# AI for New Zealand farming. URL: https://levystreet.com/industries/agriculture/ Updated: 2026-10-19 AI for agriculture in New Zealand: practical farm uses from record retrieval to NAIT and compliance evidence, the rules that apply and where AI gets it wrong. ## Short answer AI in New Zealand agriculture is most useful for farm records rather than farming decisions: finding spray, treatment and maintenance records, assembling evidence for audits and freshwater farm plans, and turning photos and voice notes into tidy logs. Farmers, vets and agronomists keep the decisions. The value comes from fewer hours hunting for paperwork and fewer gaps when an auditor or processor asks. - Start with record retrieval: questions like when a paddock was last sprayed, answered with the source. - NAIT movements must be recorded within 48 hours, so reminders and drafts help; registration stays manual. - Never let a model calculate withholding periods or application rates; use the product label. - Agree who owns farm data before connecting tools, especially on Māori-owned land. ## Where AI is useful on a New Zealand farm AI is useful on a farm where records are scattered and questions are repetitive: which paddock was sprayed with what, when the tractor was last serviced, which cows were treated and when they can go back in the vat. Those answers usually exist, in a diary in the ute, a photo on someone's phone, a spreadsheet and a farm software account. AI can read all of them and give the answer with its source. Farm businesses are mostly small businesses, and small businesses are aware of AI but light users. Research for MBIE by TRA, an online panel of about 500 businesses with up to 19 staff surveyed in April 2025, found 94% were aware of at least one AI tool, yet about a third had not used any AI tool in their business ([MBIE, July 2025](https://www.mbie.govt.nz/business-and-employment/business/support-for-business/research-and-reports-for-small-business/ai-adoption-by-new-zealand-small-and-medium-sized-businesses)). That gap is the opportunity: the work is administrative, not agronomic. ## Farm workflows that suit AI | Workflow | What AI does | What a person still decides | Data involved | Difficulty | |---|---|---|---|---| | [Farm record retrieval](/industries/agriculture/farm-record-retrieval/) | Answers questions from diaries, software exports and photos, citing the record | Whether the record is complete and what to do next | Spray and fertiliser diaries, farm software exports | Medium: scattered formats | | [Compliance evidence collection](/industries/agriculture/compliance-evidence-collection/) | Gathers and files evidence against an audit or plan checklist, flags gaps | What to submit and any corrective action | Plans, receipts, photos, maps | Medium: changing requirements | | [Equipment maintenance logs](/industries/agriculture/equipment-maintenance-logs/) | Turns voice notes and service invoices into a dated log per machine | Repairs, replacement, safety lock-out | Invoices, hour meters, voice notes | Low: simple structure | | Spray diary checks | Extracts product, rate, paddock and weather from photographed diary pages | Whether an application was appropriate | Handwritten diaries, product labels | Medium: handwriting and units | | Animal treatment records | Finds treatment history and the label withholding period on file | Treatment, milk and meat release | Treatment books, vet scripts | High: food safety stakes | | NAIT movement reminders | Matches stock sale and transfer records against recorded movements | Recording the movement in NAIT | Sale dockets, NAIT exports | Low: clear rules | | Seasonal staff induction Q&A | Answers new staff questions from the farm's own procedures | Sign-off that the person is competent | Induction notes, hazard register | Medium: must cite procedures | | Invoice coding | Suggests codes for supplier invoices in Xero or Figured | Final coding and GST treatment | Invoices, chart of accounts | Low: existing rules | The detailed guides cover [finding farm records](/industries/agriculture/farm-record-retrieval/), [collecting compliance evidence](/industries/agriculture/compliance-evidence-collection/) and [keeping equipment maintenance logs](/industries/agriculture/equipment-maintenance-logs/). ## The records rules already require Much of the value comes from rules farms already follow. AI helps keep the evidence; it does not change the obligations. - **NAIT.** OSPRI says animal movements [must be recorded in NAIT within 48 hours](https://www.ospri.co.nz/farmers-and-livestock-owners/moving-animals/recording-animal-movements), with both the sending and receiving person in charge of animals responsible for their side. Animals must be [registered within 7 days of tagging](https://www.ospri.co.nz/farmers-and-livestock-owners/tagging-and-registering-animals/registering-animals-in-nait), and tagged and registered before 180 days of age or their first off-farm movement. A reminder that compares sale dockets with NAIT records is a low-risk, high-value use. - **Freshwater farm plans.** These sit in Part 9A of the [Resource Management Act 1991](https://www.legislation.govt.nz/act/public/1991/69/en/latest/), which was amended on 21 August 2025 by the Resource Management (Consenting and Other System Changes) Amendment Act 2025. Requirements and dates are set by regulations and rolled out by regional councils, and they have changed more than once since 2023. Check with your regional council before relying on any date an AI tool gives you. - **Health and safety.** Under the [Health and Safety at Work Act 2015](https://www.legislation.govt.nz/act/public/2015/70/en/latest/), a business must engage with workers when proposing changes that may affect their health or safety (s 60). Moving hazard reporting or inductions onto an AI assistant is that kind of change. - **Data ownership.