AI Forum NZ — Generative AI Working Group A living document
Living Whitepaper
Archived

You're reading an archived edition. Read the latest version from 1 September 2026.

AI in Agriculture in Aotearoa New Zealand: A Living Whitepaper

Introduction

AI adoption in New Zealand agriculture is becoming more practical, targeted and institutionally supported. The sector is not experiencing a single, uniform technology transition. Instead, AI is being deployed where it can improve a specific decision or physical task:

  • Managing livestock and pasture.
  • Predicting crop and fruit quality.
  • Improving irrigation and soil-moisture decisions.
  • Supporting biosecurity and disease surveillance.
  • Automating packhouse and orchard operations.
  • Reducing labour, emissions and input costs.
  • Connecting data across the food and fibre value chain.

Dairy remains the most advanced adoption segment, but horticulture, arable farming, viticulture, biosecurity and agricultural robotics are expanding rapidly. The major change since the previous update on 13 July 2026 is the strengthening of the adoption infrastructure around AI: government co-investment, commercial trials, farmer incentives, export-oriented robotics and sector-led programmes are increasingly translating research into deployment.

Executive Snapshot

  • AI adoption is shifting from experimentation to operational use. Halter, Aimer, Hectre and other systems are increasingly embedded in routine farm, orchard and packhouse workflows.
  • Dairy remains the leading New Zealand use case. Livestock wearables, virtual fencing, pasture modelling and AI-enabled decision support are the most mature commercial applications.
  • Reliability has become a central adoption issue. A July 2026 Halter outage and hardware issues demonstrated that connectivity, fallback procedures and service resilience matter as much as model accuracy. (farmersweekly.co.nz)
  • Horticultural AI is expanding from software into robotics. Hawke’s Bay manufacturer Hawk Technology reported strong United States demand for AI-enabled apple-packing robots, including 250 additional machines on order. (farmersweekly.co.nz)
  • Government is funding adoption, not only research. AgriZeroNZ’s Early Adoption Accelerator provides up to NZ$51 million in Crown funding, matched dollar-for-dollar by industry investment. (beehive.govt.nz)
  • Biosecurity is becoming an important public-sector AI use case. Biosecurity New Zealand is piloting generative AI to assist with the preparation of import health standards while retaining expert human decision-making. (beehive.govt.nz)
  • The strongest strategic opportunity is value-chain integration. New Zealand’s AI sector is increasingly focused on combining farm, processing, logistics, traceability and export data rather than optimising isolated farm activities. (technewzealand.org.nz)
  • The national AI research platform remains unresolved publicly. MBIE’s public page still describes the selection process and says an announcement timeline will be provided in due course, despite earlier expectations that the platform would be established in July 2026. (mbie.govt.nz)

What Changed Since the 13 July 2026 Update

1. Halter’s adoption story gained both scale and caution

Halter announced its largest investment in its beef product to date in July, building on its satellite-enabled virtual fencing system for remote and extensive cattle operations. The company’s technology combines collars, machine learning, farm data and satellite connectivity to support virtual fencing, animal monitoring and pasture management. (halterhq.com)

The company also announced six winners of its 2026 “One Year Free” programme. The farms include dairy and beef operations affected by flooding, succession challenges, labour constraints and farm expansion. The breadth of the winners indicates that AI-enabled farm systems are being positioned not only as productivity tools, but also as resilience and workforce-support technologies. (ruralnewsgroup.co.nz)

However, a July connectivity outage and hardware issue affected some Halter customers. Halter said the outage had been resolved and that affected systems moved into backup mode, but the incident highlighted a critical issue for physical AI: if a system influences livestock movement, network and hardware resilience become part of farm safety and operational risk management. (farmersweekly.co.nz)

2. AI-enabled horticultural robotics moved further into commercial export

Hawke’s Bay-based Hawk Technology is experiencing increasing international demand for its apple-packing robots. The company reported that 50 Gen 5 Apple Packers had been installed in Washington State, with another 250 on order. The machines combine robotics, computer vision, software and AI inference to automate packhouse tasks. (farmersweekly.co.nz)

The development is significant for New Zealand because it shows the sector generating exportable physical-AI products, not merely adopting overseas software. Hawk Technology has expanded from two research and development employees seven years ago to more than 50 staff, while retaining design, fabrication, electrical and assembly capability in Hawke’s Bay. (farmersweekly.co.nz)

