Top 10 AI Prompts and Use Cases and in the Government Industry in New Zealand

AI is now central to modernising New Zealand's public services: the Government Chief Digital Officer is leading a practical, trust-first rollout via the Public Service AI Framework and fresh GenAI guidance to help agencies innovate while protecting people and data (New Zealand Public Service AI Framework guidance).

The national AI strategy launched in July 2025 sets out an economic and skills agenda - citing that generative AI alone could add NZ$76bn by 2038 - and aims to remove barriers like regulatory uncertainty and limited AI capability while building public trust (New Zealand national AI strategy launch).

For public servants and suppliers this means pairing clear principles (human-centred, transparent, risk‑based) with hands-on training so AI improves outcomes - faster triage, fewer delays - without losing accountability.

Bootcamp	Details
AI Essentials for Work	15 Weeks; practical AI skills, prompt writing, workplace applications; Early bird $3,582, then $3,942; Register for AI Essentials for Work (15-week AI Essentials for Work bootcamp)

“Use of AI technologies to improve public services is a priority for me, and this guidance will enable its safe and responsible uptake,” Ms Collins says.

Selection of the top 10 prompts and use cases followed a practical, New Zealand‑focused filter: prioritise public‑interest impact and feasibility, align with the Government's light‑touch, risk‑based approach to adoption, and favour solutions that create clear economic or service value in sectors already showing strong AI returns (for example, agriculture, healthcare, manufacturing and finance).

Criteria were adapted from international challenge frameworks - notably the “AI CONVERGENCE” selection factors such as public‑interest benefit, international outreach, disruptive relevance, value creation and stakeholder mobilisation - while procurement and deployment checks referenced the practical Algorithm Impact Assessment guidance for buying algorithms.

Emphasis was placed on ready‑to‑pilot prompts that reduce risk and speed outcomes (reflecting the Strategy's advice to adopt proven tools rather than build national models), plus measures for governance, transparency and human oversight drawn from New Zealand's responsible AI guidance so each use case is both ambitious and auditable.

The result is a compact, high‑leverage shortlist designed to deliver measurable productivity and public‑service wins quickly, rather than speculative long‑term bets; think rapid triage of the highest‑impact problems first, with safeguards built in.

Read the AI Convergence criteria and NZ procurement guidance for more on the selection rules.

“The AI Convergence challenges are unique opportunities to contribute to a world where AI concretely improves every aspect of our lives.” - Bruno Bonnell

Protecting Aotearoa's unique ecosystems starts at the border, and MPI's biosecurity screening roadmap is increasingly centred on scalable, automated detection - everything from mobile baggage X‑rays to camera systems that can spot pests smaller than 10 millimetres as containers move through a port.

Trials overseas show machine‑learning cameras can scan a container from multiple angles in four seconds, analysing millions of images and flagging roughly 4% of containers for follow‑up; New Zealand research and pilots are testing comparable approaches (camera imagery plus laser topology) to lift the proportion of containers inspected beyond the current c.5% sample rate.

Complementary work on 3D X‑ray and NIR imaging - already trialled at Auckland Airport and combined into shared image libraries - offers another non‑destructive route to find hidden pests in luggage, mail and fresh produce.

For public servants and suppliers, the pragmatic next step is piloting these proven detection tools alongside existing detector‑dog and profiling layers so high‑risk consignments are routed to people with clear audit trails and fast mitigation pathways; practical pilots make the “scan every container” ambition technically and operationally credible.

Read more on the Trellis BATDS trial and MPI's x‑ray trials for context.

Measure	Value / Source
Camera detection threshold	Smaller than 10 millimetres (Trellis)
Trial images analysed	1.7 million (Trellis)
Containers with pests in trial	4% (Trellis)
Approx. NZ manual external checks	~5% of sea containers (B3NZ)
MPI fixed x‑ray units	28 units for baggage and mail screening (MPI)

“It's about taking something that is near impossible as a human, to be able to detect small objects while containers are being constantly moved, while also dealing with rain and sun glare and rust on the containers,” - James Meszes

