Yes, some farmers use AI, but fewer than headlines suggest. A 2026 survey of 5,500 farmers in ten countries found 17% use generative AI for farm tasks and 4% pay for a solution. On a small farm the real uses are mostly paperwork and planning: voice records, crop monitoring from satellite images, grant drafts and schedules.
That survey is the McKinsey Global Farmer Insights 2026 report, published on 8 September 2026 and summarised widely in the farm and business press. According to those summaries, McKinsey notes the paying figure may include general AI subscriptions, not only farm-specific tools. A second, older kind of data covers precision-agriculture technology, such as guided steering and yield mapping, which is not the same thing as a chat assistant but is the nearest measure of how readily farms take up technology: a national government farm-technology chart (published 10 December 2024, covering 2023) found guidance autosteering on 52% of midsize farms and 70% of large-scale crop farms, with small family farms showing the lowest use in every technology category.
Put those together and the honest reading is this: adoption rises with farm size, general AI use is still a minority habit, and small farms are the least served by products priced and built for large ones. That gap is also the opening, because the cheap uses below need no machinery. Each of the eight uses has a what, an example, an indicative cost, and a check, so you can tell which are worth an afternoon.
The farm we will use for the numbers
All figures are illustrative; replace them with your own. Imagine a 25-hectare mixed farm: 15 hectares of vegetables and cereals, 5 hectares of grassland with 40 sheep, and 5 hectares of high-value salad crops grown under contract with two local buyers. The owner works with one full-time employee and takes on seasonal help for about eight weeks. Time is worth a loaded $22 an hour, and the owner spends roughly six hours a week on records, planning, messages and forms.
The table gives the shape of the eight uses before the detail. "Indicative" means a typical list price for a small farm in November 2026, so check current prices before you buy.
| # | Use | Indicative cost a month | Hours saved a week (illustrative) | Hardware needed |
|---|---|---|---|---|
| 1 | Voice records into a sheet | $0 to $20 | 1.0 | Phone |
| 2 | Satellite crop monitoring | $4.99 to $29.99 | 0.5 plus better timing | None |
| 3 | Grant and scheme applications | $0 to $20 | Lumpy: 6 to 12 hours per application | None |
| 4 | Crop and grazing planning | $0 to $20 | 0.5 | None |
| 5 | Weather and irrigation log | $0 to $20 | 0.5 | None |
| 6 | Buyer and customer messages | $0 to $20 | 0.75 | None |
| 7 | Rotas and helper instructions | $0 to $20 | 0.5 | None |
| 8 | Reading lab reports and long documents | $0 to $20 | 0.25 | None |
Most of the tools share one subscription. A paid individual assistant plan, such as ChatGPT Plus at $20 a month, covers uses 1, 3, 4, 5, 6, 7 and 8 together. The only separate cost is the satellite service in use 2. A team plan such as ChatGPT Business is $20 per seat billed annually or $25 monthly, with a two-seat minimum, which only suits a farm with at least two people using it.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowUse 1: voice records that become a sheet
The use most likely to repay quickly is turning speech into records. Farms generate records constantly: field operations, treatments, stock movements, feed, repairs, hours. Most are written late from memory, if at all. Speaking into a phone while the job is fresh is easier, and an assistant tidies the text into rows.
A voice note, spoken at the field gate: "Field three, the long field, drilled spring barley today, 14 hectares, seed rate 160 kilos a hectare, tractor two, started eight, finished two, ground fit, light showers at noon."
Turn each voice note into one row: date, field, operation, area_ha,
product_or_seed, rate_per_ha, machine, start, finish, conditions, notes.
Leave a cell empty if not stated. Do not infer anything.
After the table, list anything that looks inconsistent (for example, an area
larger than the field I have named).An illustrative result puts "barley (spring)" in the product column, 160 kg in the rate column, and flags nothing unless you have told it the field is 12 hectares, in which case it should flag the 14. Your check: read each row against what you said. Cost: free tier or $20 a month; time saved about an hour a week once the habit sticks. The record is still yours to keep correct, and any record that must meet a legal standard should be kept in the form the rules require.
