AI in Agriculture in Aotearoa New Zealand: A Living Whitepaper

Updated 10 June 2026

Introduction

AI in New Zealand agriculture is now best understood as a set of operational systems tied to specific production problems rather than a single broad adoption wave. Dairy remains the most commercially advanced segment, while horticulture is concentrating investment in disease-risk modelling, orchard robotics, and AI-enabled quality control. At the same time, public-good AI is becoming more important in emissions measurement, weed-risk analysis, and biosecurity-oriented decision support. This matters in a sector with outsized economic leverage: MPI’s latest published outlook forecasts food and fibre export revenue of NZ$62 billion in the year to 30 June 2026, including NZ$27.4 billion from dairy and NZ$9.2 billion from horticulture. New Zealand’s July 2025 AI strategy also explicitly takes an adoption-led stance, emphasising AI application over frontier-model development and identifying agriculture as a natural fit for local advantage. (mpi.govt.nz)

Executive Snapshot

• Dairy is still the centre of gravity. The strongest AI deployments remain in cattle movement and virtual fencing, in-shed animal-health alerts, pasture intelligence, and advisory/search tools grounded in trusted sector content. (halterhq.com)
• The biggest change since 2 April 2026 is stronger operating-layer deployment. Halter has moved beyond on-farm tower dependence with direct-to-satellite connectivity, and Aimer has shifted from scale-up messaging to backed co-investment and product development. (halterhq.com)
• Horticulture AI is attracting capital and sharpening around export risk and quality intelligence. Hectre’s February 2026 raise and Lincoln Agritech’s STELLA work in Hawke’s Bay both point to AI being applied where fruit quality, storage risk, and labour constraints are most economically material. (auckland.ac.nz)
• Research and public-interest AI are becoming sector infrastructure. AgResearch says it has more than 50 AI-focused projects underway, while tools such as Map and Zap® and the On-Farm Emissions Calculator show AI moving into national capability, not just startup software. (agresearch.co.nz)
• The main bottleneck is still integration, trust, and proof of value. DairyNZ’s research and conference outputs continue to emphasise that farmers need clear ROI, interoperable data, and trustworthy outputs before AI moves from early adopters into the mainstream. (dairynz.co.nz)

What Changed Since 2 April 2026

• Halter launched direct-to-satellite virtual fencing in May 2026. The company says its collars can now connect through One NZ Satellite and Starlink without on-farm infrastructure, and that it serves more than 2,000 farmers and ranchers across New Zealand, Australia, and the US, with one million collars sold. (halterhq.com)
• Aimer secured new government co-investment. MPI’s Primary Sector Growth Fund is investing NZ$600,000 in a NZ$1.675 million three-year project to develop AI-enabled tools intended to optimise emissions, productivity, and profit in pasture-based dairy and beef systems. (mpi.govt.nz)
• Horticulture AI drew fresh capital. Hectre closed an oversubscribed NZ$12 million Series A in February 2026 to expand its AI and computer-vision systems for fruit sizing, colour, and quality before fruit reaches the packhouse. (auckland.ac.nz)
• Fieldays 2026 signalled broader market visibility for AI. Organisers said they had seen a “marked increase” in AI-based solutions among innovation-award participants, and AWS joined as overall sponsor. (fieldays.co.nz)
• The national AI research-platform decision still appears unresolved publicly. MBIE’s AI Research Platform page, as visible on 10 June 2026, still shows the phase-two process and shortlisted concepts, including the agriculture-focused BioAI Platform, but no published final award announcement. This is an inference from the current public material, not a confirmed delay notice. (mbie.govt.nz)

Current News and Market Developments

1) Livestock AI has deepened from automation into connectivity

Halter remains the clearest scaled example of AI in New Zealand farming, but the May 2026 development matters because it changes where the system can work, not just what it can do. Direct-to-satellite connectivity is aimed squarely at remote and large properties where communications infrastructure was a practical blocker. Halter says internal modelling suggests this could expand access for New Zealand beef farms by at least 20%, while the same announcement bundled new tools for reproduction, animal behaviour, and precision pasture management. Strategically, that pushes livestock AI further toward being a farm operating layer rather than a standalone control tool. (halterhq.com)

