Digital AI and Physical AI in Healthcare: What New Zealand Should Prepare For Now

Written by Lisa Tamati

• 

Posted on November 20 2026

I dedicated my recent conversation with Paulo Osorio to my mum, Isobel, who died recently. You can find the full episode on the episode page, and you can watch it on YouTube. Paulo is a Partner in Technology and Transformation at Deloitte and leads Deloitte NZ's work on physical AI and robotics. This article takes one idea from our talk: healthcare needs two kinds of AI, and we need to prepare for both now.

Two kinds of AI, one problem

Most people think of AI as the digital kind. It is the chat tool, the search, the bot sorting your email. Paulo Osorio, Deloitte NZ, says that is how most people picture it. It will lift healthcare too, once it runs through the system.

But he says digital AI only goes so far. The second kind is physical AI: machines that can see a room, make sense of it and do a task in the real world. Paulo's point is that a hospital needs both.

Why records come first

When Mum was moved from Wellington Hospital to New Plymouth, half her records did not come with her. The team in New Plymouth did not know she had two unsecured aneurysms in the gut. I did not know either. I had power of attorney and still could not see the records. I found them through my local doctors.

Paulo says this is not just a New Zealand problem. Very few countries have connected health data. Your record should follow you wherever you go, even if you have an accident on a ski field in Christchurch. He says there are plans to fix this in New Zealand and that we should back them.

Why hands matter just as much

Paulo says that for half the healthcare workforce, something like 30 to 40 percent of the work is physical. That is moving things, moving equipment and moving people. These are skilled people who should be at the bedside, and they are doing logistics because we are short of staff.

I raised the WHO figure of about 11.1 million healthcare workers short by 2030. Everyone is chasing the same pool of international nurses and doctors.

Physical AI is not only humanoid robots. Paulo pointed out that smart watches and rings already make remote monitoring possible. Connect that data to AI and it can flag when someone needs help, and when they should be left to sleep. I think about Mum in that hospital room. She could not push a button. There was nothing watching over her. If something had been monitoring her blood pressure and oxygen, someone would have known.

Where to start

Paulo is clear that he is not asking for scaled adoption in 2026 or 2027. He is asking us to prepare for 2029 and 2030. His suggestions were practical:

  • Start with non-clinical tasks such as moving supplies, linen, food, medicine and rubbish. I mentioned the Moxi robot in US hospitals, which has done over a million deliveries.
  • Run small pilots in one place, with the legal, security, data and clinical people in the same room.
  • Measure cost per task completed, and also waiting times and clinical outcomes.
  • Bring frontline staff in early. Paulo says they are the ones most in favour, because they are exhausted from walking and doing laborious tasks.

On cost, Paulo said Goldman Sachs cut its estimate for building a humanoid robot by 40 percent in one year. He also said the price of AI inference has fallen about 280 times over a couple of years. He is firm that this is not about replacing healthcare workers. It is about giving them capacity to do the things only a human can do.

Why New Zealand has a window

We are small. Policy makers, health leaders, technologists and clinicians can be in the same city. Paulo says that makes it easier here than in the US, the UK or Europe to agree on pilots and how to measure success. Robots will also need real-world healthcare training data. A country that helps with pilots can get to the front of the queue.

I come at this as someone who cared for Mum for years. The cost to me was enormous, and most families cannot do it. Paulo and I both want a system where care is not limited by the number of hands available. Start small, keep the same North Star, and get off the start line.

Connect with Lisa

Newsletter https://www.lisatamati.com/lisa

Podcast https://www.lisatamati.com/ptl-podcast/

Shop https://shop.lisatamati.com

Books https://shop.lisatamati.com/collections/books