Fewer labour migrants in horticulture: what automation and AI do and do not solve

Why this dependence is now a risk

The dependence itself is not new. Much of the work in horticulture is seasonal, labour-intensive and repetitive: harvesting, sorting, packing, crop work. For years, that work has been done by labour migrants, mainly from Central and Eastern Europe. What is changing is that this inflow can no longer be taken for granted. The Dutch government has started restricting labour migration, and according to Groenten & Fruit Actueel (a Dutch trade journal for fruit and vegetable growing), for the time being this is not being done through a ban, but by making low-skilled labour more expensive, partly to make robotisation more attractive.

For a grower, that means two things at once. The availability of staff becomes more uncertain, and the cost per hour worked rises. On top of that, according to Rabobank, unclear government policy is currently growers' biggest concern. Taken together, that means automation is no longer a luxury or a pet project, but a way to make the business less vulnerable to decisions you cannot influence yourself.

What automation of cultivation, harvest and sorting can do today

Automation in horticulture is not an all-or-nothing choice. It is a matter of individual tasks being taken over by machines step by step, at a pace that varies greatly by crop and by task. Sorting and packing are the most advanced: optical sorting machines that sort by colour, size and visible defects are already common in many businesses. Internal logistics, such as the transport of trolleys and crates, is likewise well suited to automation.

Harvesting is harder. A robot that recognises delicate fruit, judges whether it is ripe and picks it undamaged from between leaves and vines is technically demanding and differs from crop to crop. For some crops there are working harvesting robots, while for others it is still very much in development or not yet economically viable. The honest picture is therefore mixed, not triumphant.

• Sorting and packing: optical sorting by colour, size and visible quality is already common practice in many businesses.
• Internal logistics: automated transport of trolleys, crates and products within the greenhouse or the hall.
• Crop work: partial automation, heavily dependent on crop and growing system.
• Harvest: working for a few crops, for others still in development or not yet profitable.

Where AI makes the difference: planning and quality

The biggest gain is often not in a spectacular robot, but in making smarter use of data you already have. Two applications stand out. The first is planning: if you have a more reliable picture of how much product comes off the crop and when, you can deploy harvesting and processing capacity in a more targeted way. That is exactly where labour is scarce and expensive, so every peak or idle period you avoid counts immediately there.

The second is quality. AI can assess images of products for defects that are hard to see consistently with the naked eye, and so support the human inspector. The aim is not to replace the human, but to make the assessment more consistent and to flag defects earlier.

In both cases, AI supplies an estimate or a signal, and the grower or team leader decides what is done with it. The model does not fully know the context of your business, such as a customer with a non-standard specification or a batch that reacts slightly differently.

No miracle cure: what to be critical of

Much is promised about automation and AI in horticulture. Be critical of vendors who suggest that a harvesting robot or an AI system will solve your labour problem in one go. In practice, automation usually replaces some of the tasks, not the entire job, and the work shifts to operation, maintenance and supervision of the technology. That calls for different staff, not for no staff.

Also be critical of savings claims and payback periods. The effect varies greatly by crop, by growing system and by business, and a generic percentage says little about your own situation. Ask a vendor for a trial with your own product and your own figures rather than for a sales brochure. When you do the sums, include not only the price of the machine but also changes to cultivation, maintenance and the people who have to work with it.

The human decides and signs, AI literacy included

Automation and AI shift the work, but not the responsibility. For quality assessment and cultivation decisions, the grower remains ultimately responsible, even when an AI system looks over their shoulder or makes a proposal. That is not only a matter of principle, but also a practical one: a batch that is wrongly let through or a missed defect affects your reputation with the buyer.

In addition, Article 4 of the EU AI Act (Regulation (EU) 2024/1689) has applied since 2 February 2025. Since the amendment by Regulation (EU) 2026/1744, in force from 27 July 2026, it has asked organisations to take measures that support the AI literacy of employees who work with AI systems. They do not have to guarantee a specific level. For a horticultural business, this means in practical terms that whoever operates a sorting or quality system, or assesses its output, understands what the model is based on, where its limits lie and when a person must take over the judgement. That applies equally to the permanent core of staff that remains once the routine work has been partly automated.

Where to start in practice

Do not start by asking which robot is the most impressive, but by asking where labour in your business is scarcest, most expensive or hardest to plan. That is usually where automation or AI pays off soonest. Then establish which data you already have, for example harvest records and sorting data, because that is precisely what planning and quality applications run on.

Then choose one specific process and first record how it is done now: how many hours it takes, who does what, and on what basis. Only with that baseline measurement can you judge whether an investment in automation or AI measurably saves time or money or reduces vulnerability. That way, the next step rests on your own figures and not on a promise; the free Practice Scan helps with that baseline measurement.