Hone AI: What Engines Are and What to Know Before a Demo
Hone AI is a San Francisco startup building Engines, long-running AI agents that take responsibility for a business outcome rather than a single task. You give an Engine a goal and guardrails, and it learns your company’s systems and works on that goal for weeks or months.
The quick answer: Hone is an interesting bet for companies with a measurable, repeatable outcome they want automated, like faster lead response or lower vendor spend. It is very young, sells through demos, and has not disclosed revenue, customer outcomes or independent performance data.
Hone at a glance
Checked against Hone’s site and press coverage on November 2, 2026.
| Hone | |
|---|---|
| What it is | Persistent AI agents assigned to business outcomes |
| Founders | Moritz Stephan (CEO), Oliver Brady and Carlo Kobe |
| Funding | $60 million seed led by Benchmark and Index Ventures |
| Valuation | $285 million, according to the company |
| Where it works | Slack, Teams and email |
| Price | Not published. Hone sells through demos |
| Best for | Companies with a clear outcome, good historical data and appetite for early tools |
How Hone defines an Engine
Hone draws a simple line between three kinds of AI. A chatbot answers when you prompt it. An agent completes a task you assign. An Engine owns an outcome, and you set the goals and the guardrails.
An Engine goes through four stages, according to Hone:
- Role. You define success, set boundaries and decide what it should own.
- Onboarding. You connect your systems and introduce the people and processes involved.
- Training. It runs against past tickets, deals and decisions. You see what it would have done, correct it, and decide when it’s ready.
- Collaboration. It works in Slack, Teams and email, where anyone can ask, correct or hand off work.
Index Ventures adds some detail: an Engine explores a company’s systems, writes its own skills and tools, and acts on events instead of waiting for prompts. A whole team shares one Engine, so a correction from one person improves it for everyone.
What Engines can be assigned
Hone’s site gives five examples:
- Inbound conversion: enrich, qualify, route and follow up with leads.
- Vendor management: analyze spend and drive consolidation and renegotiation.
- Credit risk: investigate hidden exposure and drive interventions to cut losses.
- Product adoption: find friction and drive activation and retention.
- Customer commitments: track promises from pre-sales through delivery.
These are examples rather than a fixed catalog. Hone says it will build an Engine around the outcome you name.
Hone’s quoted users are Chris Prinz at Modal, Imen Awag at Lucanet and Manny Adelstein at Cognition. The Lucanet quote stands out because it names the less glamorous part of the product: “The game changer for us when using Hone was the governance layer.” Hone says sensitive actions can require approval and that every action and permission is logged.
The team and funding behind Hone
Hone announced its seed round in October 2026. The valuation comes from the company itself, through Bloomberg’s reporting.
CEO Moritz Stephan was an early engineer at Cognition, where he helped build the core of Devin and later served as chief of staff. Oliver Brady was an early engineer at Harvey before helping build Mercor’s enterprise effort. Carlo Kobe is a Thiel Fellow who co-founded Fizz with Stephan.
Stephan’s track record with long-running agents is the strongest argument for the company. A seed round this size also means investors are paying for that team and idea well before there is much public evidence.
The hard questions for an outcome-owning AI
The idea is appealing, and it raises problems that a demo won’t settle.
- Attribution. If conversion improves, was it the Engine, a new campaign or the season? Agree how you will measure this before you start.
- Goal-chasing. A system pushed to hit one number can damage what you didn’t measure. Lead follow-up that lifts conversion but annoys customers is a poor trade.
- Self-written code. Hone says Engines write their own routines in a versioned repository. That helps reuse, and it also means someone should own reviewing what they write.
- Past data. Training on old decisions works only if those decisions were good and well recorded. It can also teach an Engine your old habits.
- Accountability. An AI can own a metric, but a person still answers for the result. Name that person up front.
Early customers named in coverage are themselves AI and technical companies, so their experience may not transfer to a more traditional business.
How Hone compares with other agent platforms
Hone sits among AI agent platforms and the newer AI employee products. Most of these let you assign and automate tasks, and many can now run on events. Hone’s distinct claim is the outcome framing, the simulated testing against your history, and shared learning across a team.
For narrow, well-defined work, a simpler automation or a general AI agent can be cheaper and quicker to prove. Hone’s model makes more sense when the work crosses several teams and systems and needs judgment over a long period.
FAQ
What is Hone AI?
Hone is a San Francisco AI startup, also described as an applied research lab, that builds persistent agents called Engines. Each Engine is assigned a business outcome and learns from company data and feedback.
What’s the difference between an agent and an Engine?
In Hone’s terms, an agent completes a task you assign, while an Engine owns an outcome over time within goals and guardrails you set.
How much does Hone cost?
Hone does not publish pricing. You book a demo to discuss it.
Who are Hone’s customers?
Bloomberg reporting names Cognition and Modal as early customers. Hone’s site also quotes a user at Lucanet.
What is Hone’s website?
Hone’s site is hone.com. The hone.ai domain is unrelated and was listed for sale when we checked.
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
