From weighing scale to visual recognitionFrom weighing scale to visual recognition

In supermarkets and wholesale stores, weighing fresh produce always adds friction: the shopper brings fruit and vegetables to the scale, someone weighs them and prints the label, and only then can the shopper go to the checkout. It's slow, error-prone and depends on staff being available at the weighing station.

Systel, Argentina's leader in commercial weighing solutions, saw a clear opportunity in that flow. If the scale could recognize on its own which product was on the tray, the whole process would get simpler: shorter lines, fewer pricing errors and a smoother experience for shoppers.

The problem was that no one at the company had ever worked with artificial intelligence. The technical know-how was there, but of a different kind: programmers, electronics engineers, industrial designers. Building an image recognition system from that base, one that could work in real supermarket conditions, was an entirely new challenge for the team in 2021.

The first move was to look for outside vendors to develop the algorithm. Systel approached several Argentine companies specializing in computer vision, but the quotes were very high and none could show similar work done before. In every case, development would be a shared learning process, with no guaranteed outcome.

Faced with that, Systel decided to build it in-house. The reasoning was simple: if the learning was going to happen anyway, it was better for that knowledge to stay inside the company.

The R&D effort produced Detect, launched in early 2025. It is a self-service scale with built-in automatic image recognition: the shopper places the product on the tray, the camera captures the image, and the algorithm identifies it, calculates the price and prints the label. The process is instant and requires no assistance.

The system was designed with a strong focus on usability: the interface is clear, the flow is intuitive and people of any age can use it without difficulty. It is sold mainly to supermarkets and wholesalers for their produce sections, and the model has already been adapted to recognize cheeses and cold cuts as well.

In parallel, the R&D team began bringing AI tools into its own internal processes: coding assistance, software development and automated quality-control testing. AI first entered the product and from there spread across the team's entire operation.

The most immediate impact was financial. The recognition system's accuracy directly reduces product identification errors at the point of sale: cheeses or cold cuts weighed at the wrong price mean real losses for the retailer. In high-volume operations with many similar products, the difference shows quickly.

In one case documented by the team, the owner of several wholesale stores recouped the cost of the equipment in the second month of operation, thanks to fewer errors when selling variable-price products.

Beyond the product, Systel now has a team that knows how to build, train and scale image recognition models, with its own platform for updating those models from anywhere in its network of installed equipment. That technology asset opens the door to new applications in segments the team is still exploring.

The company trained all its teams in using AI tools, with one core message: the technology is here to make work more productive, not to replace people. It's the same approach the company took when it introduced Office or email.

The response was positive. Most staff already used some AI-powered tool in their daily lives, which made adoption easier. At the same time, IT security policies were put in place to protect corporate knowledge. Leadership actively drives this culture: the CEO takes part, managers encourage everyday use and the R&D team shares experiences with other departments through internal talks.