Zeroset has raised $5.2 million in pre-seed funding co-led by Gradient and 2048 Ventures, with participation from Leblon Capital and other investors, as the company develops infrastructure intended to help enterprise AI agents understand how work actually happens inside organizations.
The company is developing Nebula, a system designed to capture the state and traces generated as employees and AI agents work across enterprise systems.
Zeroset’s thesis is that increasingly capable frontier AI models still struggle with enterprise automation because they lack an evolving representation of the organizations they operate in.
While a general-purpose AI model may be able to reason or carry out individual tasks, enterprise processes depend heavily on company-specific decisions, dependencies, workflows, and institutional knowledge.
Zeroset is trying to give AI systems a representation of that operating environment.
The company compares the challenge with the evolution of robotics, where increasingly capable robots required not only better policies and more computing power but also representations of their surroundings and an understanding of how those environments changed in response to actions.
Zeroset wants to apply a similar concept to businesses by creating what it describes as enterprise “worlds.”
Nebula captures the operational state and traces created as work takes place and uses that information to build models of how an organization functions and changes over time.
Those models can incorporate decisions, dependencies and unwritten processes that are specific to an individual company.
Rather than requiring AI agents to reconstruct a workflow’s state from disconnected information each time they perform a task, Nebula is intended to provide a persistent representation of what has happened and how the organization’s state has evolved.
Zeroset also believes the company that generates the enterprise knowledge should control it.
Its approach lets organizations use frontier AI models while also training open models on their own operating knowledge, without changing the underlying enterprise intelligence system.
Nebula is currently available through a closed research preview, as Zeroset works with organizations to automate complex workflows and explore how AI can learn from the work their employees and agents already perform.
The company ultimately sees this infrastructure as a foundation for increasingly specialized enterprise intelligence.
Zeroset’s goal is to create a system in which AI agents can act with a deeper understanding of an individual organization, while models continuously learn from the company’s own operational activity.
The financing will support development of that vision as enterprise AI increasingly shifts from isolated assistants and individual tasks toward autonomous agents operating across longer and more complex business processes.
