NeoCognition raises $40M
- NeoCognition raised $40 million to build agentic AI systems that learn domain expertise like humans. - The seed round positions the firm to develop agents aimed at bounded, expert workflows rather than general chat. - That funding reflects a market language shift toward agent promises that newsrooms will judge by operational legibility (techcrunch.com).
NeoCognition has raised a $40 million seed round and emerged from stealth to build artificial intelligence agents that learn on the job instead of only answering prompts. (techcrunch.com) The company announced the financing on April 21, 2026. The round was co-led by Cambium Capital and Walden Catalyst Ventures, with Vista Equity Partners and other investors participating. (prnewswire.com) NeoCognition is led by Yu Su, an Ohio State University professor and Sloan Research Fellow, and includes co-founders Xiang Deng and Yu Gu, according to the company’s launch materials and coverage of the deal. (prnewswire.com) (aijourn.com) An AI agent is software that takes actions across tools and steps, not just a chatbot that returns text. NeoCognition says current agents are still unreliable in real work and says its systems are meant to learn a workplace’s rules, constraints, and routines over time. (techcrunch.com) (waldencatalyst.com) The company’s pitch is narrower than “build a general assistant for everything.” It says the target is bounded, expert workflows where an agent can build what Su called a model of a “micro world” and improve through repeated use. (techcrunch.com) That framing matches a wider shift in the artificial intelligence market in 2026. Investors and founders are talking less about all-purpose copilots and more about systems that can handle specific business processes with measurable accuracy. (techcrunch.com) (waldencatalyst.com) NeoCognition is also selling its academic pedigree. The company says Su’s lab worked on agent-related research including Mind2Web, MMMU, and SeeAct, and says that work has been used in frontier-model development at OpenAI, Anthropic, and Google. (prnewswire.com) The backer list reinforces that message. NeoCognition and investors said angel investors and advisors include Intel chief executive Lip-Bu Tan, Databricks co-founder Ion Stoica, and researchers Dawn Song, Ruslan Salakhutdinov, and Luke Zettlemoyer. (prnewswire.com) Investors are making a bet that “agentic” claims can turn into software that actually completes work. NeoCognition now has $40 million to prove that specialized agents can be more reliable, cheaper to run, and easier to trust than today’s general-purpose assistants. (prnewswire.com) (techcrunch.com)