The company’s experience also provides a more nuanced view of automation and employment. Its customers are using robotics partly to address labour shortages, while the manufacturer reports growing demand for workers with software, machine-code, mechatronics and AI skills. (farmersweekly.co.nz)

3. Government support is becoming more explicitly adoption-oriented

AgriZeroNZ’s Early Adoption Accelerator is designed to move emissions-reduction technologies from development into commercial farm use. The programme provides up to NZ$51 million of existing Crown funding, with matching industry investment. AgriZeroNZ had already invested NZ$79.9 million in 18 companies, research projects and trials by June 2026. (beehive.govt.nz)

Although not every supported technology is AI-based, the programme is relevant to AI adoption because it creates a pathway for data-rich livestock wearables, emissions measurement, predictive systems and other farm technologies to be tested under commercial conditions.

The June 2026 Responsible Dairy programme provides a similar mechanism within dairy. The seven-year, NZ$45.85 million programme will work with 35 to 40 partner farms, test stacked technologies and develop evidence for more productive and lower-footprint farm systems. Halter and Gallagher are among the technology partners. (dairynz.co.nz)

4. Biosecurity agencies are beginning to use generative AI

In July, the Government announced a four-month pilot of a generative AI tool to assist with the development of import health standards. The tool is intended to reduce document-heavy work and improve consistency, while key decisions remain with Biosecurity New Zealand experts. (beehive.govt.nz)

This is an important development because it expands agricultural AI beyond the farmgate. Biosecurity standards influence the importation of plants, animals and biological products, and therefore affect access to genetics, production inputs and new agricultural technologies.

5. Aimer’s pasture platform is moving from product to adoption programme

Aimer has received NZ$600,000 from MPI within a NZ$1.675 million project to test and refine its AI-enabled pasture-management system across hundreds of New Zealand farms. Project partners include Ravensdown, Cropmark Seeds and Fonterra. (mpi.govt.nz)

AIMER combines pasture measurements, growth rates and farm inputs to build a digital representation of the farm’s pasture system. It provides decision prompts such as when to move stock, cut silage or defer grazing. (mpi.govt.nz)

The commercial importance of the programme lies in adoption support. Farmers are not only being given access to an AI tool; the project is designed to demonstrate how it works with existing farm practices and data systems.

Current State of AI Adoption

Dairy and livestock

Dairy remains the most mature AI segment in New Zealand agriculture.

The leading applications include:

  • Virtual fencing and remote stock movement.
  • Animal-location and behaviour monitoring.
  • Heat and health detection.
  • Pasture cover estimation.
  • Feed planning and grazing allocation.
  • Reproductive performance analysis.
  • Digital farm models and decision assistants.

Halter’s platform processes large volumes of animal and pasture data through collars, cloud services and machine-learning models. Its satellite-enabled beef system reduces the need for towers or cellular coverage on remote properties. (halterhq.com)

Aimer is pursuing a different but complementary model: smartphone-based pasture measurement combined with satellite data, paddock modelling and predictive analytics. Its AI tools are designed around New Zealand’s pasture-based production system rather than imported housed-livestock or broadacre assumptions. (aimer-farming.com)

The evidence suggests that dairy farmers are most receptive to AI when it:

  • Uses familiar farm data.
  • Produces a clear operational recommendation.
  • Works offline or with limited connectivity.
  • Supports rather than replaces farmer judgement.
  • Fits existing advisory relationships.
  • Demonstrates a visible economic return.

Generative AI and farm administration

Generative AI adoption among dairy farmers remains earlier-stage than embedded operational AI.

A DairyNZ-commissioned study interviewed farmers, rural professionals and AI specialists. It found that current GenAI use is concentrated among innovators and early adopters, with ChatGPT the most commonly used tool. Reported uses include:

  • Summarising technical information.
  • Analysing farm spreadsheets and test results.
  • Interpreting animal-health and reproduction data.
  • Reviewing feed and nutrient information.
  • Drafting standard operating procedures.
  • Writing communications.
  • Exploring farm scenarios and budgets.

