AI Last. AI consulting for health teams in Aotearoa.
You didn't train to copy and paste.
I help health teams remove paperwork first and bring AI in last. It costs less that way, and patient names stay out of AI tools. Start with a free 15 to 20 minute scan of where your team's hours go.
15 to 20 minutes, one person, no patient information. You get a one-page map of which task to fix first.
Or book a call first
02 / The old way
Most AI projects start by picking a tool.
So you are left choosing between fast and safe.
The paperwork in a small health practice keeps growing. ACC reports, clinical notes, funding forms, rosters, emails. Most of that time cannot be billed, so the owner or a senior clinician does it at night, after a full day with patients.
ACC and the funders add new forms. The professional bodies add new rules. The team adds its own meetings and checklists over the years. Very little is ever removed. And nobody counts the hours, so staff time gets treated as free.
Then someone says "let's use AI". In a small practice this often creates new problems. Staff try free ChatGPT on their own, each in a different way. Nobody checks the results. Patient information ends up in tools that were never approved for health information. The safer tool you already pay for, like Microsoft Copilot, feels slow, so people stop using it. One practice I worked with told me exactly that: Copilot was hard to use and slow, so staff went back to ChatGPT.
Use free AI and risk patient data, or keep copying, pasting and typing reports by hand.
AI put on top of a messy process does not fix the mess. You end up paying a licence to do, faster, work that should never have existed. A one-day AI workshop feels inspiring, and on Monday the work is the same. I have run those workshops, so I am guilty of this myself.
Microsoft Copilot: safer, but slow
Free ChatGPT: fast, but risks patient data
03 / The new rule
AI last.
A step you have removed costs nothing to automate.
Pharmacists have a word for it: deprescribing. You stop the medicines a patient doesn't need before you add a new one. I do the same with paperwork. First we find where the hours go. Then we remove the steps nobody needs and simplify what is left. AI comes in last, on one narrow task, after the work is clean. And less goes into the AI, so less can go wrong. That is why AI last pays off, and why it is safer.
I call the whole method UPSWING. The part that does the cutting is called SWIFT, and you will see it in step 2 below.
04 / How it works
Three steps. The first two are free.
After that, you pick the next task on the map, with me or without me.
Step 1. Free scan.
One person on your team fills in the AI Quick-Scan. It takes 15 to 20 minutes and works on a phone. It asks about your day, which admin tasks eat the most time, and how your top three are done now. No patient information. Within two working days you get a one-page Opportunity Map: your three tasks scored on payoff and risk, in your own words.
Step 2. SWIFT.
30 minutes, free, in person in Hamilton or on Google Meet. We go through your map together and start cutting your top task live with SWIFT, the method I named above. You keep the map and the cutting advice whether or not you hire me.
Step 3. First Win.
14 days, one task, fixed price. We remove and simplify first, then put AI on what is left. Hours are counted before and after, and the hours you save come back every week the tool runs. Your team works through it with me, so the AI skills stay in your practice when I leave.
05 / What I can count
What I can count
I am a new company with an old method.
Most of my counted numbers come from education, because that is where I worked until July 2026. Health is where I work now, and I have one health client's result so far. Here is everything I can count.
At one health provider, an ACC contractor, one report took 90 to 120 minutes to write. We cut the phone calls first, then brought in AI. In 3 to 5 test reports, each took 12 to 20 minutes, and each was checked against the version a person wrote. That was in December 2025, with older AI tools than today's.
108 educators at one teachers-only day in Horotiu, October 2025. 83% satisfaction.
157 science teachers across Auckland, Hamilton, Wellington, Christchurch and Dunedin, November 2025 and April 2026. After the April sessions the funder surveyed the room: 36 answered and 94% said their skills grew. The report is public.
AI training for 63 education leaders in 18 countries through EmpowerEd, in partnership with the Asia-Europe Foundation.
Re:Gen AI, 31 July 2026, University of Waikato. Over 100 registered, 86 in the room.
No client reviews yet. The counted results of my first clients will go here, with their permission.
Bachelor of Pharmacy, University of Otago. Cancer pharmacist and clinical-trials work before teaching. Physics and Chemistry teacher at Rototuna Senior High School until July 2026. Master of Technological Futures, academyEX, 2026. AI Associate, in partnership with BAIC Ltd. Lead, AI Collective Hamilton. AI Forum NZ working group member.
