Hone, founded by former OpenAI and Cognition staffers to create AI agents that can help run a business, raised a $60M seed at a $285M valuation
First reported by Bloomberg ·
Autonomous AI agents that can own business outcomes are now moving from research to enterprise application.
Hone, a startup founded by former employees of OpenAI and Cognition, has secured $60 million in seed funding at a valuation of $285 million. The company aims to develop AI agents capable of autonomously pursuing complex business objectives over extended periods, such as weeks or months. This funding round was co-led by Benchmark and Index Ventures. The core mission of Hone is to create AI systems that can take ownership of business outcomes, moving beyond simple task execution to achieve defined goals. The primary challenge for Hone lies in ensuring the reliability and trustworthiness of these autonomous AI agents to the extent that businesses will delegate significant responsibilities to them.
The substantial seed funding for Hone, with its ambitious goal of AI agents "owning outcomes," signals a significant market push toward more autonomous and goal-oriented AI applications. This move indicates a maturing landscape where AI is expected to not just assist but actively manage and achieve business objectives over longer time horizons. The success of such ventures will hinge on demonstrating tangible, reliable business results, potentially reshaping how companies structure their operations and delegate responsibilities in the future.
This development directly affects businesses looking to leverage AI for strategic advantage, potentially enabling them to automate complex decision-making and execution. For AI developers, it presents a new frontier focused on long-term reliability and accountability, moving beyond task-specific models. The market will be watching closely to see if Hone can overcome the inherent challenges of building trust in autonomous systems capable of handling significant business functions.
AI-written summary. May contain errors.