The report found that decision support and contextual analysis were more common than fully autonomous farm management. Farmers generally accepted that outputs could be inaccurate and used their own judgement to sense-check results. (dairynz.co.nz)

This reinforces the distinction between:

  1. Embedded AI, which operates inside a farm platform or sensor system; and
  2. Self-directed GenAI, where a farmer actively prompts a general-purpose model.

Embedded AI is further along because it is trained and configured around a defined workflow. General-purpose GenAI remains useful for productivity and analysis, but trust, data handling and accuracy are still limiting factors.

Horticulture, viticulture and packhouses

Horticulture has a smaller number of visible AI deployments than dairy, but the commercial value of individual applications can be substantial.

Hectre is one of the strongest examples. The Auckland-founded company raised NZ$12 million in Series A funding in February 2026. Its AI and computer-vision tools assess fruit size, colour and quality before produce enters the packhouse. The company reported that its systems process data representing billions of pieces of fruit annually across 22 countries. (auckland.ac.nz)

Hectre’s strategy is moving toward a broader fruit-quality information system, including spectroscopy research intended to identify defects and maturity before storage losses occur. The company argues that better information can improve grading, storage, sales and grower returns. (auckland.ac.nz)

Hawk Technology’s apple-packing robots demonstrate the next stage: AI is no longer limited to observation and prediction but is increasingly connected to physical manipulation and automation. (farmersweekly.co.nz)

Other horticultural use cases include:

  • Disease-risk modelling.
  • Crop-load estimation.
  • Fruit counting and sizing.
  • Vineyard disease detection.
  • Automated grading.
  • Targeted spraying.
  • Harvest and packhouse workflow management.

Irrigation and soil intelligence

The University of Canterbury is developing the ANZ Soil Moisture Data Assimilation System, which combines ground sensors, satellite signals and AI-based modelling. The system is intended to produce field-scale soil-moisture estimates multiple times a day. (canterbury.ac.nz)

Potential applications include:

  • Improving irrigation timing.
  • Reducing water waste.
  • Supporting pasture-growth decisions.
  • Improving drought resilience.
  • Managing irrigated arable and horticultural crops.

The project illustrates a broader movement toward environmental intelligence: AI systems that integrate multiple sources of imperfect data to produce a more useful operational picture.

Biosecurity, emissions and public-good applications

AI is increasingly being used in agricultural systems where the benefits are shared across the sector.

Examples include:

  • AI-supported biosecurity standard development. (beehive.govt.nz)
  • AI-enabled cameras for pest and hornet surveillance.
  • Emissions calculators and farm-level carbon accounting.
  • AI weed detection and precision control.
  • Disease-risk prediction.
  • Public soil-moisture and climate information systems.
  • Traceability and farm-to-fork data exchange.

The AI Forum’s 2026 Blueprint identifies a farm-to-fork tracking prototype involving the University of Waikato, Nanyang Technological University and Massey University. The aim is to support cross-border information sharing while protecting commercially sensitive data, with potential applications in trade digitisation, carbon tracking and producer feedback. (aiforum.org.nz)

Research and Institutional Developments

National AI research capability

Five concepts were selected for the second phase of New Zealand’s proposed national AI Research Platform. Agriculture is prominent in at least three of them:

  • The BioAI Platform, led by the Bioeconomy Science Institute.
  • The Physical AI proposal led by the Universities of Waikato and Canterbury.
  • The University of Canterbury-led national platform concept focused on complex real-world environments.

The proposals emphasise AI that can operate outdoors, handle uncertainty and support agriculture, horticulture, forestry and environmental monitoring. (mbie.govt.nz)

However, the final platform decision is not yet visible in the public MBIE material reviewed for this update. The public timetable records that phase-two proposals were due on 31 March 2026, but still states that an announcement timeline will be provided in due course. This should be treated as an unresolved policy and infrastructure development rather than a confirmed cancellation or delay. (mbie.govt.nz)

Sector-level AI strategy

The AI Forum’s May 2026 Blueprint describes New Zealand as having relatively high AI use but comparatively weak trust, governance and depth of integration. Across the wider economy, adoption estimates vary considerably by survey, while many organisations continue to use AI informally rather than through formal strategies. (aiforum.org.nz)

For agriculture, the Blueprint highlights:

