Arga Labs Raises $10M to Train Enterprise AI Agents

Training AI agents to work reliably within enterprise software remains a major challenge, and startup Arga Labs is building tools to make the process easier.

The company announced a $10 million seed funding round led by General Catalyst, with participation from Box Group, Emergence, Gradient, and SV Angel. Arga Labs is developing training environments that enable AI agents to practice complex workflows across business applications before deployment.

The startup creates digital twins of enterprise software platforms, including Salesforce, Workday, and email systems. Instead of testing agents in limited API environments, Arga recreates entire software environments with permission systems and webhooks intact, allowing agents to learn from more realistic scenarios.

Building digital twins for AI agent training

Arga Labs CEO and co-founder Philip Li said enterprise workflows often involve multiple systems and ambiguous situations that AI agents struggle to handle. He described scenarios in which employees create sales leads on different platforms, and agents must determine whether records represent the same company, avoid duplicate outreach, and choose the correct contacts.

Traditional reinforcement learning methods require running the same scenario thousands of times and keeping successful strategies. However, enterprise software systems are difficult to reset or duplicate, making large-scale agent training challenging.

Arga’s approach is to create controlled digital replicas of these applications. Because the company manages the environment, it can reset scenarios, modify conditions, and run multiple training environments simultaneously.

The goal is to help AI agents learn how to operate across connected business tools, rather than mastering isolated tasks inside individual applications.

Creating infrastructure for enterprise AI systems

Arga Labs compares its approach to the development of AI coding tools, which have benefited from advanced systems for testing, deploying, and analysing code changes. Similar training infrastructure has not yet been widely available for business software applications.

The company believes more sophisticated testing environments could help AI agents become more capable at handling enterprise workflows. General Catalyst managing director Yuri Sagalov said repeatable sandbox environments are becoming increasingly important as businesses adopt agent-based systems.

“Having a repeatable sandbox environment is very important, and much more important with agents than it was with humans,” Sagalov said.

With its new funding, Arga Labs plans to continue developing its platform to help companies train and evaluate AI agents before deploying them in real-world business operations.