AI Startup Zeroset Emerges From Stealth With $5.2 Million to Improve Enterprise AI Agents
Zeroset Raises $5.2 Million
AI startup Zeroset has emerged from stealth with $5.2 million in pre-seed funding as it works to solve a growing problem for enterprise AI agents: understanding how individual companies actually operate.
The round was co-led by Gradient and 2048 Ventures, with participation from Leblon Capital. Zeroset plans to use the funding to hire researchers and systems engineers, support early enterprise deployments and fund model training.
Founded in January 2026 by Akshat Kannan and William Zhang, the company currently has five employees. Its first product, Nebula, remains in a closed research preview as Zeroset gradually expands access.
Nebula Maps Enterprise Workflows
Nebula connects with workplace systems including Microsoft 365, SharePoint, Outlook, Teams and GitHub. Instead of simply retrieving documents, the platform tracks how information, activities and decisions change across those systems over time.
Zeroset's goal is to give AI agents a continuously updated picture of a company's operational state. This can include the sequence of actions, handoffs and decisions behind a workflow, allowing agents to use previous cases as context when carrying out similar tasks.
The company is targeting enterprises and AI-native organisations that use long-running agents, with initial applications including manufacturing, supply-chain operations and financial research.
Building a Context Layer for AI Agents
Zeroset argues that enterprise AI has a different context problem from consumer-facing AI. Internal workflows are often spread across CRM systems, ticketing platforms, documents, messaging applications and other tools, while important decisions can depend on processes that are not formally documented.
Nebula is designed to connect those traces into a representation of how work changes over time. Gradient describes the system as a way to recreate enterprise workflows and states so that agents can operate with more relevant context.
Zeroset says its approach puts it in the broader AI-agent infrastructure market alongside companies such as Mem0 and Zep, although its focus is specifically on enterprise workflow state and operational context.
Funding to Support Early Enterprise Deployments
The new funding will help Zeroset expand its technical team and move beyond its current research phase. The startup also expects to combine licensing with usage-based pricing tied to the amount of data and agent activity running through Nebula.
Model training is already one of the company's largest expenses. Kannan told Business Insider that one peak training run cost about $100,000 over two weeks.
Zeroset's next challenge will be proving that its workflow infrastructure can deliver reliable results inside real enterprise environments. With Nebula still in closed preview, broader customer adoption and long-term performance will determine whether the approach can become a core layer for enterprise AI agents.
