Making AI agents effective is proving more difficult than anticipated for many organizations. Fortunately, a new wave of startups is addressing this challenge by developing superior methods to test and train agents prior to deployment, especially within the complexities of the modern business landscape.
Arga is at the forefront of this movement. The startup announced on Wednesday that it has secured $10 million in a seed funding round led by General Catalyst, with additional support from Box Group, Emergence, Gradient, and SV Angel.
Arga specializes in creating training environments for enterprise software such as Salesforce, Workday, and email clients. While many testing environments rely on stateless API endpoints, Arga builds full-scale digital twins of these programs. This process effectively clones the entire software, keeping permission systems and webhooks intact, which results in a more robust method for training agents across multiple systems.
Philip Li, the company's CEO and co-founder, illustrated the product's utility with a specific scenario: a prospective client creates a lead in Salesforce while a colleague reaches out separately through HubSpot.
"Can the agent correctly identify that these two are the same company?" Li asked. "Are they able to check whether or not they've only sent the email once? Are they able to identify who to send the email to out of the two opportunities?"
Agentic systems frequently struggle with such ambiguities, which Li views as a critical area where Arga's tools can facilitate significant improvement.
Typically, an AI agent could be trained for a task like this using reinforcement learning (RL), which involves running the scenario tens of thousands of times to allow only successful strategies to persist. However, the nature of enterprise software makes achieving this scale of testing nearly impossible. There is no easy method to "reset" a system like Salesforce or Outlook when attempting to rerun a scenario, let alone clone it.
Arga’s solution addresses this by digitally recreating the software, functioning much like a crash test dummy standing in for a person. Because Arga maintains complete control over the environment, the resulting recreation is simple to reset or modify. The company can also operate multiple environments simultaneously to train agents on the complex interactions between different programs. The goal is to replicate a person's full work environment, ensuring specific tasks overlap across various programs and knowledge systems.
This approach can be viewed as a method to bridge the reinforcement gap between coding and business applications. One reason AI coding tools have advanced rapidly is that sophisticated tools exist for deploying, reversing, and analyzing new code, making it easier to set up RL environments for coding. These tools do not yet exist for most business software, but their development is imminent. Once available, AI systems are likely to become much more proficient at utilizing these programs, potentially revolutionizing other industries just as they have transformed coding.
Yuri Sagalov, a managing director at General Catalyst who oversees the firm's seed investing program, highlighted the increasing necessity for agentic testing tools like Arga.
"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."
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.