Lexia and its Voice AI Gilbert: The French Revolution of the Office
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
An Announced Revolution but Still Incomplete
At the end of 2022, the arrival of ChatGPT sparked great expectations regarding the transformation of businesses through artificial intelligence. However, in 2026, the reality is more nuanced. Although AI has penetrated the professional world, its use often remains limited and sometimes even rudimentary.
Mathis Escriva, co-founder of Lexia and only 22 years old, attributes this situation to the complexity of text-based interaction. According to him, a gap persists between technological advancements in the lab and their concrete adoption in businesses. Traditional computing forced users to adapt to text interfaces, but current generative AI requires rich contexts in the form of prompts, an exercise that can prove difficult for many. "People were not comfortable with the idea of typing their requests to the AI," he explains, based on his observations during training sessions and an internship at a major CAC 40 company.
Lexia and Innovation Through Voice
To address this challenge, Mathis Escriva and his team created Lexia with a bold idea: to eliminate the keyboard from the equation. Instead of forcing users to master the art of prompting, why not let AI adapt to our most natural mode of communication, voice? Thus, Gilbert was born, a tool aimed at marrying the efficiency of AI with the spontaneity of oral expression. Lexia considers text as a "by-product," while voice provides the necessary context and nuances.
Gilbert: The Intelligent Dictaphone
Gilbert presents itself as an "intelligent dictaphone" capable of capturing the "vocal capital" of businesses. Every day, numerous crucial pieces of information are exchanged orally, leaving no written trace. Mathis Escriva emphasizes that "80% of decisions are made by voice, without even going through an email."
Whether during meetings, client appointments, or informal discussions, Gilbert allows users to capture these exchanges with a single click on a smartphone or computer, instantly transforming them into structured and actionable data. The tool is designed to be intuitive and easy to use, requiring no complex training. "A technology that requires training is a bad technology," asserts Escriva. With Gilbert, the user speaks naturally, and the AI takes care of structuring the notes, identifying actions to be taken, and integrating them into tools like Slack, Notion, or CRMs.
Sovereignty and Energy Efficiency at the Heart of Innovation
Unlike other market players that rely on American APIs, Lexia develops its own models using open-source Small Language Models (SLM). These models, more specialized than LLMs, allow for a tailored approach suited to complex industry jargon. "We have complete control over our technological pipeline," explains Escriva, illustrating how they combine a high-performing model in mathematics with another skilled in storytelling to create an excellent math teacher.
This approach ensures an accuracy rate of 97% to 99% and guarantees data sovereignty, crucial in a context where voice is a sensitive biometric data. The data is processed in France, and can be hosted on a secure private cloud or directly on clients' servers.
Moreover, Lexia's SLMs consume ten times less energy than generalist technologies, without compromising performance, thus facilitating rapid and efficient integration.
Government Support and a Promising Future
Since its launch in January 2026, Gilbert has experienced rapid adoption, with already 900 active users. This growth is supported by Bpifrance and a recommendation from the French government to its administrations. Lexia offers flexible solutions, ranging from an individual subscription at 20 euros to custom packages for specific needs, such as those in the medical sector, which has data hosting constraints.
To maintain this momentum, Lexia recently introduced the Model Context Protocol (MCP), allowing Gilbert to connect directly to clients' preferred AIs and their business tools. "We offer a tool that communicates with all the systems in the company," Escriva rejoices.
In addition to the software, Lexia plans to physically equip meeting rooms through hardware partnerships, integrating professional microphones to optimally capture audio. The goal is to provide businesses with a complete device, transforming oral communication into structured, ready-to-use data.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.