Health NZ can sharpen surgical decisions with locally tuned risk prediction: tools like nzRISK, built from national hospital data and already used at Auckland City Hospital, give clear 1‑month, 1‑year and 2‑year mortality estimates so clinicians and whānau can weigh benefits and harms with concrete numbers rather than guesswork; the model was developed from New Zealand patients and population data (covering around 99% of admissions) to make the outputs relevant across ethnicities and procedure types, and has been extended and validated in peer‑reviewed studies showing strong discrimination for perioperative mortality (NZRISK national study (PubMed)) - a vascular surgical variant achieved AUROCs of about 0.87, 0.83 and 0.82 for 30‑day, 1‑year and 2‑year mortality respectively - so the same predictive engine can drive smarter pre‑op triage, focused clinical decision support and reduced low‑value surgery while keeping clinicians in control; see the nzRISK calculator and clinician resources for how this is already being operationalised in New Zealand hospitals (nzRISK surgical risk tool for New Zealand patients (Orion Health video)).

Measure	Value / Source
Prediction horizons	1 month, 1 year, 2 years (NZRISK)
National data coverage	≈99% of hospital admissions 2011–2016 (Orion Health)
Vascular model sample size	21,597 cases (NZ vascular risk study)
Vascular model AUROC	0.869 (30d), 0.833 (1yr), 0.824 (2yr) (JVascSurg)

“NZRISK is an unbiased, highly accurate risk prediction tool. As a clinician, NZRISK provides me with valuable insights to help me and my patients in the shared decision-making process before surgery.”

Auckland Transport can make OIA compliance less manual and more reliable by automating the routine choreography around requests: automatic acknowledgements, deadline tracking for the statutory 20 working‑day response window, prompt routing when a transfer is needed, and clear publishing workflows so responses that are cleared for release are posted on schedule.

That approach dovetails with the OIA's guiding principle that information should be made available unless there's a good reason to withhold it, and with Cabinet Manual guidance on timely transfers and consultation between agencies - automation doesn't replace judgment, but it preserves audit trails, evidence of consultation and the “no surprises” checks that ministers and officials are expected to observe.

Practical wins include faster replies to Aucklanders, fewer Ombudsman complaints, and better data for the Directory of Official Information; see the Ministry of Justice guidance on OIA requests and the Cabinet Manual's Official Information Act section for the rules automation needs to meet.

Measure	Value / Source
Statutory response time	20 working days (Ministry of Justice)
Transfer timeframe	Transfer promptly and within 10 working days (Cabinet Manual)
Directory update frequency	Every two years (Ministry of Justice)
Publication timing for released responses	No earlier than 10 working days after response sent (Ministry of Justice)

In Auckland, Vector's rollout of Grid Aware - Google X's Tapestry asset‑inspection and planning suite - shows how AI can turn scattered visual feeds (helicopter, drone, street‑view and satellite imagery) into a single, searchable picture of the distribution network, slashing average inspection times from roughly 45 minutes to around five per asset and getting crews to faults before customers notice outages; that speed not only speeds preventative maintenance but seeded Tapestry's richer grid‑planning tool so local planners can simulate resilient, low‑cost upgrade scenarios rather than guessing from patchy records.

The human‑in‑the‑loop annotation process, where experienced Vector crews taught the model to recognise real field cues, keeps clinicians - sorry, crews - in control while letting ML spot subtle defects at scale, and the same approach (seen in PJM and other Tapestry partners) supports faster interconnection and scenario planning across transmission and distribution.

For New Zealand agencies and suppliers, the practical takeaway is clear: combine multimodal imagery, robust annotation workflows and MLOps pipelines to deliver measurable uptime and smarter, auditable asset decisions (Grid Aware Vector deployment in Auckland - Latitude Media report, Google X Tapestry project page, PJM and Google Tapestry collaboration report).

Measure	Value / Source
Average inspection time per asset	Reduced from ~45 minutes to ~5 minutes (Latitude Media)
Data sources combined	Helicopter, drone, satellite, street‑view and field images (Tapestry / Latitude Media)

“A field worker is going to do their job faster and more easily this afternoon in Auckland because of this machine learning work.” - Page Crahan

In Christchurch, organisations like Orbica can turn raw spatial data into practical, public‑facing tools - think living digital twins that help planners test development scenarios, ratepayers explore valuation and service footprints, and asset managers prioritise maintenance with 3D clarity; modern 3D GIS platforms make this possible by streaming large, interoperable datasets, visualising buildings and underground assets, and running spatial analytics for flood modelling, emergency response and risk assessment (3D GIS overview (ArcGIS)).

