Zeroset raises $5.2M to teach AI agents how companies actually work
What's the deal? ZerosetDealroom has a profile for this one. Try Dealroom → has emerged from stealth with a $5.2 million pre-seed round to build models that give AI agents a constantly updated picture of a company's operations. The round was co-led by Gradient and 2048 Ventures, with participation from Leblon CapitalDealroom has a profile for this one. Try Dealroom →.
The problem: Cofounder and chief executive officer Akshat Kannan said today's AI models have little knowledge of how an individual business runs, because that information is scattered across customer relationship management systems, ticketing tools, and internal docs. "Models off the shelf have no understanding of the dynamics of how an enterprise actually works," he told Business Insider. "They have no understanding of how the work gets done."
What's the endgame? Zeroset's first product, Nebula, connects to software a company already uses — including Microsoft 365Dealroom has a profile for this one. Try Dealroom →, SharePointDealroom has a profile for this one. Try Dealroom →, OutlookDealroom has a profile for this one. Try Dealroom →, TeamsDealroom has a profile for this one. Try Dealroom →, and GitHub — and tracks how information and activity across those systems change over time. The aim is to give agents the sequence of events and decisions behind a workflow, so they can see how similar cases were handled and then do the work autonomously.
Who's behind it? Zeroset was founded in January by Kannan, 18, who dropped out of Stanford this year, and William Zhang, 21, who left the University of Texas last year. The company has five employees and remains in a closed research preview, expanding access gradually rather than through a broad launch.
Where the money goes: Zeroset plans to hire researchers and systems engineers, fund early enterprise deployments, and pay for model training. Kannan said training is one of its highest costs; in one peak run, it spent about $100,000 in two weeks.
The business model: The startup will combine a license component with usage-based pricing that scales with the data and agent activity running through Nebula. Its initial focus is enterprises and AI-native companies deploying long-running agents in manufacturing, supply-chain operations, and financial research.
What could go wrong? Zeroset indirectly competes with Mem0 and ZepDealroom has a profile for this one. Try Dealroom →, both building memory infrastructure for AI agents. Winning enterprise deals from a five-person team still in research preview is a steep climb.
The signal: As companies push AI agents from one-off tasks to long-running work, the gap between generic models and how a specific business operates becomes the bottleneck. Zeroset's bet — backed by Gradient, a fund GoogleDealroom has a profile for this one. Try Dealroom → founded in 2017 to target early-stage AI startups — is that the next layer of value sits in context, not just capability.
Read more: Business Insider
Image credit: Generated with Gemini
