A coach for the gaps between appointments
People with type 2 diabetes see a clinician now and then. In between, they're the ones managing meals, exercise, medication and glucose every day, mostly on their own. Researchers at the Auckland Bioengineering Institute (ABI) at Waipapa Taumata Rau, the University of Auckland, are building an AI coach for those in-between weeks. It's called DiNa, and the university announced on 13 November 2026 that it's now in a small clinical trial.
Dr Gonzalo Maso Talou leads the project, working with PhD student Kekayan Nanthakumar and others. The team keeps its goal narrow. "DiNa is not about curing diabetes," Maso Talou says. The coach is there to help people keep up the habits they've already agreed with their healthcare team, long after the motivation from the clinic visit wears off. He describes the longer-term aim as helping people "make sustainable lifestyle changes and stay engaged with their care."
There's plenty of need here. Health New Zealand's Virtual Diabetes Register counted more than 365,000 people with diabetes in Aotearoa in 2025, a rise of over 40 percent since 2011. Māori, Pacific and Indian communities carry a disproportionate share of that burden.
What the trial looks like
The first round is small. According to the university, 10 adults with type 2 diabetes are taking part for three months. Half have been given DiNa and the other half an existing diabetes management tool, so the team has something to compare against.
The recruitment listing on the Diabetes New Zealand website gives more detail:
- People are randomly assigned to DiNa or to mySugr, a popular diabetes logging app.
- It runs for 12 weeks, and participants take part from home.
- Each person borrows a continuous glucose monitor (CGM) and a smartwatch at no cost for the length of the study.
- Participants get a koha of up to $280 in vouchers.
- To take part you need to be 18 or older, have type 2 diabetes, live in the Auckland area, be comfortable using English, and have a compatible phone or tablet with reliable internet.
- The study runs under Health and Disability Ethics Committee approval 2025 FULL 20497, titled as a pilot of digital twins for managing chronic metabolic disease.
It's the first of three planned rounds, and bigger studies are scheduled for 2027. So far the university reports positive responses, with participants saying monitoring got easier and felt like less work. That's early, informal feedback from a handful of people. No outcome data has been published yet.
How it works day to day
DiNa brings together several streams of health data and lets people talk to it about them.
For food, users photograph a meal. The software identifies what's on the plate and estimates calories, carbohydrates, protein and other nutrients. If it misreads something or misses an item, users can correct it out loud. Activity comes in automatically from data phones and wearables already collect, so nobody has to log a workout. A quick check-in records whether prescribed medication has been taken, and glucose readings come from the CGM.
With all of that, the coach can suggest a good time of day to exercise, help someone work out what's blocking a goal, adjust a plan when life changes, and offer encouragement. The Diabetes New Zealand listing adds goal setting, progress tracking, and showing people how food, sugary drinks and missed doses affect their glucose.
Safety relies on keeping the coach on a short leash. The university says DiNa's answers come from curated New Zealand diabetes resources developed with clinical partners, and the trial listing names Healthify as one of them. When a question needs a clinician's judgement, DiNa refers the person to a health professional.
From a lifelike avatar to plain voice
The project's history explains an interesting design choice. In August 2022 the university reported that two ABI projects had won more than $4 million from MBIE's Catalyst Strategic Investment Fund, set up in 2021 for internationally connected research on AI in healthcare. One project covered type 2 diabetes and cardiovascular disease. Soul Machines, the New Zealand company known for animated digital people, was a partner, and the plan back then was a culturally appropriate "digital health navigator" that would appear on screen, read sensor data and talk with patients in real time.
Early prototypes did show virtual faces. Users told the team they'd rather just talk, finding voice more natural and less demanding, so the avatar was dropped. Today's DiNa is voice-only.
The digital twin underneath
DiNa sits inside ABI's broader digital twin programme, which Maso Talou co-leads with Dr Finbar Argus. In this case a digital twin is a computer model of how one person's body responds to food, exercise and medication. Current versions estimate glucose responses from a meal's carbohydrate, fat and protein. "Two people can respond very differently to exactly the same meal," Maso Talou says, which is why personal models could matter.
Ideas for later include ranking meal options by their likely glucose impact, suggesting recipes based on someone's current glucose and the ingredients they have at home, and letting people explore what a food swap might do to their numbers. The researchers mostly frame this as an educational tool. Data from the trials will feed back into refining the models.
ABI's twin research has wider government backing too. In September 2025 Science Minister Dr Shane Reti announced $4.5 million from the Catalyst Fund for a New Zealand and United States digital twin programme linking ABI with the University of Texas's Oden Institute. It focuses on physics-informed AI for health, clinical decision tools and agriculture. That announcement doesn't mention DiNa, so we can't say how much of that money reaches the diabetes work.
Costs and what's still unknown
The project has more than $4 million in MBIE funding. Participants pay nothing, since devices are lent and vouchers offered. There's no product, price or launch date. If results hold up, the team hopes DiNa could be built into health services and made widely available.
Several things a buyer or regulator would want to know haven't been made public:
- Which AI models handle food recognition, speech and conversation, and where they run.
- How accurate the photo-based nutrient estimates are.
- How the team tests that answers stay safe and that referrals to clinicians actually happen when they should.
- Where glucose, activity and voice data is stored and who can see it.
- Whether the tool improves adherence or health outcomes. Ten people over three months can't answer that.
- How it'll serve people who'd rather use a language other than English, given who's most affected.
Lessons for builders
Health tech teams in New Zealand and Australia can pick up a few practical points before any results land:
- Test the interface with real users early. A lifelike avatar sounded like a plus, yet users preferred plain voice. Finding that out early saved effort.
- Make logging nearly effortless. Photo capture with spoken fixes, passive activity data and one-tap medication checks attack the main reason people give up on tracking apps.
- Keep answers to vetted local content. Restricting a health chatbot to curated, clinically reviewed New Zealand material, with clear routes to clinicians, is a sensible way to manage risk.
- Pick a real comparator. Measuring against mySugr gives a more honest read than measuring against nothing.
- Budget time for ethics approval. A registered HDEC approval is what any health AI pilot here should expect to need.
We'll report back when the second and third rounds produce data.
Sources
- Researchers trial AI coach for diabetes (University of Auckland, 13 November 2026)
- Research participant opportunities (Diabetes New Zealand)
- How digital people could improve access to healthcare (University of Auckland, 19 August 2022)
- Digital twin NZ-US collaboration receives Catalyst funding (University of Auckland, 18 September 2025)
Build with people working on the same things
Join the Auckland Society for AI: NZ$599 a year with a ForgeBox and the members' projects area. Our Auckland meetups are open to everyone.