Arga Labs: Building Better AI Training Environments for Enterprises

Making AI agents work in practice is proving far more challenging than many companies anticipated. However, solutions are emerging. Arga Labs is one such company, offering specialized training environments for enterprise software. The company recently secured a $10 million seed round led by General Catalyst, with participation from Box Group, Emergence, Gradient, and SV Angel. Arga Labs creates digital twins of enterprise software like Salesforce and Workday, enabling more robust agent training across multiple systems. CEO and co-founder Phillip Li uses an example where a potential client creates a lead in Salesforce while a colleague contacts them separately through HubSpot. "Can the agent correctly identify these as the same company?" Li asks. "Can it verify that only one email was sent? Can it identify the right person to email out of the two opportunities?" Agentic systems still struggle with this level of ambiguity, but Arga Labs' tools are positioned to address these challenges. Traditionally, reinforcement learning has been used to train agents for such tasks—running scenarios tens of thousands of times and filtering only successful strategies. However, the nature of enterprise software makes this scale of testing nearly impossible. Resetting systems like Salesforce or Outlook for repeated testing is impractical, and cloning them is even more difficult. Arga Labs' solution is to create a digital recreation of the software, replicating its structure like a crash-test dummy replicates a person. By having full control over the environment, Arga Labs can easily reset or modify it. The company can also run multiple environments simultaneously, training agents on the complex interactions between different programs. The goal is to replicate a person's full work environment, with overlapping tasks across different programs and knowledge systems. This approach aims to bridge the reinforcement learning gap between coding and other applications. AI coding tools have advanced rapidly due to existing sophisticated tools for deploying, reversing, and analyzing code, which makes it easier to set up reinforcement learning environments for coding. However, similar tools do not yet exist for most business software. Once they do, AI systems are expected to significantly improve their ability to use these programs, revolutionizing industries just as they have revolutionized coding. Yuri Sagalov, managing director at General Catalyst and head of its seed program, highlights the growing need for agentic testing tools like Arga Labs. "A lot of the economic value from agents comes from using business applications," he says. "A repeatable sandbox environment is very important, and much more important with agents than it was with humans." Arga Labs' solutions are poised to enhance AI agents' ability to efficiently utilize enterprise software, potentially revolutionizing how businesses operate across sectors.