An agricultural business, whether it is an arable farm, a greenhouse horticulture operation, a dairy farm or a contracting business, runs on a combination of physical work outdoors or in the barn and administrative work at the kitchen table or in the office. The hours go into crop monitoring, livestock care, planning of cultivation or milk flow, registration for certification and subsidy, and administration around manure policy, crop protection products and animal welfare. Weather, soil, disease pressure and market prices steer the outcome more than in many other sectors, which makes the work less repeatable than it appears on paper.
That combination of extensive rule-bound administration and extensive circumstance-bound physical work is exactly where AI ambitions in this sector now run into difficulty. The ambition often sounds simple: less time on forms, more time on the crop or the animal. The question is whether the business is already ready for that.
Three categories of work run through every agricultural business, and the distribution differs greatly per business.
The first part is work AI can already take over: filling in registrations based on sensor data, compiling a manure accounting record from existing figures, recognising deviating growth patterns in satellite images. Businesses that already record their data in a structured way see this part shift the fastest.
The second part is work where AI makes a proposal and a person approves or rejects it with reason: an advisory dose of crop protection, a feed schedule for the livestock, an estimate of harvest timing. Oversight remains necessary here because the consequences of a wrong choice directly affect yield or animal welfare, and because local circumstances can rarely be fully captured in data.
The third part remains human work: assessing a sick animal based on behaviour, negotiating with a buyer, making a cultivation decision under uncertain weather conditions. This is work where experience and responsibility come together in a way that cannot be handed over to a system.
The difference between businesses lies not in the sector but in readiness. A business with documented protocols, structured data and a team that knows who is authorised to reject an AI proposal shifts work faster than a business where that foundation is lacking, even if both businesses state the same ambition.
The cleaning industry and the recreation industry also involve physical work under varying conditions, but there the regulatory burden is lighter and the margin often thinner. In the agricultural sector it is the reverse: the regulatory burden is heavy and detailed, and precisely for that reason registration work is relatively well suited to being taken over by AI, while the physical judgement of crop and animal is not. Anyone wanting to know how that ratio plays out in other sectors can compare it with the ambitions in the IT sector or with the ambitions in financial services, where the work happens largely behind a screen and the readiness question is different.
The management of an agricultural business often speaks in a sentence everyone understands and no one fills in the same way: "we want less time spent on administration." One person means full automation of the manure accounting record, another means a faster form that a person still fills in themselves. Both are legitimate ambitions, but they require different capabilities: different data infrastructure, different decision rights, different oversight. Without laying down that ambition in layers, from vision to concrete target state, the discussion about what exactly needs to happen remains undecided.
This is also where decision rights need to be sharply defined: who is authorised to reject an AI recommendation on crop protection, and on what grounds. You record that in the manner described in how you record decision rights during a change. And because automating administration saves costs but does not automatically affect revenue, it is useful to think through in advance how growth relates to margin in a strategy, so that the ambition is not only about hours but also about where that freed-up capacity should go.
If an AI ambition affects the workforce structure, separate statutory requirements apply to that; that assessment falls outside what a readiness check does.
The underlying question is always the same: which work in this specific business can truly be taken over by AI, which part requires ongoing oversight, and which part remains human work. That question is answered per task with the work scan from FTE TO AI, not with a sector average.
Before a business turns an AI ambition into a project, it is useful to know where the organisation is already ready and where it is not. The free readiness check consists of eight short questions, one per dimension, and produces a picture of where you are furthest along and where the gap is largest. The full ambition check, with the four layers and the five confidence gates, is under development.
Vertel wat u wilt bereiken, dan kijken we samen wat daarvoor moet staan.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.