  • Flat productivity growth.
  • Limited capital for productivity-enhancing technology.
  • Climate adaptation and emissions pressure.
  • Connectivity challenges in isolated terrain.
  • The importance of data sovereignty and kaitiakitanga.
  • A need for collaboration across farmers, researchers, investors and technology companies.
  • The importance of applying AI across the full primary-sector value chain. (aiforum.org.nz)

The Blueprint’s agricultural vision for 2030 is a sector in which high-quality data and deep farm knowledge support better decisions, lower emissions, improved profitability and increased resilience. (aiforum.org.nz)

Case Studies

Halter: AI as livestock and farm infrastructure

Technology: Smart collars, virtual fencing, machine learning, animal monitoring and satellite connectivity.

Primary value: Labour reduction, pasture utilisation, livestock management and remote-farm access.

Current signal: More than 500,000 cattle are reported to be using Halter collars across New Zealand, Australia and the United States. (halterhq.com)

Strategic assessment: Halter is among New Zealand’s most advanced examples of AI moving from pilot to operating infrastructure. Its main challenge is no longer demonstrating novelty; it is proving reliability, interoperability and consistent return on investment at scale.

Aimer Farming: pasture intelligence for grazing systems

Technology: Smartphone computer vision, satellite data, paddock digital twins and predictive analytics.

Primary value: More frequent pasture measurement, improved grazing decisions and better feed planning.

Current signal: MPI is co-funding a project to test the system across hundreds of farms, with Fonterra, Ravensdown and Cropmark Seeds involved. (mpi.govt.nz)

Strategic assessment: Aimer is a strong example of New Zealand-specific AI design. Its value comes from aligning AI with pasture-based farming, offline mobile use and existing farmer routines.

Hectre: AI before the packhouse

Technology: Computer vision, fruit sizing, colour assessment, quality analysis and planned spectroscopy.

Primary value: Better storage, grading, market allocation and reduced fruit waste.

Current signal: Hectre raised NZ$12 million in 2026 and reported customers in 22 countries. (auckland.ac.nz)

Strategic assessment: Hectre demonstrates that New Zealand agricultural AI can scale internationally when it converts a costly information gap into a measurable supply-chain advantage.

Hawk Technology: physical AI in apple packing

Technology: Robotics, machine vision, AI inference and automated fruit packing.

Primary value: Packhouse productivity, labour substitution and exportable automation capability.

Current signal: Fifty Gen 5 Apple Packers had been installed in Washington State, with a further 250 on order. (farmersweekly.co.nz)

Strategic assessment: Hawk Technology shows the potential for New Zealand to export agricultural automation systems, not just food products. It also highlights the need to retrain workers for robotics, software and maintenance roles.

Scanabull: low-friction livestock measurement

Technology: 3D LiDAR and AI-based cattle weight estimation through a phone scan.

Primary value: Faster and potentially less stressful weighing and more frequent livestock measurement.

Current signal: Scanabull won the Prototype Award at the 2026 Fieldays Innovation Awards. (fieldays.co.nz)

Strategic assessment: Scanabull represents the widening innovation funnel: small, focused tools may achieve adoption more quickly than large all-in-one platforms if they solve a clear operational problem at low friction.

1. Workflow-embedded AI is outperforming generic AI

The strongest commercial examples are attached to an existing action:

  • Move cattle.
  • Measure pasture.
  • Grade fruit.
  • Weigh livestock.
  • Detect disease.
  • Manage irrigation.
  • Prepare biosecurity standards.

This reduces the behavioural change required from farmers and makes value easier to measure.

2. AI is moving from prediction toward intervention

New Zealand agriculture is progressing from analytics to systems that act in the physical environment:

  • Halter collars influence livestock movement.
  • Hawk robots manipulate and pack fruit.
  • AI systems support targeted weed control.
  • Drones and autonomous systems are being developed for orchard and vineyard operations.

This transition increases both the economic opportunity and the need for safety, fallback procedures, certification and liability frameworks.