I work in partnership with BAIC Ltd, an AI governance firm led by two specialists in international risk management and ISO 42001, the international standard for managing AI. You deal with me. The team behind me helps check that what I deliver is safe.
06 / What you get
What you get
The steps nobody needs are gone, and AI handles most of the copying and pasting that remains.
90 to 120 minutes
12 to 20 minutes
- We cut the phone calls first
- then brought in AI
A person checks it before it leaves.
- Time back inside the working day. At one health provider, a report that took 90 to 120 minutes took 12 to 20 after we cut first and then brought in AI.
- Lower AI costs. You only pay for AI on work worth keeping. The tools you already pay for, like Microsoft 365, are often enough.
- Patient names stay out of AI tools. Your staff are trained on what is safe to enter. Names and identifying details come out before any text goes in. A person on your team checks every report before it goes to ACC or to a patient. This is in line with NZ health privacy rules.
- A number you can trust. We count how long the task takes before anything changes. Then we change one thing and count again. For each process you keep a one-page record: the new steps, the hours saved, and what the AI is and is not allowed to do.
- Your team stays. Before we start, we agree in writing that the time saved goes back to patient care. I train one person on your team to carry the work on, so the method stays with you after I leave.
Questions people ask me
Is my patient data safe?
Patient names and identifying details come out before any text goes into an AI tool. Your staff are trained on what is safe to enter and what is not. A person on your team checks every report before it goes to ACC or to a patient. We write down which tools are used and where the data is stored, and you keep that page. A patient can say no to AI and still be treated the usual way. This is in line with the Privacy Act 2020 and the Health Information Privacy Code. I was a cancer pharmacist who worked on clinical trials, and I treat your patient records the way we treated trial records: coded, and never in a place they do not need to be.
What does the Quick-Scan ask, and how long does it take?
15 to 20 minutes for one person, on a phone if you like. It asks about your business, walks through yesterday, lets you tick the admin tasks that eat time, and asks a few questions about how your top three are done now. No patient information. You get a one-page map back within two working days.
What does it cost?
The scan is free. The map meeting is free. A First Win on a low-risk task is $3,950 plus GST, fixed. You pay half to start and the other half only if we hit the target we set together, measured two weeks after the tool goes live. After that, some teams keep me on a monthly retainer. We price that per team.
What counts as high risk?
Anything that touches patient records, legal records, child-safety or vetting records, or any decision about a person. Those tasks need extra safeguards and testing, so I quote them separately after a risk check, and I do not guarantee a result on them. I will tell you which side of the line your task sits on at the map meeting.
What if cutting the paperwork fixes the problem and we never need AI?
Good. That happens. Every First Win still ends with a live AI tool, so if the cutting removes the whole task, the AI goes onto the next task on your map, in the same 14 days, for the same fee.
Will my staff lose their jobs?
No. Before we start, we agree in writing that the time saved goes back to patient care. I also train one person on your team to carry on the work, so the method stays with you.
We already pay for Microsoft Copilot. Do we need something new?
Probably not. The tools you already pay for are often enough once the work is simplified. One practice I worked with had Copilot and nobody used it, because it felt slow on a messy process. The fix was the process, and then the tool.
Who checks what the AI writes?
A person on your team, every time, before anything goes to ACC or to a patient. AI drafts. People sign.
Do you work outside Hamilton?
Yes, anywhere in New Zealand. Hamilton and Waikato in person, everywhere else on Google Meet. The scan and the map are the same either way.
I run an education organisation, or a business that is not in health. Is this for me?
The scan works for any team. I work with subject associations and education institutions on a different offer (professional learning and curriculum resources, no First Win), so book a call and we will talk about that instead. Other businesses are welcome to take the scan and see what the map says.
The deal, in plain words
The scan is free. The map meeting is free, and you keep the map and the cutting advice whether or not you hire me.
If we go ahead with a First Win on a low-risk task, the price is fixed at $3,950 plus GST. You pay half to start. You pay the other half only if we hit the target we set together, measured two weeks after the tool goes live. If the task turns out to be high risk, I will say so and quote it separately, because it needs extra testing and I will not promise a result on it.
Every First Win ends with a working AI tool, a one-page record of what changed and the hours saved, and one person on your team who can run it. There is no lock-in after that.
07 / Start here
Cut the copy and paste so you do the work you love. Your people deserve your time.
Start with the scan. 15 to 20 minutes.
You get a one-page map of where your team's hours go and which task to fix first. If you would rather talk to me first, I am happy to chat.