Use 2: satellite monitoring, priced per hectare
Satellite services compare regular images of your fields and flag areas where plant vigour looks low or changes. You do not buy hardware. One example, Farmonaut, lists monthly plans on its pricing page: Mini at $4.99 a month for 5 hectares and Basic at $29.99 a month for 30 hectares (monthly billing, read November 2026; annual billing is cheaper). Other services exist and prices change, so compare at least two.
For the 25-hectare farm, the choice is concrete. Watching the whole farm on Basic costs $29.99 a month, or about $360 a year. Watching only the 5 hectares of contract salad crops on Mini costs $4.99, or about $60 a year. Since the salad crops carry the highest value per hectare and the contract penalty for a poor harvest, the owner starts with Mini, learns what an alert looks like, and widens only if it proves useful.
What to expect and what to doubt. A vigour map shows where plants look different from the rest of the field. It does not say why. A pale patch might be waterlogging, a nutrient shortage, a blocked drain or a tractor pass. You still walk the field. The service saves you walking all 25 hectares weekly and helps you choose where to go first. Cloud cover and field size limit what you see. The tutorial on crop monitoring without drones covers the limits in depth, and the pricing comparison puts two services side by side.
Use 3: grant and scheme applications
Applications are where an assistant earns a year of subscription in an afternoon, and also where mistakes cost most. The assistant is good at structure: turning your rough notes into the headings a scheme asks for, checking that every criterion has an answer, and tightening wording to a word limit. It is poor at facts about your farm and about the scheme.
The safe way to use it is to supply both the criteria and your facts, and forbid anything else:
Below are the funder's assessment criteria (pasted from their guidance)
and my farm notes. Draft the "project description" section in 300 words.
Use only facts in my notes. Where a criterion has no matching fact in my
notes, list it under MISSING instead of writing something. Do not state any
figure that is not in my notes.Imagine the owner wants a grant towards a polytunnel and a water-saving irrigation system, total cost $18,500. The notes give the area (0.2 hectares), the current losses (about 15% of salad crop in a dry month, from the owner's records), and expected savings. The assistant produces a tidy 300 words and a MISSING list that names "evidence of match funding" and "planning status". Those two gaps are worth more than the prose, because a rejected form usually fails on a missing item. Your checks: read the funder's own page for the deadline, eligible costs and rules, and have a second person read the final form. The tutorial on drafting a farm grant application from your own data goes through the routine step by step. For larger applications, an adviser who knows the scheme is worth the fee. Time saved: 6 to 12 hours per application, in lumps, and nothing in a quiet month.
Use 4: crop and grazing planning
Planning is about order and timing: which crop follows which, when to sow, when the sheep move, how many days of grazing a paddock gives. An assistant can build the first draft of a rotation table or a sowing calendar from your constraints, and re-do it in seconds when something changes.
I farm 15 hectares of arable and vegetables in five equal blocks (A to E)
and need a four-year rotation. Rules: no brassicas after brassicas within
three years; a legume or green manure in one block each year; potatoes
no more than one block per year. Last year: A potatoes, B spring barley,
C brassicas, D beans, E fallow. Build a table for the next four years and
explain any rule you had to bend.An illustrative answer sets out the four years in a grid and notes that it had to place brassicas in block E in year one because the others were blocked by the three-year rule. The tutorial on crop planning with AI extends this to sowing dates. You check it against your own knowledge of the land, because the assistant does not know that block B holds water in winter. Grazing plans work the same way: give it paddock sizes, flock numbers and rest periods, and it drafts a move calendar. For the sheep, say 40 ewes on 1.5 hectares of grass for a three-day move, then check the result against what the grass can really carry, which only you can judge by looking at it. Cost: nothing beyond the assistant.
Use 5: weather and irrigation log
Decisions about spraying windows, harvest and watering turn on weather, and records of what you did and why protect you later. An assistant can combine your own log entries with weather data you paste in and give a summary: how many dry days, when the last soaking was, when the next window opens.