3. Connectivity is a strategic differentiator

Rural connectivity remains a structural constraint. Halter’s direct-to-satellite system and Aimer’s offline smartphone capability both address the reality that AI tools must function in remote or low-coverage environments. (halterhq.com)

4. Adoption programmes are becoming as important as invention

Responsible Dairy, AgriZeroNZ’s Early Adoption Accelerator and MPI’s Aimer co-investment all point to a more mature innovation model. The central question is increasingly not whether a technology can work, but whether it can be tested, financed, supported and trusted by ordinary operators.

5. Data integration is the next major frontier

New Zealand’s agricultural AI opportunity extends beyond the farmgate. The most valuable future systems are likely to combine:

  • Farm-management records.
  • Animal and pasture data.
  • Weather and satellite information.
  • Processing and packhouse data.
  • Emissions and traceability information.
  • Market and logistics data.

The AI Forum and Tech New Zealand both identify fragmented data and weak value-chain integration as major constraints. (technewzealand.org.nz)

Constraints and Risks

  • Reliability: Hardware failures, outages and poor connectivity can undermine confidence in systems that influence livestock or machinery. (farmersweekly.co.nz)
  • Interoperability: Farmers often manage multiple platforms and want integrated information rather than additional standalone applications.
  • Independent validation: Many performance claims remain company-reported. Longitudinal, independent studies are still needed.
  • Cost and scale: Smaller farms may struggle to justify subscription, hardware, connectivity and integration costs.
  • Data ownership: Farm data may have commercial, cultural and strategic value. Data sovereignty and Māori governance principles need to be incorporated early.
  • Human oversight: Generative AI remains vulnerable to inaccurate or poorly contextualised outputs. DairyNZ research supports a decision-support model rather than fully autonomous decision-making. (dairynz.co.nz)
  • Workforce transition: Automation may reduce some repetitive tasks while increasing demand for robotics, data, software and maintenance skills. (farmersweekly.co.nz)
  • Regulatory uncertainty: Physical AI and autonomous agricultural machinery will require clearer standards covering safety, accountability and liability.
  • SME adoption gap: Wider New Zealand research continues to show that many small and medium-sized businesses are not yet planning significant AI investment. (mbie.govt.nz)

Outlook: 2026–2028

Over the next two years, AI adoption in New Zealand agriculture is most likely to advance through five pathways:

  1. More embedded AI in existing farm platforms
    Farmers will increasingly encounter AI through tools they already use for pasture, animal health, reproduction, accounting and farm planning.

  2. Expansion of physical AI in horticulture
    Packhouses, orchards and vineyards are likely to see greater use of machine vision, robotic handling, targeted spraying and autonomous scouting.

  3. Greater emphasis on emissions and resilience
    AI will support emissions measurement, feed optimisation, pasture allocation, drought response and environmental compliance.

  4. More public-private commercial trials
    Co-investment programmes will become central to moving technologies from research to practical farm use.

  5. Early development of agentic farm assistants
    Farm-specific AI assistants may begin coordinating data from multiple systems, but mainstream adoption will depend on clear boundaries, auditability, farmer control and trusted data sources.

Conclusion

As of 18 August 2026, AI adoption in New Zealand agriculture is best described as practical, selective and increasingly embedded in sector infrastructure.

Dairy remains the leading market, with Halter and Aimer demonstrating two complementary models: AI-enabled livestock operating systems and pasture-intelligence platforms. Horticulture is advancing through fruit-quality systems, disease prediction and increasingly capable robotics. Public-good applications in biosecurity, irrigation, emissions and environmental monitoring are expanding the role of AI beyond private farm software.

The most important development since the previous update is not a single breakthrough model. It is the strengthening of the pathway from innovation to adoption:

  • AgriZeroNZ is funding early commercial deployment.
  • Responsible Dairy is creating large-scale farm testbeds.
  • MPI is supporting Aimer across hundreds of farms.
  • Hawk Technology is exporting AI-enabled agricultural robotics.
  • Biosecurity New Zealand is testing generative AI in regulatory work.
  • Sector bodies are developing stronger frameworks around trust, data and adoption.

The central strategic lesson is clear: AI will gain ground in New Zealand agriculture where it is workflow-specific, low-friction, connected to trusted data, resilient in rural conditions and able to demonstrate measurable economic or environmental value. The next phase will be determined less by novelty than by reliability, integration, farmer confidence and the ability to scale proven tools across the food and fibre value chain.

← Back to all articles ⚑ Flag an issue