YBA.ai : the French AI agents that challenge the global tech giants.

The Normandy startup YBA.ai, in collaboration with INSA Rouen Normandie and the LITIS laboratory, announces top-performing capabilities of its contextual AI agents, an advanced contextual reasoning benchmark that tests the ability of AIs to reason by cross-referencing multiple sources of information. YBA.ai achieves 59.08% correct responses on the MuSiQue benchmark—matching Microsoft's trained model and surpassing Salesforce, Google, Huawei, and Peking University. With the highest scores in Exact Match (53.2%) and F1 (69.5%), YBA with its MAGAR technology sets a new standard in augmented reasoning.

Résultats de comparaison

Détails de l'étude :

https://www.yba.ai/articles/yba-magar-surpasses-industry-benchmarks-outperforming-leading-tech-giants

Reda El Messaoudi Aubert, cofounder of YBA, claims that it has never been easier for businesses to automate their time-consuming business processes and to support their employees daily in high-value tasks.

1. AI agents rooted in reality

At YBA.ai, AI agents do not just generate text. They connect to internal sources (CRM, ERP, document databases, emails, customer histories...), analyze the business context, and then produce concrete and justified actions to achieve objectives. A true revolution in businesses!

In sales, AI agents respond to customer requests by leveraging the knowledge base, the tone of interaction, and the account history; they automatically generate business strategies and quotes in just a few minutes, analyzing the company's history, competition, and business rules; and they write personalized responses to tenders. On the customer service side, they read incoming emails, consult the company’s knowledge base, the ordering system, and provide personalized responses (procedures, availability, delivery), while prioritizing ambiguous cases. These capabilities extend to all teams — operations, finance, logistics — to free up time and allow employees to focus on what matters. “Our agents are capable of understanding the business, terminology, and constraints of each user. They adapt to reality, not the other way around,” explains Reda El Messaoudi Aubert, co-founder of YBA.ai.

2. À propos du benchmark MuSiQue

The MuSiQue (Multihop) benchmark was developed by researchers such as Arsh Trivedi, Niranjan Balasubramanian, Tushar Khot, and Ashish Sabharwal. It represents one of the most complex tests, as it evaluates artificial intelligences by challenging them with advanced reasoning capabilities. Unlike simple question-and-answer, a MuSiQue question requires the system to:

Raisonne à Travers Plusieurs Documents : Les informations nécessaires pour répondre à la question sont dispersées et doivent être récupérées à partir de différentes sources.

Integrate Evidence Elements: The system must perform intermediate reasoning steps and connect semantically diverse facts to formulate a coherent final response.

3. À propos de YBA.ai

YBA.ai develops contextual AI agents that adapt to the specific needs of each business. In collaboration with INSA Rouen Normandy and its LITIS laboratory specialized in AI agents. YBA designs a European technology, independent and results-oriented, to democratize genuinely useful AI. More productivity and less mental load for teams.

The mission of YBA, "it is in its name: 'Yes Boss Assistant', or 'Oui Chef !'. Our goal is to provide technology to employees to free up their time so they can focus on what matters. During a trip to Brazil, discovering a hotel symbolically named '700K Hours', I fully realized the importance of every moment. Back in the office at Salesforce, I realized that more than half of my daily routine had little meaning or value for the company." This is how the idea of YBA was born, an assistant designed to lighten the burden of thankless tasks and restore meaning to our work. The secret? Understand everyone's context.

Novembre 2025

Contacter

yba@yba.ai

06 50 00 11 46