Enterprise AI agents are moving from concept to practical use, but reliable training remains a major hurdle. Startup Arga is aiming to solve that challenge with a new approach to testing and simulation.
The company announced a $10 million seed round led by General Catalyst, with participation from Box Group, Emergence, Gradient, and SV Angel. The funding will support Arga's work on training environments designed for business software workflows.
Arga creates digital twins of enterprise tools such as Salesforce, Workday, and email platforms. Instead of relying on simple API-based testing, the startup recreates the full software environment, including permissions and web hooks, so AI agents can practice in conditions that closely mirror real operations.
According to co-founder and CEO Philip Li, the platform is built to handle situations where agents must connect information across systems. For example, an AI assistant may need to recognize that two separate records refer to the same company, avoid duplicate outreach, and choose the right contact from multiple opportunities.
This kind of training is difficult in live enterprise systems because scenarios are hard to reset and repeat. Arga's model makes it possible to run the same workflow many times, adjust variables, and scale simulations across multiple applications at once.
Investors see repeatable sandbox environments as a key layer for the next generation of agentic software. As enterprise AI becomes more capable, tools like Arga could help turn complex business workflows into a more structured and trainable digital frontier. The future of work may increasingly depend on these realistic learning spaces.
