Hone raises $60M to build AI agents for business work that lasts months

Why it matters

Hone is raising seed capital to build agents that manage open-ended work over time. Customers will need to verify that the agents improve business metrics and operate within reliable safeguards. Hone has not disclosed revenue, customer outcomes, or independent performance data.

Hone raised a $60 million seed round to build AI agents that pursue business goals over weeks or months. CEO Moritz Stephan, a former chief of staff at Cognition, says the aim is to build AI that "owns outcomes." Benchmark and Index Ventures led the round, which values Hone at $285 million, according to the company, as reported by Bloomberg.

Hone's challenge is to make autonomous work reliable enough for companies to hand over real responsibilities. Its early customers include Cognition and AI inference startup Modal, Bloomberg reported. Hone is working with Cognition on agents for go-to-market work as the AI coding company grows. The report gives no revenue, customer results, or independent performance data for Hone's product.

A founder drawn to long-running work

Stephan's path to Hone includes a project with a longer clock than most software startups. As a teenager, he co-founded Team Tumbleweed, a European Space Agency business-incubator project that developed a wind-powered Mars rover prototype. The project grew to more than 50 people across four countries. business@school reported that a planned 2020 test in Israel's Negev desert was postponed because of COVID-19. A prototype was later tested there during the Austrian Space Forum's AMADEE-20 mission in 2021, according to the mission's Tumbleweed experiment records. Separately, Austria in Space reports that an improved prototype was tested in an Oman desert during the AMADEE18 mission. Stephan later studied computer science at Stanford and conducted research at its AI Lab under Chelsea Finn, applying meta-reinforcement-learning models to grade student coding submissions. The research was accepted at NeurIPS 2022, according to the Stanford AI Lab and the NeurIPS paper record. He then joined Cognition, where he was chief of staff to CEO Scott Wu.

That operating experience helps explain Hone's starting point. Coding agents such as Cognition's Devin can take on defined software work; Stephan wants to apply agent systems to responsibilities that cut across a business. Bloomberg's example is sales: Hone envisions an agent that does more than score incoming leads, instead handling the work of improving lead conversion over time.

Hone's own product description calls these persistent agents "Engines." Hone says its Engines can be assigned an outcome, connected to company systems, and given boundaries for what they can do. Its site describes a process that includes simulating an agent's decisions before it takes on more responsibility, with sensitive actions requiring human approval and actions logged for review. These are Hone's descriptions of its system, not independently evaluated results.

The long-horizon test

A business objective can stay open for months while the conditions around it change. That creates a different technical challenge from completing a discrete coding or customer-service task: an agent must keep track of context, adjust its approach, and know when a person needs to step in. Hone's sales example makes the promise tangible and raises a question: how can a customer judge whether the agent is improving the right metric without creating new problems elsewhere?

A June 2026 research paper offers a measure of that gap. CEO-Bench simulates operating a fictional startup for 500 days, including pricing, marketing, and budgeting decisions. The authors report that most evaluated models struggled, and every model finished below the benchmark's rule-based baseline. Because the test is a simulation, it does not evaluate Hone directly. Its results show that current models still struggle with long-horizon business decisions.

Benchmark partner Peter Fenton is joining Hone's board and told Bloomberg that the company is working on safeguards against cybersecurity incidents involving agents. Index partner Shardul Shah is also joining the board. Cognition CEO Scott Wu is a personal investor. Cognition is also among Hone's disclosed early customers, and Stephan worked there as chief of staff.

The $285 million valuation is attributed to Hone, and Bloomberg did not label it post-money. The report did not include the round's terms, Hone's revenue, or measurable customer outcomes. Without those details, readers cannot assess Hone's progress toward its broader promise: a system entrusted with work that usually belongs to a team. Customers will need to be able to see and verify the outcomes its agents claim to own.