The agricultural sector consists of two kinds of work that have little to do with each other. On one hand the work in the barn, the greenhouse or the field: feeding, milking, harvesting, machine maintenance, observing animals for behaviour and health. On the other hand the work at the desk: manure accounting, subsidy applications, crop registration, invoicing, planning deliveries, communication with buyers and the accountant. Anyone wondering which work AI will take over first must keep these two streams separate, because the outcome is completely different for each.
The physical work changes slowly. Milking an animal, pruning a crop or driving a tractor is still largely human work, even when sensors and robots are already involved. The administrative and analytical work changes faster, and it is precisely this work that takes up more hours at many businesses than the owner realises. A dairy farmer who knows how much time per week goes into filling out forms and retyping data between systems often sees for the first time where capacity can be freed up.
On an agricultural business, the hours do not only go to production. A large part goes to:
These tasks are rarely part of the conversation about AI on the farm, while they often form the largest share of an agricultural entrepreneur's office hours. Assessing which part of this work can be taken over depends on how structured the data already is, how many systems run independently of one another, and how often a task is carried out based on fixed rules versus based on experience and judgement.
At businesses where data comes in digitally and in a structured form, software can already take over tasks today such as compiling manure accounts from sensor data, automatically filling out standard forms, flagging deviations in milk yield or growth curves, and proposing schedules based on weather forecasts and available capacity. This is work where the input is unambiguous and the outcome can be checked against a standard.
Other tasks remain under human oversight: a system can flag a deviation in animal behaviour, but the assessment of whether this means illness, stress or something else remains with the livestock farmer or the veterinarian. A system can compare a quote for fertiliser or feed on price, but the choice of supplier also depends on relationship, delivery time and experience from previous seasons. This oversight is not an intermediate phase that disappears once the technology improves; it is a fixed place in the process for as long as the assessment requires context that is not in the data.
And part of the work remains human work, full stop: the physical handling of animals and crops, machine maintenance, judging the moment to harvest based on smell, colour and feel. Technology can support this with data, but the action itself does not shift.
Two agricultural businesses of comparable size can differ greatly in what can already be taken over now. A business that already works with accounting software, sensors and a central registration system has the foundation in place to automate tasks. A business that still works with paper forms and separate Excel files must first put that foundation in place before AI can take anything over. The difference therefore lies not primarily in the sector, but in the state of the data and systems within the business.
This pattern is not unique to agriculture. In the ICT sector and in financial services too, the quality of the underlying data determines how much work is actually taken over. Where an agricultural business struggles with fragmented sensor data, a marketing department often struggles with a comparable problem when scoring leads on quality: the technology is there, but the input must be in order before it can take anything over.
This analysis says nothing about who a business should employ or dismiss. It is about tasks and hours, not about roles or people. If a shift in work has consequences for staff, separate legal requirements apply on which this page makes no statement.
The first step is to get a clear view of where the hours in your own business go: how many to registration, how many to planning, how many to the physical work itself. Only with that overview can it be assessed which part of the work can already be taken over today, which part can be done with oversight, and which part remains human work.
The free quickscan from FTE TO AI provides an initial indication of this: twelve questions, no account required, with an estimate of what share of the hours in your profile can be taken over by AI today. The full work scan, which breaks down a business's work to task level and calculates fte capacity per task, is still under construction.
Vraag maar. Ik ken de kennisbank van deze site; wat ik niet weet, zeg ik erbij.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.