Focused animal-health AI also continues to show the kind of economics that mainstream farmers notice. Bovonic says participant results for QuadSense include payback within six months for some users, average labour savings of 3.7 hours per week, a 37% drop in bulk SCC, and reduced antibiotic use on participating farms. Because these are company-reported participant outcomes rather than independent sector-wide figures, they are best read as evidence that narrow, workflow-native AI is converting more easily than general-purpose AI. (bovonic.com)

2) Pasture intelligence is becoming a more serious productivity category

Aimer is important because it is building AI around New Zealand’s pasture-based farming model, not around imported row-crop logic. Its platform uses smartphone-based computer vision and proprietary algorithms to estimate pasture, forecast growth, generate feed wedges, and support grazing plans. In March 2026, the company said its 2026 priorities included expanding its customer base and rolling out satellite- and drone-based assessment; by May, MPI had committed public funding to a three-year project with emissions, productivity, and profit optimisation at its centre. (aimer-farming.com)

The commercial logic also lines up with sector economics. DairyNZ’s May 2026 productivity commentary says New Zealand dairy’s competitive edge still rests on homegrown feed and pasture, and that AI-driven pasture tools could improve efficiency, though many remain unproven at scale and expensive. That is an important market signal: pasture AI is no longer just an agritech niche, but it has not yet fully crossed into cost-assured mainstream adoption. (dairynz.co.nz)

3) Dairy institutions are becoming active AI deployers

DairyNZ is no longer just analysing AI adoption; it is using AI in farmer-facing delivery. Its DAiSY assistant now draws from more than 1,100 pages and 880 tools and resources from DairyNZ’s knowledge base while restricting responses to DairyNZ website content. That matters because institution-led AI can lower trust barriers by grounding outputs in familiar, cited, sector-owned material. (dairynz.co.nz)

DairyNZ’s December 2025 International Precision Dairy Farming Conference also underscored how central AI has become to sector thinking. The event brought together more than 400 people from 22 countries, including 90 New Zealand dairy farmers, and highlighted AI, automation, climate, and modelling, while explicitly stressing that adoption depends on trust, integration, and clear value. That is still the best short summary of the dairy market in mid-2026: high interest, real deployment, but uneven diffusion. (dairynz.co.nz)

Horticulture, Arable and Orchard Systems

1) Horticulture AI is clustering around quality, disease and export assurance

Hectre is a strong signal that horticulture AI remains a live capital category. The Auckland-founded company raised NZ$12 million in February 2026 to expand AI and computer-vision tools that capture fruit size, colour, and quality before fruit enters the packhouse. The technology is aimed at one of horticulture’s most expensive data problems: poor visibility into quality before storage and grading decisions are locked in. (auckland.ac.nz)

Lincoln Agritech’s STELLA work shows the other major horticulture path: AI for invisible risk detection. In Hawke’s Bay, the project is combining automated spore sampling, UAV imagery, satellite imagery, environmental monitoring, and AI-powered risk models to predict bull’s-eye rot in apples before symptoms appear in storage or export markets. With New Zealand apple exports reaching a record NZ$1 billion in 2025, this is a direct example of AI being tied to export protection, not just operational convenience. (lincolnagritech.co.nz)

2) Arable and seed applications are becoming more concrete

A useful 2026 signal outside dairy and orchard systems comes from Plant & Food Research’s Hyperseeds work with the Foundation for Arable Research. The programme uses hyperspectral imaging plus AI to pre-screen seeds for contaminants, targeting a long-standing bottleneck in seed-quality assessment. This is narrower than livestock AI, but strategically important: it shows AI adoption widening into high-value agricultural quality assurance where labour-intensive inspection has clear limits. (plantandfood.com)

3) Robotics remains part of the horticulture AI stack

The broader autonomy story still matters. Robotics Plus’ 2025 acquisition by Yamaha Motor created the base for Yamaha Agriculture, with Prospr positioned as an autonomous orchard and vineyard platform for spraying, weed control and other specialty-crop tasks, while keeping headquarters and IP development in Tauranga. Even without a fresh 2026 corporate event of the same scale, this remains one of the clearest signs that New Zealand horticulture AI can move from local development to globally backed deployment. (roboticsplus.co.nz)