** The [NZ Farm Data Code of Practice](https://www.farmdatacode.org.nz/) asks accredited organisations to disclose their practices on data rights, processing and sharing. For ahuwhenua trusts and Māori incorporations, [Te Mana Raraunga's principles](https://www.temanararaunga.maori.nz/s/TMR-Maori-Data-Sovereignty-Principles-Oct-2018.pdf) say Māori data should be stored in Aotearoa whenever possible, which affects which tools you choose. ## Getting answers out of Figured, FarmIQ and MINDA Farm data in New Zealand is spread across specialist systems: LIC's MINDA for herd records, Fonterra Farm Source for supply data, FarmIQ for livestock and land, Figured and [Xero](/integrations/xero/) for financials and budgets. Each holds part of the answer. Most offer exports, and some offer APIs. A practical AI setup pulls regular exports into one searchable store and answers questions across them, with a link back to the original record. That is a [retrieval system](/services/rag-development/), not a chatbot trained on your farm; our lesson on [building a searchable knowledge base](/learn/building-a-searchable-business-knowledge-base/) shows the idea at small scale. Paper is the other half. Photographed diaries, docket books and service invoices need [document processing](/services/ai-document-processing/) that copes with handwriting, mud and creased pages, and flags anything it could not read rather than guessing. ## A first project: one place to ask about farm records - **Inputs**: Twelve months of spray and fertiliser diaries (photographed), farm software exports, machinery invoices. - **Systems**: A shared drive or Microsoft 365 folder, the farm's existing software exports, a retrieval tool with citations. - **Human review**: The farm manager checks extracted diary lines weekly for the first two months. - **Outputs**: Answers to plain questions ("what went on paddock 14 this season") with a link to the source page or record. - **Measures**: Time to answer the ten questions you are asked most, answers matching the source, unreadable pages flagged. Illustrative example: a dairy operation with a manager and three staff fields about 15 record requests a month from its accountant, auditor, processor and contractors (an assumption). At 20 minutes each to find and check the record, that is 5 hours a month. If retrieval brings each to 5 minutes including checking the cited source, the saving is about 3.75 hours a month, plus fewer requests answered from memory. ## Where AI gets farming wrong - **Advice it has no business giving.** Ask a general chatbot about drench or spray rates and it will answer. Control: the assistant only answers from your records and labels, and refuses agronomy and animal health questions. - **Units and rates.** Diaries mix L/ha, mL per 100 L and total litres. Control: extraction keeps the unit as written and flags conversions for a person. - **Handwriting and photos.** Smudged pages produce plausible but wrong dates and paddock numbers. Control: confidence flags and weekly human checks of extracted lines. - **Out-of-date rules.** Freshwater and other requirements change between seasons. Control: answers about obligations link to the regulator's current page, never the model's memory. - **Connectivity.** Rural coverage is patchy. Control: capture offline, process later, and keep safety steps on paper or local devices. Related work in our portfolio: [Eco-index](/portfolio/eco-index/) was a now-closed New Zealand biodiversity programme (2021 to 2026) whose open data and AI-trained vegetation detection models guided native ecosystem reconstruction. ## Questions people ask ### Can AI work on a farm with poor mobile coverage? Partly. Capture can happen offline: photos, voice notes and forms saved on a phone and uploaded when the ute reaches Wi-Fi or coverage. The AI processing then runs in the cloud and the results are waiting next time someone checks. Design for delay rather than live answers, and never make a safety or animal welfare step depend on a connection. ### Can AI tell me what to spray or how to treat sick stock? It should not. General chatbots answer agronomy and animal health questions confidently, but they do not know your soils, your stock history or the product registered for the job, and they can invent rates and withholding periods. Use AI to find what your vet, agronomist or the product label already said, and keep the decision with those people. ### What farm records are worth digitising first? The ones you are most often asked for: spray diaries, animal treatment records, fertiliser applications, staff inductions and machinery servicing. These are also the records that auditors, processors and regional councils request. Photographing paper diaries and letting AI extract dates, paddocks, products and quantities into a spreadsheet is a sensible start, with a person checking every line. ### Who owns the data when AI reads my farm records? Check the terms of every tool before you connect it. The NZ Farm Data Code of Practice asks accredited organisations to disclose their practices on data rights, processing and sharing, which is a useful test for any provider. For Māori-owned land and trusts, Te Mana Raraunga's principles on Māori data sovereignty are a good guide to consent and storage questions. ### How much admin time can AI save a farm business? It depends on how scattered the records are now. A farm with one manager who knows where everything is gains little. A larger operation with several staff, paper diaries and a manager fielding questions from accountants, auditors and processors can save a few hours a week. Time a month of record requests first so you can measure the change honestly. ## Sources - [AI adoption by New Zealand small and medium-sized businesses](https://www.mbie.govt.nz/business-and-employment/business/support-for-business/research-and-reports-for-small-business/ai-adoption-by-new-zealand-small-and-medium-sized-businesses) - [Recording animal movements](https://www.ospri.co.nz/farmers-and-livestock-owners/moving-animals/recording-animal-movements) - [Registering animals in NAIT](https://www.ospri.co.nz/farmers-and-livestock-owners/tagging-and-registering-animals/registering-animals-in-nait) - [Resource Management Act 1991 (version as at 12 September 2026)](https://www.legislation.govt.nz/act/public/1991/69/en/latest/) - [Health and Safety at Work Act 2015](https://www.legislation.govt.nz/act/public/2015/70/en/latest/) - [NZ Farm Data Code of Practice](https://www.farmdatacode.org.nz/) - [Principles of Māori Data Sovereignty](https://www.temanararaunga.maori.nz/s/TMR-Maori-Data-Sovereignty-Principles-Oct-2018.pdf) Published by Levy Street (https://levystreet.com), Wellington, New Zealand.