NeoCognition raises $40M seed for human-like AI agents

The race to build more dependable and capable AI agents is drawing significant venture capital, with investors actively recruiting top academic talent to launch new companies. This trend is exemplified by the emergence of NeoCognition, a startup founded by Ohio State professor Yu Su, which has secured a substantial $40 million seed round. Su initially hesitated to commercialize his research but was persuaded by recent advances in foundational models that could enable truly personalized AI systems.

NeoCognition’s funding was co-led by Cambium Capital and Walden Catalyst Ventures. The round also included participation from Vista Equity Partners and notable angel investors such as Intel CEO Lip-Bu Tan and Databricks co-founder Ion Stoica. The company is positioning itself as a research lab focused on developing self-learning AI agents that can achieve far greater reliability than current offerings.

Su argues that today’s AI agents are fundamentally unreliable generalists. “Every time you ask them to do a task, you take a leap of faith,” he stated. He estimates that current agents from various providers successfully complete tasks as intended only about half the time. This inconsistency makes them unsuitable for deployment as trusted, independent workers. NeoCognition’s core mission is to solve this problem by creating an agent system that can autonomously learn to become an expert in any given domain, mirroring the human capacity for rapid specialization.

Human intelligence is broad, but its true power lies in our ability to focus and master specific fields. When entering a new profession, people quickly learn its unique rules, relationships, and consequences. “For humans, our continued learning process is essentially the process of building a world model for any profession, any environment,” Su explained. He believes that for AI agents to become true experts, they must autonomously learn to construct a model of any given “micro world.” This capability for autonomous specialization is what Su sees as the missing link for achieving reliable, independent AI.

While it is possible to engineer task-specific agents for narrow verticals, NeoCognition is pursuing a more flexible approach. The startup is building generalist agents capable of self-learning and specializing in any domain without requiring custom engineering for each new application. The primary target customers are enterprises, including established SaaS companies, which could use the technology to create AI agent-workers or enhance their existing software products.

The strategic value of having Vista Equity Partners as an investor is significant for this go-to-market strategy. As a major private equity firm in the software sector, Vista can provide NeoCognition with direct access to a large portfolio of companies actively seeking to integrate AI into their offerings. The startup currently employs a compact, highly specialized team of about 15 people, most of whom hold PhDs, reflecting its deep research roots.