Ifood and Prosus announced the launch of the Large Commerce Model (LCM), a large-scale generative artificial intelligence model. Developed and optimized for digital commerce, the LCM customizes the user experience from product searching and recommendation to the creation of customized consumption journeys.
Based on language models (LLMS), the LCM learns from billions of real interactions such as clicks, searches and purchases, to identify behavior patterns and interpret users' purchase intentions through anonymous data. Technology evolves with each use, recording actions, reasoning about them and improving with each interaction.
The LCM transforms the and e-commerce, which was previously based on searches for keywords and generic recommendations. With it, the consumer has a personalized experience, as if he had a “personal shopper”. The technology works like a “digital brain” capable of going beyond the simple recommendation of products, revealing the “why” and the “how” of each order. This is possible because the model has long-term memory and is trained with data that reflects Brazilian reality and culture.
According to Diego Barreto, CEO of iFood, the LCM was trained in the database of Prosus' global business transactions, learning from more than 500 million users and more than 10 trillion data tokens. The model is integrated with QWQ3, an open source model with 32 billion parameters.
first results
The initial results with the LCM already show a positive impact. The conversion rate in the “You can also like this” suggestions grew by 66%. In addition, custom LCM-driven notifications resulted in an increase in clicks that generated four times more requests, surpassing previous records. Technology is also competitive in relation to other market references, being 60 times cheaper to operate than other generative AI models.
The LCM was developed to be agnostic and scalable, which allows its advances to be applied to other Prosus Group businesses, such as Decolar, Sympla and OLX.
iFood launches Ailo, AI assistant that allows requests for conversation
iFood announced the entry into the “age of agents” with the launch of the there, a generative artificial intelligence assistant who inaugurates a new way of making requests: through conversation. Ailo understands each user's preferences and offers personalized suggestions, making the experience of asking even faster and more practical.
The assistant goes beyond the understanding of direct intentions, managing to decipher subjective desires of consumers. For example, it is already possible to say “I want a romantic dinner today” or “I would like a light food that arrives in up to 30 minutes” and receive appropriate options.
For the first time, orders can also start outside the app. Ailo is in public tests within the platform and also on WhatsApp, which expands the possibilities of interaction with ifood. As it is agnostic, the wizard will be able to work on different channels and platforms in the future.
Tests with AILO began in June 2025, with more than 70,000 people participating and more than 100,000 interactions registered. The first results point to significant gains:
There are 48% more chances of a search to convert into an order compared to the traditional stream of the application.
The process to finalize a request with AILO on WhatsApp is 33% faster than the current application flow.
The expectation is that, with more interactions, the Aylus will become progressively more intelligent and efficient.
Technology behind Ailo
Ailo personalization is driven by the Large Commerce Model (LCM), developed in partnership with Prosus. LCM is the technology foundation that makes Ailo increasingly intelligent, as it provides iFood with an in-depth understanding of the choices and intentions of its 60 million users. Technology advances beyond the simple recommendation of products, revealing the “why” and “how” of purchases, which results in an intuitive experience.
Ailo is built on a hybrid architecture that combines different AI models such as Anthropic, OpenAI and AWS. Each of these models acts in different layers, from the understanding of natural language to the generation of responses, ensuring robustness and speed. The difference lies in the integration with the LCM, which combines the fluidity of conversation of global models with the unique customization for the Brazilian consumer.
To participate in the Ailo tests, users can send a “hi” message to WhatsApp at +55 11 91150-4025, in addition to having a profile on ifood.
