Arga Labs raises $10M seed to build training environments for enterprise AI agents

Making AI agents work in practice is harder than many companies expected, and a new crop of startups is looking for better ways to test and train those agents before deployment, particularly against the complexities of the modern enterprise. Arga Labs is one such company, announcing a $10 million seed round Wednesday led by General Catalyst, with participation from Box Group, Emergence, Gradient, and SV Angel.

Arga Labs builds training environments for enterprise software such as Salesforce, Workday, and email clients. Where most testing environments settle for a stateless API endpoint, Arga builds a full-scale digital twin of the program, cloning an entire enterprise application with its permission systems and web hooks intact, aiming for a more robust way to train agents across multiple systems. CEO and co-founder Phillip Li described the kind of ambiguity Arga’s tools are meant to resolve: a prospective client creating a lead in Salesforce while a colleague reaches out separately through HubSpot, and whether an agent can correctly identify the two as the same company, avoid emailing twice, and pick the right contact.

Normally, an agent could be trained for a task like this through reinforcement learning — running the scenario many times and letting only the successful strategies through — but the nature of enterprise software makes that scale of testing difficult, since there is no easy way to reset a system like Salesforce or Outlook to run the same scenario again, let alone clone it. Arga’s approach is to create a digital re-creation of that software, replicating its structure the way a crash-test dummy replicates a person; because Arga controls the environment completely, it can reset or modify it easily and run many environments at once, training agents on interactions between different programs. Part of the reason AI coding tools have advanced quickly, the thinking goes, is that sophisticated tools already exist for deploying, reversing, and analyzing code, making it easier to set up reinforcement-learning environments for coding tasks — tools that don’t yet exist for most business software.

General Catalyst managing director Yuri Sagalov, who also runs the firm’s seed program, said he sees a growing need for agentic testing tools like Arga’s. “I think that a lot of the economic value from agents is from using business applications,” Sagalov told TechCrunch. “Having a repeatable sandbox environment is very important, and much more important with agents than it was with humans.”

Why it matters

If reinforcement-learning environments did for AI coding tools what Arga is attempting for business software, the bottleneck holding back agents from reliably using enterprise systems like Salesforce and Workday may be a training-infrastructure problem rather than a model-capability one.