Research Overview

DairyNZ / Perrin Ag: GenAI is useful, but mainstream adoption is still low

DairyNZ’s December 2025 Perrin Ag report remains the clearest recent research synthesis on farmer-facing GenAI in dairy. It says adoption is currently low, concentrated in innovators and early adopters, and most real-world uses fall into three buckets: decision support, task enhancement, and communication support. Decision support is the dominant use case. The same report says that in the next 3–5 years, GenAI is more likely to support farmer judgement than replace it, because of persistent limits around tacit knowledge, hallucinations, bias, trust, and infrastructure. (dairynz.co.nz)

Precision-dairy research: fragmentation is the practical problem

One of the strongest pieces of sector evidence is still the Southland dairy chatbot case study published in the 2025 precision-dairy conference proceedings. Farmers in that work were routinely monitoring 4 to 7 applications across weather, soil moisture, pasture and livestock data, and the strongest interest in AI centred on bridging silos: natural-language access to farm records, anomaly detection, automated compliance reporting, and better synthesis across fragmented systems. The implication is straightforward: the next wave of value will come less from adding more sensors than from making existing farm data usable across tools. (dairynz.co.nz)

Bioeconomy science: AI is becoming research infrastructure

AgResearch has said it has more than 50 AI-focused projects underway, including work using CT scans and AI to extract livestock traits linked to methane, feed efficiency, welfare and meat quality. Since 1 July 2025, AgResearch has operated as a group within the Bioeconomy Science Institute, reinforcing the trend toward larger-scale, bioeconomy-oriented AI capability in agriculture. (agresearch.co.nz)

That research layer is not purely analytical. AgResearch’s Map and Zap® uses AI for weed identification and laser-targeted control, and the Institute’s weed-risk work uses AI to screen large volumes of literature and model risks across New Zealand’s exotic flora. Together, these programmes show AI shifting from exploratory science into tools with operational, regulatory, and commercial relevance. (agresearch.co.nz)

Emissions and climate tools are moving toward standardisation

The On-Farm Emissions Calculator, launched in October 2025 by MPI, the Bioeconomy Science Institute and the Ag Emissions Centre, is the first tool to apply the government-mandated farm-emissions standard released in December 2024. That matters because standardised emissions estimation is likely to become a base layer for later AI-enabled advisory and optimisation tools, especially in grazing systems where climate and profitability decisions are tightly coupled. (agresearch.co.nz)

Case Studies

Halter: AI as a farm operating system

Halter now combines virtual fencing, livestock guidance, animal monitoring, reproduction tools and pasture-management features with direct-to-satellite connectivity. The May 2026 launch is significant because it expands applicability to remote and rugged operations, especially beef, rather than merely refining an already proven dairy use case. (halterhq.com)

Aimer Farming: pasture AI built for New Zealand systems

Aimer shows what low-friction AI looks like in a pasture economy: smartphone-first measurement, forecast pasture intelligence, and planned satellite/drone extensions, now backed by direct MPI co-investment. Its importance lies less in headline scale than in how closely it maps to New Zealand’s cost base and pasture-led competitive model. (mpi.govt.nz)

Bovonic QuadSense: focused AI with visible farm economics

QuadSense is a strong example of narrow AI diffusing through economics rather than hype. Its value proposition is not “AI” in the abstract; it is earlier mastitis detection, lower SCC, time savings, and reduced antibiotic use. That is the pattern much of agriculture is following. (bovonic.com)

Lincoln Agritech / STELLA: AI for export-risk management

STELLA is a high-value case of AI improving early detection of disease that conventional sensing struggles to see. In New Zealand’s export horticulture sectors, that kind of prediction capability is strategically important because it protects market access, reputation and storage outcomes. (lincolnagritech.co.nz)

