Customized language models and service channels are some of the key tools offered by the French retail chain in the South American country.
“The physical store is not going to disappear. E-commerce is simply growing, but without cannibalizing the physical format. There are more branch openings, but many people realize that for a macro purchase they can use digital platforms,” is how Juan Nesis, Chief Data Officer of Carrefour Argentina, describes the current state of the retailer in the country.
The French supermarket chain is currently one of the leading multi-format retailers in the South American giant with more than 600 branches under the brand names Carrefour, Market, Express and Maxi. Much of this expansion is due to a commitment to digital tools such as generative artificial intelligence and machine learning . This is a vision that is part of the concept of a digital retail company , proposed by the French subsidiary at the beginning of 2024.
“This idea is 100% implemented in the number of projects that are now emerging in the different countries where Carrefour operates with digitalization. For example, launching apps for customers or employees, as well as implementing laboratories with AI solutions throughout the company,” Nesis told AméricaEconomía within the framework of Ned24, organized in Lima.
In other words, Nesis believes that a digital retail company understands how it can make connections between customer needs and new technologies such as generative AI and cloud applications. It is an approach that has brought results to Carrefour in recent years, especially in Latin America.
For example, in the second quarter of 2024, the company accumulated sales of US$ 25,194 million, 10.8% more at constant rates. Likewise, its revenues rose by 44.5% in the same period, marking a trend opposite to the sales declines experienced in European markets. However, each Carrefour subsidiary operates independently, which allows autonomy in digital operations and establishing plans for a single market.
According to Nesis, this factor is a great advantage for Carrefour Argentina, but it does not eliminate the challenge of thinking about how to apply digital tools in sustainable strategies. “The big challenge is whether an internal solution can be used by Carrefour’s 17,000 employees in the 667 stores. It has to work well, have good performance . So, the challenge is basically infrastructure,” says Nesis.
In terms of digital solutions for customers, in 2023, Carrefour France launched Hopla, an AI-powered chatbot that aims to improve the shopping experience. It consists of a robot that offers individualised shopping advice by creating product descriptions and another that supports internal purchasing processes. This system is the product of a collaboration between Carrefour and Microsoft and the consultancy Bain & Company, OpenAI partners. Ultimately, Hopla operates as a source of suggestions for defining the shopping basket based on each customer's budget.
But in Argentina, Carrefour has also devised AI-based solutions for the company's staff. “Today, language models have been very useful for us to process the mass of information coming from customers and to extract insights that we couldn't see before. So our experience in building internal chatbots has been super positive and we are now building an external chatbot for customers through WhatsApp to be the main channel,” Nesis says. The ultimate goal is for customers to be able to manage retail or wholesale purchases through the popular messaging app.
In fact, there are already initiatives underway to meet specific customer demands. At the beginning of 2024, a chatbot was created to process user feedback dynamically. Thanks to this tool, Carrefour discovered that despite the abundance of sweet rolls in its branches, the truth is that many people wanted the savoury one. This is a clear example of how retail companies can take advantage of machine learning to boost their sales on a large scale. Another advantage of this technology could be the processing of internal documentation.
It is a complex process, but it follows an identifiable path that other retailers can imitate. “Language models are like large neural networks: they receive information, pass it through different layers and give us something back. At Carrefour, when we work with artificial intelligence, we don’t build a neural network, but we take existing ones. For example, ChatGPT, which is a gigantic neural network that has already been trained. What we do is take it through a Python code as a programming language. Then, we fill it with our data so that it can train itself with the information,” explains the manager.
On the other hand, among the projections for 2030, the head of data assures that Carrefour plans to spread its machine learning and image or audio processing models more widely to customers. In recent years, these tools have been applied to the company's staff. In Argentina, the company even has training programs in information technology (IT) for work teams. “There are internal training sessions that we do at Carrefour's headquarters, at the central office, where there are more than 1,000 employees. We sometimes get together there to learn about certain topics,” says Nesis.
