimus, its general-purpose robotic humanoid under development.

Spark is seeing productivity gains from using generative and agentic AI in its business, but Banerjee says it was yet to consider ways it could implement physical AI into its operations.

He believes AI robots will be commonly used in the workplace within five years, although he is not sure that would extend to New Zealand.

Basic levels of AI adoption is not yet widespread across New Zealand, namely among smaller firms, so it could be some time before that is a reality in this part of the world, he says.

While some experts call that problematic, and puts the country at risk of being left behind, Banerjee does not believe being slow to adopt AI technology is a concern.

“We don't need to be the front runner. But if something is working, and is leading to better productivity and making life easier so people can concentrate on things which are of value, I think we should bring it in.”

Use cases

Kenny Lee, co-founder and chief executive of Washington-based start up Aigen, has developed a Mars rover-type robot for the agriculture sector that uses advanced AI to identify and remove weeds mechanically.

Powered solely by solar, the robot is physical AI that aims to reduce the amount of pesticides farmers are using, and save farm owners labour costs, enabling them to perform a task staff would otherwise need to do.

Lee began developing the technology in late 2022, and currently has 50 of its robots deployed, working on six large farms producing tomatoes and pima cotton crop in North Dakota and California.

A single weed-removing robot from Aigen cost US$50,000 ($88,680).

Aigen is looking to take its business global with partners, and has its eye on Australia and Brazil as its first markets of expansion outside of the US.

Lee says physical AI has the ability to transform industries, such as farming, able to make life easier for employees and business owners.

Amazon Web Services chief executive Matt Garman told The Post, he was “very bullish” about the opportunities physical AI would bring for businesses.

“There’s huge potential, and it’s one of the next frontiers where AI is going to have an enormous impact,” Garman said in response to questions about the intersection between advanced AI technology and physical machines.

He says there will come a time when the cloud computing business is able to offer machines as part of its product services.

“We have many, many robots deployed inside of [Amazon] fulfilment centres today, and are one of the largest robotics companies in the world because of that. We’re very attuned to the potential for robots and very excited about it. It’s an area that we focus on a lot; we think about it a lot, and I’m very bullish about it.

“From an AWS perspective, we get to learn from our retail [business] about how do you think about robotics, how do they think about physical AI on top of their robotics, and we have the potential to somehow, some day, expose that to our customers as well.

“It's an area over the next couple of years that I think will progress and we’ll see a lot more.”

Rahul Pathak, vice president of data and AI at AWS holds a similar view. He says there will come a time when AI built into physical machines will be commonplace across most workplaces.

“I can definitely imagine AI applied to humanoid robotics. Will we see people humanoid robots sitting and typing away keywords? I don't know if that's the most efficient way to use AI and robots, but we will see more everywhere that we use robotics today; in logistics, manufacturing and health and safety,” says Pathak.

“The big power of agentic AI is the ability to reason about the situation you're in, and so robots will get a lot smarter about dealing with unknown situations.

“Anywhere that we're using robotics in earnest today is an opportunity to infuse that with AI.”

Aimee Shaw travelled to Seattle courtesy of Amazon.