When he had to choose the subject of his thesis, Juan Manuel Baruffaldi thought about his father’s farm, a small operation in southern Santa Fe. Herbicide was applied across the entire field, even though weeds covered only 10% or 15% of the area. He wanted to find a way to treat only those patches.
Baruffaldi had studied Computer Science and worked with facial recognition. He wondered whether he could use the same technology to distinguish a weed among crop plants. He began developing the system and, in 2017, founded DeepAgro in Casilda.
That led to SprAI, which is fitted to the sprayer the farmer already uses. Cameras are mounted on the boom, one meter above the ground, and detect weeds as the machine moves through the field. Using those images, the system decides which nozzles to activate to apply the herbicide. The operator starts it from the cab.
According to DeepAgro, the system reduces herbicide use by 70% per growing season. It has already operated across more than 1.2 million hectares and has over a hundred clients, from Chaco, Salta and Santiago del Estero to southern Buenos Aires province. At the Aapresid congress in Rosario, the company presented financing options for those who need to spread out the purchase cost.
Money no longer spent on liters
To explain how much can be saved, Baruffaldi shows his clients’ results. Some reduced herbicide spending by as much as $150,000, depending on the area they farm. As a benchmark, he estimates savings of forty dollars per hectare per year or sixteen dollars for each selective application.
The payback period varies according to the farm and how the machine is used. He describes a farmer with a thousand hectares who recovered the investment in one year, while another with 2,500 hectares took two.
The equipment also changes how spraying is carried out. Some farmers made a single pass using the dose needed to eliminate the most resistant weeds. They ended up applying that same amount throughout the field, even where less would have been enough.
With selective spraying, they can treat small weeds first and return fifteen days later to target those that remain. According to Baruffaldi, savings on that second pass can reach 95%, with more effective treatment.
Reducing the amount of herbicide also leaves the crop less exposed to potential damage from the product, known as phytotoxicity. Fewer chemicals are also applied to the environment.
Baruffaldi explains that equipment purchased in 2022 can run the same software as a new unit without replacing the installation. Images collected in the field help improve weed detection. Those improvements then reach other users through updates.
The first field in history
Between 2017 and 2019, the five founders financed the project with their savings. They worked half days at another company and then continued development in Casilda. The remaining test was to take the equipment into the field and see how it performed.
In November 2019, they took it to the soybean field on Baruffaldi’s father’s farm. The application reduced herbicide use by 82%. DeepAgro describes that experience as the first selective spraying application using artificial intelligence in soybeans.
The pandemic interrupted their travels. When they could go out again, they loaded the prototype into his father’s red Ford pickup and traveled through the northern farming region. They visited Quimilí, Charata and other towns to demonstrate the equipment at work.
The first investors were farmers and agricultural businesses that had seen the trials. Some became clients in 2021, and commercial sales began in 2022.
DeepAgro now has around fifty people, almost all in Argentina. Before starting to sell in Brazil, the team traveled 5,000 kilometers through Mato Grosso in ten days to learn firsthand how farmers worked and what they needed.
In the United States, they heard a comparison Baruffaldi often mentions: Argentina had helped spread no-till farming and could once again contribute a technology capable of changing agricultural practices.
Paying for the machine with what it saves
The agreement with Nera provides financing for purchasing the equipment, with the idea of covering installments through herbicide savings. This allows farmers to adopt it without paying the full amount upfront.
Farmers can buy the equipment from DeepAgro or its distributors. The company also works with Metalfor, an Argentine sprayer manufacturer, and is negotiating with manufacturers abroad to have SprAI incorporated into their machines before they are sold.
Meanwhile, it is testing new uses. The system already distinguishes broadleaf weeds from grasses. Some clients are trialing selective insecticide and fertilizer applications, planned for the next growing season.
Requests often emerge after the first applications. Those who see the savings start asking about field data, drone surveys or insect detection. Baruffaldi says that contact with users, from installation through after-sales support, helps the team decide what to develop and what to correct.
In Rosario, he announced that they aim to add between one and two million hectares in the next stage. If they reach two million, they will have added more land than the entire area covered since the 2019 trial on his father’s farm.
