While AI agents are increasingly capable of finishing tasks, a new startup argues that current digital assistants lack true comprehension of the organizations they serve.
Zeroset has emerged from stealth mode and secured $5.2 million to develop models that continuously update AI agents with a real‑time view of a company’s operations.
According to Zeroset cofounder and CEO Akshat Kannan, contemporary AI models lack insight into the specific workings of any business, as relevant data is often dispersed among CRM platforms, ticketing systems, and internal documents, and they are rarely trained on a company’s proprietary data.
Kannan told Themoneytimes that pre‑built models have no grasp of enterprise dynamics and cannot understand how work gets done.
Zeroset’s Nebula product integrates with a company’s existing suite of tools — such as Microsoft 365, SharePoint, Outlook, Teams, GitHub, and other workplace messaging platforms — and monitors how data and activity evolve across those systems over time.
Rather than relying on a simple repository of documents, Nebula aims to give AI agents a deep understanding of workflow sequences and decision points, enabling them to recall how similar cases were handled and execute tasks autonomously.
Zeroset was launched in January by 18‑year‑old Kannan, who left Stanford this year, and 21‑year‑old William Zhang, who exited the University of Texas last year, to build the startup. The firm remains in a closed research preview, planning to gradually open access to customers instead of a single, broad release.
Zeroset intends to offer a hybrid licensing model with usage‑based pricing that scales according to the volume of data and agent activity flowing through Nebula, targeting enterprises and AI‑native firms that run long‑duration agents in sectors such as manufacturing, supply‑chain logistics, and financial research.
The startup indirectly competes with firms such as Mem0 and Zep, which are developing memory infrastructure for AI agents.
The $5.2 million pre‑seed round was led jointly by Gradient and 2048 Ventures, with Leblon Capital also participating. Gradient, a venture capital fund established by Google in 2017, concentrates on early‑stage AI and machine learning startups.
With a team of five, Zeroset plans to allocate its new capital toward hiring researchers and systems engineers, financing early enterprise deployments, and covering model training expenses.
Kannan noted that training represents a major expense, citing a recent run that cost roughly $100,000 over two weeks.
View the 11‑slide pitch deck that helped secure the funding.