Salesforce AI Research announced on Mar. 26 the launch of AI Foundry, a new initiative aimed at advancing system-level artificial intelligence for enterprise use. The program brings together research teams, strategic customers, and academic partners to develop and validate new capabilities that move quickly from foundational research to product development.
This initiative addresses challenges unique to business environments, where the requirements for accuracy, consistency, and reliability differ from those in consumer applications. While recent advances in artificial intelligence have focused on large language models and consumer-facing tools like personal assistants, Salesforce said that the most significant obstacles for businesses now exist at the system level.
“The problems that matter most for businesses don’t live at the model level anymore,” said Silvio Savarese, Chief Scientist at Salesforce. “They live at the system level, where components work together to deliver accuracy, consistency, and reliability at scale. AI Foundry is the engine we’ve built to make that a reality.”
AI Foundry focuses on three main areas: simulation environments such as eVerse—which tests agents under realistic business conditions; ambient intelligence—embedding proactive and context-aware capabilities into workflows; and agent-to-agent ecosystems—developing protocols so autonomous agents can interact across organizational boundaries with security measures in place. The initiative also works closely with legal experts and ethical technology offices to define frameworks for agent negotiation.
Itai Asseo, Vice President of Salesforce AI Research said: “Many of the old rulebooks simply don’t apply anymore. AI Foundry connects foundational research to real business problems by collaborating closely with our strategic customers in rapid iteration cycles.” The project also expands Salesforce’s academic grant program by connecting external researchers with industry challenges.
With this effort, Salesforce aims not just to improve model benchmarks but also to build systems capable of meeting complex operational needs in enterprises.
