Valtra puts crop-sensing AI on the tractor order form

The next AI purchase on a farm may arrive already bolted to the tractor. Valtra is offering PerPlant’s crop sensor through its Unlimited factory-customisation programme, with sales starting in October. It is an option for the A105 and A115 and G, N, T, Q and S series in selected European markets; customers elsewhere can enquire about availability.[1]

That changes the purchase as much as the technology. A farmer ordering the option can have the sensing equipment fitted with the tractor. Fewer installation choices fall to the customer after delivery. For a small technology supplier, appearing in an established manufacturer’s ordering process is also a route into a machinery purchase the farmer is already making.

From a crop image to an application decision

The sensor is a roof-mounted box with two cameras. PerPlant’s Agronomy on the Edge demonstration project explains the operating loop. Cameras observe the crop, an onboard computer identifies plants and weeds, and the system combines those observations with positioning and historical satellite information. It calculates application instructions locally and sends them to the implement. Cloud software is used to review field maps afterwards. Keeping the immediate calculation on the tractor means a patch of poor mobile coverage need not interrupt that decision.[2][1]

Seeing a weed is only part of the job. The sprayer or spreader has to understand the instruction and be capable of acting on it. The link uses ISOBUS, the agricultural machinery communication standard. Its industry foundation, AEF, explains that interoperability depends on the functions supported across the connected equipment. An ISOBUS label alone does not establish that every tractor, terminal and implement combination can perform every function. Buyers should establish what their actual combination supports.[3]

This makes the implement an economic part of the AI system. A camera may distinguish a small patch within a field, but the value of that distinction depends on how selectively the attached machine can apply a product. The useful buying question is what the complete equipment combination will do on the farm’s crops and tasks. Image recognition has to end in an agricultural operation.

The ESA project record describes pilots in Denmark, the UK and the Czech Republic during the 2026 spraying season. Its account also distinguishes the sources of information: satellite records give historical field context, while cameras supply the view at the tractor. That combination supports a system which can use previous observations without waiting for a remote computer to interpret every new image.[2]

Read the savings against the application

Valtra cites PerPlant’s reported input savings of 40–87%, depending on the implement. These are supplier figures about affected inputs, not a reduction in the farm’s entire cost base. The launch page does not provide the trial design needed to turn that range into an independent estimate for a particular farm.[1]

The return will depend on how much land receives a relevant treatment, what that treatment costs, and how much can be avoided while achieving the required result. Saving a high proportion of one product does not remove the tractor, labour or land costs. Conversely, a modest saving repeated over many hectares can be useful. The sensible comparison is the farm’s existing application programme against the same job with the complete sensor-and-implement setup, including its purchase and running costs.

A second product travels with the tractor

The cameras also create maps of biomass, weeds and field conditions during ordinary tractor work. Valtra says this can gather observations without a dedicated mapping pass. A journey undertaken to do one job can therefore leave information for the next.[1]

In Jennifer Marston’s 14 October interview for AgFunderNews, PerPlant chief executive Rasmus Emil Hansen described a further business called Nexus, intended to let agronomists and agricultural suppliers work with the field data and build tailored AI tools. He said the company was aiming for production in early 2027. That is a planned extension of the data business, separate from the factory option customers can now order.[4]

The two uses reinforce each other in a practical way. A sensor bought to make an application more selective can keep collecting observations during other field work. Those records may help the farmer and advisers decide where to look or what to measure next. Factory fitting makes the initial purchase easier to organise; repeat observation gives the equipment a role that can outlast a single spraying decision.

Report an error or suggest a correction