Hectre: AI before the packhouse

Hectre’s fruit-quality systems push AI upstream in the horticulture value chain, helping growers and packers make earlier decisions on storage, sales, and grading. Its February 2026 raise is a useful commercial proof point that investors still see growth in practical horticulture AI. (auckland.ac.nz)

Core Trends

1) Workflow-embedded AI is outperforming general-purpose AI

The strongest systems in market are tied to jobs farmers and growers already do every day: shifting cattle, detecting mastitis, measuring pasture, predicting disease, screening seed, or estimating emissions. Tools succeed when they fit existing operating rhythms and produce actions, not just insights. (halterhq.com)

2) Interoperability is becoming the sector’s main bottleneck

Research with dairy farmers points to fragmentation across multiple apps; AgriTechNZ continues to treat trusted data exchange and common definitions as foundational sector work. The limiting factor is increasingly not the availability of data, but the ability to move it cleanly across systems and turn it into decisions. (dairynz.co.nz)

3) Public-good AI is moving into the operating layer

AI in agriculture is no longer only a private-software story. Weed-risk modelling, emissions standards, and AI research-platform proposals tied to agriculture, aquaculture and forestry all suggest that AI is becoming part of New Zealand’s agricultural infrastructure. (agresearch.co.nz)

4) Horticulture adoption is narrower than dairy, but often more strategic

Where dairy uses AI frequently for day-to-day operating decisions, horticulture tends to deploy it where a single error can be extremely expensive: disease that shows up after export, poor storage choices, grading losses, or labour-intensive orchard operations. That makes adoption narrower, but often easier to justify economically. (lincolnagritech.co.nz)

5) The market is showing stronger capital confidence than mainstream farm-wide diffusion

Recent funding and visibility around companies such as Halter and Hectre show that investors are confident in New Zealand agritech. But farmer-level mainstreaming is still uneven, especially for GenAI-style tools, where DairyNZ research continues to describe adoption as low outside early adopters. (finance.yahoo.com)

Constraints and Risks

• Trust and validation remain decisive. DairyNZ’s recent work repeatedly stresses that AI adoption depends on trust, integration and clear value, while its Perrin Ag report warns that hallucinations and bias limit near-term autonomy. (dairynz.co.nz)
• Interoperability and data quality are unresolved. Farmers still work across fragmented app environments, and sector bodies are still investing in data definitions, standards and exchange frameworks. (dairynz.co.nz)
• Many tools still need better proof of value. DairyNZ’s May 2026 productivity analysis says AI-driven pasture tools may help but many remain unproven at scale and expensive. (dairynz.co.nz)
• SME adoption remains a national weakness. MBIE’s AI state-of-play material says 68% of SMEs had no plans to evaluate or invest in AI, which matters in a country where much of the agricultural supply chain is SME-based. (mbie.govt.nz)
• National research coordination is still in formation. Agriculture-relevant concepts are prominent in MBIE’s AI Research Platform process, but the final public outcome was not yet visible on MBIE’s platform page as of 10 June 2026. (mbie.govt.nz)

Conclusion

As of 10 June 2026, AI adoption in New Zealand agriculture is best described as practical, selective and increasingly infrastructural. Dairy remains the most mature segment, led by virtual fencing, animal-health detection, pasture intelligence and trusted advisory AI. Horticulture is progressing through a different set of use cases: pre-packhouse computer vision, disease-risk prediction, and autonomous orchard operations. Public-sector and institute-led AI is also becoming more consequential, especially in weeds, emissions, and biosecurity-related decision support. (halterhq.com)

The most important shift since the previous update on 2 April 2026 is not a single breakthrough model. It is the way AI is being reinforced at multiple layers at once: better field connectivity, public co-investment in pasture AI, stronger horticulture capital formation, visibly rising AI presence at Fieldays, and a national research architecture that still appears to be consolidating around agriculture-relevant options. (halterhq.com)

The near-term winners are still the same kinds of systems identified in earlier editions, but the evidence is stronger now: tools that combine workflow fit, trusted data grounding, interoperability, and visible on-farm economics are the most likely to move from early-adopter enthusiasm into New Zealand agricultural mainstream use. (dairynz.co.nz)