By combining elevation, LiDAR and textured reality meshes with interactive web scenes, councils can explain tradeoffs to communities (a councillor or resident can literally tilt and fly through a proposed streetscape) and expose transparent, auditable layers for rates and asset condition - use cases echoed across 3D GIS literature from scenario planning to maintenance optimisation (Top applications for 3D GIS data), while practical visualization techniques (LOD management, texture streaming and WebGL delivery) keep tools snappy for busy planners and public portals.

The “so what” is simple: better spatial insight means faster decisions, clearer public communication, and fewer surprises when capital projects hit the ground.

Fisheries New Zealand faces a classic surveillance gap at sea - industrial vessels that go “dark,” patchy patrol coverage and vast, costly waters to police - and the recent fusion of satellite imagery, synthetic‑aperture radar and AI now gives regulators a practical way to close that gap: global studies show strictly protected MPAs see far less industrial fishing when satellites and machine learning are applied, and combining AIS with SAR reveals many vessels that would otherwise slip under the radar (National Geographic article on satellite detection of illegal fishing).

Tools from organisations like Global Fishing Watch vessel-tracking platform turn billions of vessel positions and optical imagery into public, searchable maps, while commercial providers and radar constellations (ICEYE, KSAT and others) deliver near‑real‑time detections through clouds and at night - meaning enforcement can be targeted, cheaper and far faster than random patrols.

The so‑what is stark: satellites can spot a single “dark” trawler on the high seas the way a lighthouse spots a ship in fog, letting Fisheries New Zealand prioritise patrols, gather airtight evidence and protect coastal fisheries with much less guesswork.

“The ocean is no longer too big to watch. With cutting-edge satellites and AI, we're making illegal fishing visible and proving that strong marine protections work.” - Juan Mayorga

National Emergency Management Agency (NEMA) work in Aotearoa hinges on fast, joined‑up situational insight: rapid impact assessments carried out “within the first 8 or 48 hours” give responders a broad picture to steer immediate action, and the forms and workflows are designed to plug into electronic systems such as EMIS so damage, needs and priorities are recorded with audit trails (Civil Defence rapid impact assessment guidance (New Zealand)).

That front‑line data is fused with tools for rapid building safety decisions - MBIE supplies approved digital assessment forms and the recognisable white/yellow/red placards (Can Be Used / Restricted Access / Unsafe - Do Not Enter) that can appear on doors within hours to keep communities and crews safe - and these templates are available as Survey123 digital forms for fast field use.

For bigger picture planning and scenario work, NIWA's RiskScape provides a flexible multi‑hazard modelling engine used for rapid impact scenarios and Pacific PARTneR work, meaning planners can run likely loss and exposure scenarios rather than guess; pairing RiskScape with authoritative basemaps and the key datasets LINZ maintains (population grids, building outlines, elevations, roads and imagery) turns scattered feeds into a practical emergency map for response, recovery and future resilience investment (NIWA RiskScape multi‑hazard modelling software, LINZ key datasets for resilience and climate change (population, buildings, elevation, roads)).

Measure	Value / Source
Rapid impact assessment window	Within 8 or 48 hours (Civil Defence)
Rapid building assessment outputs	Approved digital forms + white/yellow/red placards (MBIE)
Multi‑hazard modelling	RiskScape software for scenario & rapid impact assessment (NIWA)
Key datasets	Population, buildings, addresses, elevation, roads, imagery; rapid building assessments not national (LINZ)

WorkSafe New Zealand's evolving approach makes automation and data-driven tools a practical fit for safer, faster inspections and clearer compliance: its position on the “use of new technology” expects PCBUs to adopt proven tools that eliminate or minimise risk, to check suppliers' safety work, and to engage workers at every step (WorkSafe New Zealand guidance on use of new technology); at the same time the regulator is shifting to a partnership model that rewards businesses who can show real-time, auditable safety practice rather than paperwork alone (WorkSafe partnership model and data-driven safety approach).

Practically, this means automated inspection workflows, digital hazard registers, remote monitoring and lone‑worker tech can reduce travel, speed follow‑ups and create the evidence WorkSafe now expects - while still meeting duties under the Health and Safety at Work Act to eliminate or minimise harm.

