A five-month-old artificial intelligence startup has emerged from stealth with $60 million in seed funding and a lofty ambition: building AI systems that don't just complete tasks but own entire business outcomes.
Hone, based in San Francisco, announced Tuesday that it closed the round led by Benchmark and Index Ventures, with backing from a roster of prominent investors including Elad Gil, Lux Capital, and SV Angel. The funding values the young company at $285 million, according to Hone via Bloomberg. Benchmark partner Peter Fenton and Index's Shardul Shah are joining the board.
The startup is part of a growing wave of companies attempting to move beyond today's AI chatbots and assistants. Rather than tools that respond to prompts or handle discrete assignments, Hone is building what it calls "Engines," persistent AI agents designed to manage complex, ongoing business functions like inbound sales conversion, vendor negotiations, or credit risk assessment. Think of them less as intelligent assistants and more as autonomous colleagues with specific portfolios.
"What we are doing is building AI that owns outcomes," CEO Moritz Stephan told Bloomberg.
That framing has caught the attention of two of Silicon Valley's most selective venture firms. Shah drew a direct comparison to Cognition, the buzzy AI coding startup valued at $2 billion that has captured significant mindshare in developer circles. "What Cognition is to software development, Hone can be to every other function in an organization," Shah said.
The pitch centers on a fundamental shift in how companies might deploy AI. Instead of giving an agent a single instruction, Hone's customers assign ongoing goals. The system then coordinates across tools and people, making decisions within carefully defined guardrails. Crucially, according to the company, these Engines rehearse changes in a sandbox environment before executing them in the real world.
In a research post published alongside the funding news, Hone claimed its replay-and-simulation approach reduced errors from 33 percent to 5 percent across 125 internal test scenarios. The company also promises standard enterprise security features: SOC 2 compliance, detailed audit logs, and assurances that customer data won't be used for model training. Deployment options include Hone's own cloud infrastructure or a customer's private virtual network.
Still, Fenton suggested the technology demands caution. Systems like Hone's need "appropriate immune systems," he told Bloomberg, an acknowledgment that autonomous agents with real business authority carry risks alongside their potential.
The founding team brings experience from several notable AI and enterprise software companies. Stephan previously worked at Cognition and conducted agent research at Stanford's AI Lab. Co-founder Oliver Brady was an early engineer at legal AI startup Harvey before leading enterprise initiatives at Mercor, according to Index Ventures. The third co-founder, Carlo Kobe, is a Thiel Fellow who previously co-founded Fizz. The company's roster includes alumni from Cognition, Mercor, and quantitative trading firm Jane Street.
Hone's customer list remains thin, as expected for such a young company. Its website features testimonials from Cognition, cloud platform Modal, and software firm Lucanet, though the details of those commercial relationships have not been independently confirmed. A job listing from earlier this fall described Hone as a pre-launch team of fewer than 15 people, targeting broader availability sometime in 2026—though timelines in the AI sector have a tendency to slip.
The funding underscores venture capital's continued appetite for foundational AI infrastructure, even as enthusiasm cools in other corners of the startup world. Whether Hone can deliver on its promise of truly autonomous business agents remains an open question, one that will likely take years to answer. For now, it has the backing and the runway to find out.