Paste a week of rainfall figures and your irrigation entries, then ask for a short table of water applied against rain received per bed, and a flag where a bed got neither. It will not forecast. It will not know soil type unless you say. What it gives is a clear weekly record in two minutes. Cost: the assistant. Time saved: perhaps half an hour a week in a dry season.
Use 6: messages to buyers and customers
If you sell direct, you write the same kinds of message every week: what is available, what is late, where to collect. Draft once, reuse with fresh numbers. For the farm's two contract buyers a short Monday note covering volumes and any issues takes five minutes to prepare from the harvest sheet.
Write a Monday availability note to two salad buyers. Facts: baby leaf
about 40 kg, rocket about 25 kg, lettuce heads 120 (none for Friday),
delivery Tuesday and Thursday, orders by Sunday 6 pm. Plain text, under 80
words, no emojis, no promises beyond these figures.Your check: quantities against the real plan, not the assistant's wording, since it may say "plenty" for a short crop. Cost: the assistant; time saved about 45 minutes a week.
Use 7: rotas and helper instructions
Seasonal help is usually the part of the year when organisation matters most and time is shortest. An assistant can turn a list of tasks and available people into a week's rota, and turn a rota into clear, short instructions each helper can read on a phone. Give it the tasks with durations, who can do what (not everyone can drive the tractor), and each person's available hours.
A typical slip shows the limit (illustrative). Told early in a long chat that one helper can work only Monday to Thursday, the assistant gives that helper a Friday shift in the third revision, because the rule was stated once and then buried under later changes. Re-state the constraints each time you change the rota and read every name against its days before you share it.
Use 8: reading reports and long documents
Soil test results, feed analyses, contracts and insurance schedules are long, jargon-filled and read rarely. An assistant can summarise a report, explain a term and list the numbers that fall outside the ranges the lab prints. It is a reader's aid, not an adviser.
Paste the lab's own table and ask: "Which values are outside the range the lab prints, and what do the abbreviations mean? Do not recommend products or doses." That last sentence is the point. An illustrative reply on a soil report might read: "pH 5.6 is below the range of 6.0 to 6.5 printed on the report; phosphorus is at index 1, below the target index printed beside it; potassium and magnesium sit inside their printed ranges." Check every value it calls out against the paper report, and delete any sentence that drifts into advice, such as a lime rate, because assistants sometimes add one even when told not to. For product choice, application rates, withdrawal periods and anything that must follow a label or law, use the product label itself and a qualified adviser. A wrong dose from a confident chat answer is the most expensive mistake in this list, and the routine in spray records from voice notes shows where the human check goes. Cost: the assistant; time saved about 15 minutes a week averaged over the year.
What the eight uses cost together
For the 25-hectare farm, a sensible bundle is one $20 assistant plan plus the $4.99 satellite plan for the salad ground: about $25 a month, or $300 a year. Hours saved across uses 1, 2, 4, 5, 6, 7 and 8 come to about 4 a week (illustrative: 1.0 + 0.5 + 0.5 + 0.5 + 0.75 + 0.5 + 0.25 = 4.0), which at $22 an hour is $88 a week, or roughly $380 a month. Grant work comes on top in the months it applies.
Be cautious with those hours. They assume the habit sticks and the checking takes the time described. A farm that dictates notes twice a month will save a fraction. If it takes you ten minutes a day to check outputs, take that off first. Even at half the hours the bundle covers its cost, but only if you use it, which brings us to the part that goes wrong.
The habit that fails first: the unchecked row
The typical failure on a farm is not a dramatic AI error. It is a quiet one. A voice note says "sprayed field three" and the assistant, asked to fill a product column, writes in a product name from an earlier note. The row looks complete and sits in the sheet for six months. When you need the spray record, it points to a product you did not use.
The rule that prevents it is the one repeated in each prompt above: leave a cell empty if not stated, do not infer. Add a second rule for yourself: any record that could matter to a buyer, an inspector or an insurer gets read against the original within a day. An empty cell you fill by hand is better than a full cell you cannot trust.
Three more farms, three different first choices
A 4-hectare market garden
Satellite monitoring is too coarse for beds this small, so skip use 2. Start with uses 1, 6 and 4: voice records, weekly availability notes for a box scheme and a succession-sowing calendar. A $20 plan covers it, and the saving is mostly in planning.