Recent coverage also highlights WorkSafe's real-world reliance on satellite‑enabled lone‑worker solutions and managed device fleets for inspectors working in remote or hazardous locations (WorkSafe lone-worker satellite-enabled solutions case study), a vivid reminder that automation isn't just back‑office efficiency but can keep an inspector safe on a distant ridge while authorities get instant, auditable situational data.

Measure	Value / Source
WorkSafe position	Expect PCBUs to adopt new tech to better manage risk (WorkSafe)
Planned inspector hires	Up to 60 new inspectors in 2025 (industry reporting)
Typical inspection time (SME)	About 1–2 hours (construction inspection guidance)

“Our close relationship with our monitoring station is key to what we do and how we operate. We know that they are 100% confident in what to do in the many different situations that may arise when monitoring lone, remote and isolated workers.” - Petra Hakansson

Castlepoint's Explainable AI brings a practical, audit‑ready layer to sensitive document screening in Aotearoa, letting agencies sift legacy emails, case files and shared drives for misconduct and abuse indicators much faster than manual review.

Built as a manage‑in‑place autoclassification engine trusted by government portfolios and regulators, Castlepoint helps protect vulnerable people by flagging high‑risk material while preserving provenance and compliance trails; real‑world deployments include automated discovery for abuse and misconduct investigations, where a NewZealand.AI case study reports keyword detection at 100% accuracy, and Castlepoint's own case studies catalogue similar work across health, local government and national security contexts.

For NZ agencies the value is immediate: faster investigations, reduced harm to victims, and clear, explainable evidence chains that stand up to scrutiny. See the Castlepoint Explainable AI case studies and the NewZealand.AI abuse detection case study.

like finding a needle of complaint phrases in a haystack of legacy records.

Use case	Result / Source
Automated detection of abuse & misconduct keywords	NewZealand.AI abuse detection case study reporting 100% accuracy
Enterprise records governance for government portfolios	Castlepoint Systems case studies

Classic Group can help New Zealand agencies turn stalled, paper‑heavy procurement into a tightly governed, auditable workflow that actually speeds service delivery: think structured requisition intake, real‑time budget and contract validation, configurable approval chains and ERP sync so POs aren't “lost on a manager's desk” but flow straight into finance systems for payment and reporting - no rip‑out of core ERPs needed (Ivalua purchase order automation guide).

AI‑enhanced invoice matching and anomaly detection cut duplicates and overpayments, while supplier portals and live dashboards give procurement and finance true spend visibility and faster reconciliations (ProcureDesk guide: Automating Purchase Orders (7‑Step Framework)).

The practical payoff for public servants is clear: fewer manual checks, stronger compliance trails for audits, and liberated staff time to focus on policy and service improvements - sometimes halving PO cycle times or better, depending on the pilot and scope.

“Having implemented Ivalua for 14 months, we can see that we are able to streamline our processes and achieve optimization. This comes with higher automation and autonomy, which is a big benefit for us. Working with Ivalua, we have the opportunity to work with better and higher quality data.” - Miljenko Galic

Practical next steps for New Zealand agencies and beginners are simple: start small, stay governed, and invest in people. Begin with a focused pilot or sandbox - test a single process end‑to‑end while applying human‑in‑the‑loop checks and an Algorithmic Impact Assessment from the start - then scale the wins; New Zealand's Public Service AI Framework and coverage of the framework explain why human oversight and transparency matter (New Zealand responsible AI framework - Global Government Forum).

Pair pilots with governance and legal checks (see MBIE's work on addressing uptake barriers and the Responsible AI Guidance) and expect to close clear skills gaps: only about 19% of Kiwi organisations are fully AI‑ready today, so targeted training is essential (Cisco AI readiness report for New Zealand).

For a practical learning route, consider a short, applied course that teaches prompt craft, human‑centred design and workplace AI workflows - Nucamp's AI Essentials for Work is a 15‑week, hands‑on option to build the governance and prompt skills needed to run safe pilots and hold suppliers to account (Nucamp AI Essentials for Work registration).

Think of early pilots as safe harbours: controlled, observable experiments that protect the public while proving tangible service improvements.

Bootcamp	Key details
AI Essentials for Work	15 weeks; practical AI for any workplace; early bird $3,582 then $3,942; AI Essentials for Work syllabus • AI Essentials for Work registration

“AI could add $76 billion to our GDP by 2038, but we're falling behind other small, advanced economies on AI‑readiness and many businesses are still not planning for the technology,” - Dr Shane Reti