Arga Labs Raises $10M to Train Enterprise AI Agents

Arga Labs Raises $10M to Train Enterprise AI Agents

Arga Labs has raised $10 million in seed funding to build training environments that help AI agents better navigate complex enterprise software.

The round was led by General Catalyst, with participation from Box Group, Emergence, Gradient, and SV Angel.

Building Digital Twins of Enterprise Software

Arga Labs develops training environments for enterprise applications including Salesforce, Workday and email clients.

Instead of relying on basic, stateless API environments, the company creates full-scale digital twins of enterprise software, replicating elements such as permission systems and webhooks. This allows AI agents to train on more realistic workplace scenarios involving multiple applications and data sources.

Addressing the Enterprise AI Training Gap

Enterprise software presents challenges for reinforcement learning because systems such as Salesforce and Outlook are difficult to reset, clone and repeatedly test at scale.

Arga Labs’ environments can be reset, modified and run simultaneously, allowing companies to train AI agents across complex workflows and interactions between different business applications.

For example, an agent could be trained to determine whether leads created across Salesforce and HubSpot represent the same company, while also checking previous communications and identifying the correct recipient.

Creating More Capable AI Agents

Arga Labs believes repeatable sandbox environments could help close the gap between the rapid progress of AI coding agents and the slower development of agents designed to operate business software.

The company’s approach aims to give enterprise AI systems more realistic environments in which to test and improve their ability to complete complex tasks across multiple applications.