A 90-hectare arable farm with no staff
Here use 2 earns its place. Farmonaut's Basic plan covers 30 hectares and its Standard plan 80 hectares ($64.99 a month on monthly billing when checked), so 90 hectares needs a larger tier or two plans; compare the options on the vendor's page. Use 1 for operations records and use 8 for lab results are the other priorities. This is the profile closest to where surveys show adoption of precision tools is higher.
A small flock-and-hay farm
Livestock records have their own rules and dedicated apps. The tutorial on free AI for small farms explains what the free tiers handle before you pay for anything, and uses 3, 5 and 7 give the most value: grants for fencing and water, a log of grass and weather, and helper instructions at lambing.
How to try one use this week
- Pick the use that matches your worst weekly chore. For most owners that is records or messages.
- Run it free for one week. Time the chore before, and the chore plus checking after.
- Keep the unedited originals, so you can see what the assistant changed.
- After a week, decide: keep, change the prompt, or drop it. Only then consider a paid plan or a satellite service.
If you would like help working out which of the eight fits your farm and how to set it up so you can run it yourself, that is the sort of job my AI implementation consultation covers, and the contact page is the way in. Whatever you choose, keep the rule that has run through every example here: the assistant organises and drafts, and you decide, check and sign.
Questions about farmers and AI
What share of farmers use AI?
A 2026 survey of 5,500 farmers in ten countries, published by McKinsey in September 2026, found that 17% use generative AI for farm tasks and 4% pay for a solution. Older precision-agriculture surveys measure different tools, so the figures cannot be added together. Treat all of them as a snapshot, not a ceiling.
Do I need a drone or sensors to use AI?
No. The cheapest uses need only a phone and a chat assistant: voice notes turned into records, grant drafts and planting calendars. Satellite monitoring needs no hardware either, since it uses images from space. Drones and sensors add cost and are worth considering later, if you can name the decision they would improve.
Is AI advice on crops or animals reliable?
Not as the final word. A chat assistant can explain terms and organise options, but it does not know your soil, weather or herd, and it can state wrong things confidently. For treatments, doses, animal health and legal limits, confirm with your agronomist, vet or the product label.
Can AI fill in grant applications for me?
It can help you draft the narrative sections and check them against the scheme's criteria, but you must supply the facts and verify deadlines, eligibility and figures on the funder's own page. Never let it invent numbers about your farm. A grant adviser is worth asking for larger applications.
What is the cheapest sensible way to start?
Use a free or $20-a-month assistant for one week of voice-note records. It costs nothing in hardware, shows quickly whether you will keep up the habit, and builds the data you will need if you later add monitoring or planning tools. Decide on anything bigger after that week.
Further reads
- Can Farmers Use AI for Farm Records and Paperwork? — The records and paperwork side of farm admin.
- AI Chatbot for Farming: What It Can Answer and Where It Fails — What a chatbot can answer for a farmer and where it fails.
- AI Chatbot for Farmers: Answers From Your Own Farm Notes and SOPs — Ask questions of your own notes and routines.
- AI for Farm Business Management: A Monthly Margin Review by Enterprise — A monthly margin review by enterprise.
- AI Farm Business Plan: Draft One for a Lender From Real Records — Turn real records into a plan a lender can read.
- Farm Safety Plan: Draft Risk Assessments and Toolbox Talks With AI — Risk assessments and toolbox talks for farm staff.
- How AI Can Help Farmers Save Water: Build an Irrigation Log — Build the weekly water and rain log step by step.
- AI Automation for Agriculture: Rotas, Timesheets and Harvest Messages — Rotas, timesheets and harvest messages for seasonal help.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: McKinsey Global Farmer Insights 2026 (published 8 September 2026), as reported by Bloomberg and other news summaries; a national government farm-technology chart of 10 December 2024 (2023 data); Farmonaut pricing page (monthly prices, read 5 November 2026); ChatGPT and ChatGPT Business pricing pages (read 5 November 2026).