The rise of AI agents in large companies reveals a persistent paradox. On the one hand, prototypes are multiplying, exploratory uses are progressing, and on the other, the move to scale remains very fragile. If the teams have efficient models and accessible tools, the reliability of the agents remains the major obstacle to their operational adoption. To address this, Basalt offers a platform that structures AI engineering around a continuous loop of experimentation, evaluation and monitoring.
Thus, if AI agents quickly reach a first level of quality, companies lack methods to improve the final points necessary for scaling up. Teams must test prompt variations, compare models, analyze extreme situations and correct recurring errors. This approach is often carried out with improvised tools, which are difficult to industrialize. To improve this critical phase in the adoption of AI by the company, Basalt offers a centralized approach that makes it possible to document tests, measure progress and avoid regressions.
The platform stands out for its integration of business teams into the engineering cycle. Companies must define what a good result is, which requires direct involvement of experts in the relevant field. The quality of an agent is not only a question of algorithmic performance. It depends on the relevance of the answers, compliance with internal rules and the ability to manage exceptions. The Basalt platform facilitates this collaboration by providing a space where engineers, PMs and operational staff can validate scenarios, adjust prompts and document observed behaviors.
Production monitoring constitutes the other pillar of the approach. Companies rarely have detailed visibility into the actual behavior of their agents. Some errors only appear in specific contexts, which are difficult to anticipate in a test environment. Basalt helps detect these situations, isolate problematic responses and automatically populate new scenarios.
Basalt announces a fundraising of 5 million dollars, or 4.25 million euros, led by Entourage and Peak Capital, with the participation of Hexa, Alpha Star and Kima Ventures. The company was founded by Guillaume Marquis and François de Fitte, and is based in San Francisco, with a technical team located in Paris. The funding should make it possible to accelerate the development of the collaborative AI engineering platform and support recruitment in France and the United States.
