Photo: LumiNola/Getty Images
An Indigenous approach to implementing an AI-powered productivity assistant within Hauora Māori Service could offer a blueprint for scaling generative AI across Te Whatu Ora Health New Zealand's 80,000-strong public health workforce.
Hauora Māori's digital team adapted Microsoft 365 Copilot — internally dubbed BroPilot — to reflect tikanga Māori (an Indigenous practice) and support daily reporting, governance, and programme work. The tool is now being used by both Māori and non-Māori staff across the service.
The culturally grounded implementation was designed to ease heavy workloads while promoting AI adoption without imposing a standard technology rollout.
The digital team developed BroPilot around trust, local ownership, and hands-on learning, responding to initial staff hesitancy and scepticism about AI by allowing them to test the tool on real work in a secure enterprise Copilot environment.
Working with more than 200 specialists with expertise in tikanga Māori, the team used Māori language and knowledge resources, including Dame Naida Glavish's "Tikanga Best Practice Guidance," as the basis for customising the platform and having Copilot create 16 measurable standard operating procedures grounded in Māori culture and values.
To build confidence in using the AI assistant, the digital team conducted regular Monday and Wednesday drop-in sessions where staff could use Copilot on their own documents and test prompts. BroPilot has also been used to summarise long documents, extract actions, refine reports, and turn dense material into clearer outputs for review. Some staff have used it to create role-specific personas, including assistants for research, kaupapa-centred advice, daily tasks, and executive-level documents.
Beyond administrative work, BroPilot has been used as a cultural safety support tool, helping staff from different backgrounds sense-check what tikanga Māori may apply in healthcare scenarios, including end-of-life care.
In an interview with Healthcare IT News, Troy Baker, senior ICT specialist at Hauora Māori Service, discussed how BroPilot was designed, governed, and introduced to staff, as well as the lessons from its Māori-centred rollout that could inform scaling genAI across the wider public health system.
He also shared the practical use cases emerging from the deployment, the role of cultural safety and trust in AI adoption, and what still needs to be addressed before an Indigenous-led approach can be expanded more broadly.
Q. What, specifically, would need to be in place before Health NZ could scale the BroPilot model more broadly across the organisation's 80,000 workforce, and how would you adapt the change management approach for teams with very different workflows, digital maturity, and cultural contexts?
A. Our experience delivering BroPilot through the ongoing Monday and Wednesday Copilot drop‑in sessions was that the biggest requirement for scale is not technology – it is enabling conditions for trust, safety, and accountability.
Before scaling, three things need to be in place. First, a safe enterprise environment with clear guardrails. Copilot has worked because staff trust that Enterprise Copilot operates within Health NZ security, privacy, and data boundaries. That safety creates permission to experiment and learn without fear – something repeatedly reinforced in the Monday and Wednesday drop-in sessions.
Second, local ownership rather than central prescription. The sessions deliberately meet people where they are. Some participants are advanced and building agents, while others are new and openly anxious. Scaling BroPilot requires enabling local champions, allowing teams to shape how Copilot fits their workflows, cultural context, and risk tolerance, rather than rolling out a standardised script.
And third, time‑protected learning embedded in real work. What works is not training detached from reality, but hands‑on use with actual documents, reporting, risks, and community kaupapa (purpose). The drop‑in model demonstrates that capability builds fastest when experimentation is anchored in everyday mahi (work), rather than generic examples.
Change management must therefore be adaptive, relational, and phased, recognising that digital maturity, clinical risk, and cultural context vary widely across Health NZ. Scaling BroPilot means scaling principles, not forcing uniform behaviour.
Q. The BroPilot rollout was described as values-led and built around hands-on experimentation with real work. What have you learned so far from it that will shape the next phase of deployment?
A. BroPilot was developed to encourage the slow adopters to gain confidence, build trust and excel in their own positions. It is an alternative to a mainstream method of learning – one size does not fit all.
The strongest insight we gained is that confidence comes from relevance, not instruction. Across the Monday and Wednesday drop-in sessions, people move from hesitation to regular use once Copilot helps them solve something real, whether that is a reporting burden, a confusing brief, a capability matrix, or community engagement notes. Once staff see Copilot assists without taking over, trust follows.
Another key lesson is that values must be operationalised, not just stated.
BroPilot works because tikanga Māori principles – mana, wairua safety, tino rangatiratanga, whakapapa – are actively built into how the tool is introduced, discussed, and governed. Participants are repeatedly reminded that Copilot supports them, but responsibility remains with the human.
Finally, the rollout shows the importance of intentional imperfection. BroPilot is not positioned as flawless or authoritative. That reduces fear, invites critique, and keeps users engaged as responsible decision‑makers rather than passive consumers of AI output.
These insights shape the next phase toward deeper role‑based use, clearer patterns of good practice, and stronger cultural safety scaffolding as adoption widens.
Q. A recurring challenge with genAI rollouts in healthcare is moving staff from curiosity to sustained, accountable use. How are you planning to manage that transition at scale, particularly around training, governance, prompt practices, human oversight, and resistance from staff who may still worry about trust, overreliance, or cultural missteps?
A. From our weekly observations, the shift from curiosity to maturity happens when accountability is explicit and supported.
Training must move beyond “how Copilot works” into how to think with it. In sessions, we focus on prompting as a thinking process – providing context, intent, and constraints – and on checking outputs critically, not accepting them at face value.
A core message reinforced consistently is that Copilot does not own decisions; people do. Whether they are writing reports, assessing grants, or summarising community kōrero, users are reminded they remain accountable, just as they are when using calculators or spreadsheets.
Sustained use improves when teams share prompts, agents, and patterns that reflect their values and workflows. This has emerged organically in sessions through prompt diaries, shared agents, and collaborative learning rather than top‑down enforcement.
Concern about trust and cultural worries is often grounded in valid concerns – about over‑reliance, misinterpretation of Māori kōrero, or cultural misuse. These are addressed directly, particularly around transcription, Māori data sovereignty, and consent. Trust grows when people see that they can opt in, opt out, and shape boundaries collectively.
This work has built on the HealthX Microsoft Copilot initiative, which started in August last year and, since January, has expanded to more than 2,050 licences across Health New Zealand. HealthX has helped enable and coordinate a managed M365 Copilot rollout by identifying priority use cases and working closely with senior clinical executives, Hauora Māori Service, and Digital Services.
This has ensured Copilot licensing could progress within existing enterprise controls, with deployment carefully sequenced to leadership and digital teams first. Adoption was supported through Microsoft‑provided webinars and online learning resources, with HealthX capturing early learnings to inform future scale.
_
Editor's note: Baker's responses have been edited for clarity. Italics are the author's emphasis.
