Hone Raised $60M to Let an AI Engine Own Inbound Conversion. What Goes on the Order Form Before It Touches Pipeline?

Hone wants to hand an AI the number. On Oct 8 it launched "Engines," AI it says is assigned an outcome and keeps working it for weeks. One of them is an Inbound Conversion Engine. Hone said on Oct 8 that it raised a $60 million seed round from Benchmark and Index. The company is valued at $285 million, according to the company, via Bloomberg on Oct 8.

For a CRO, that is a direct bet on more pipeline from the same team. It also raises a new question. If software owns your conversion rate, who checks its work?

What Hone says it sells

Hone's homepage sorts AI at work into three categories: Chatbots "answer questions," Agents "complete tasks," and Engines "own outcomes," where "you set the goals and guardrails."

The rest of this section is also company-stated, from Hone's launch post and homepage.

• Engines own outcomes. You set the goal and the guardrails. The Engine onboards itself, learns your systems and history, and keeps going. Hone: "Our ultimate eval is the P&L."
• Inbound Conversion Engine. It "enriches, qualifies, routes, and follows up with inbound leads." Hone lists the outcome as "Faster lead response. Higher conversion."
• Practice on history first. "Before it touches anything live, an Engine runs against your past work: old tickets, old deals, old decisions." Changes "are tested against past cases in simulation before they go live."
• Controls. Custom policies "so that sensitive actions are always approved first." Every action and permission change is logged.

Index Ventures, an investor, says an Engine can take "a role like procurement coordinator or outbound rep" and is "tested against prior decisions before going live," according to its Oct 8 investment post.

The Proof Gap

The vendor names the P&L as its test. So hold it to that test, in writing, before it touches a live lead.

The Order Form

Write these before go-live:

1. Outcome definition. The exact conversion metric, the baseline period, and the guardrail metrics that cap it: lead quality, meeting-held rate, complaints and unsubscribes.
2. Backtest on our leads. Results from the simulation on our past inbound, shared before go-live.
3. Approval policy. A named list of actions that need human approval first.
4. Action log. Export rights to every Engine action and permission change, kept after termination.
5. Holdout. A share of inbound the Engine does not touch, so lift is measured against something.
6. Kill criteria and exit. Who can pause it, what stops it on its own, and which routines and history we keep when we leave.
7. Price basis. Hone publishes no price. Get the billing unit in writing. If the bill ties to an outcome, the outcome definition in line 1 is the billing definition.

For operators: run the replay test this week

Pull last quarter's inbound leads with their final status. Ask Hone, or any vendor in this lane, to run its backtest on that file. Then compare three numbers against what your team did: time to first response, meetings held, and cost per held meeting. Log every case where the Engine would have routed or followed up differently. Those cases are your approval-policy list. Track reversals the same way you log them in the Reversal Ledger.

The Eight-Seat Read

• Sales: Decide which follow-ups reps keep and which the Engine sends.
• Marketing: Inbound conversion sits on your handoff. Watch meeting quality beside meeting count.
• RevOps: You own routing rules, permissions, the action log and the backtest gate.
• Enablement: Train reps on what the Engine hands them and how to correct it.
• Customer Success: Inbound promises become your commitments. Read the follow-up copy.
• Partnerships: Partner-sourced leads need their own routing rule.
• CRO, CEO and founder: Fund a pilot with a holdout and a stop rule. A board line needs lift against a holdout.
• Revenue Finance: Price per held meeting, with the holdout as the baseline.

The other side

Gartner forecast on June 25, 2025 that over 40% of agentic AI projects will be canceled by the end of 2027, "due to escalating costs, unclear business value or inadequate risk controls." Anushree Verma, Senior Director Analyst at Gartner, said current models "don't have the maturity and agency to autonomously achieve complex business goals or follow nuanced instructions over time." Gartner also warned about "agent washing," the relabeling of older automation as agents.

Hone bets that an AI can hold a business goal over months. Gartner doubts that claim, and Hone has published no customer result yet. Gartner also forecast on Sep 29, 2026 that 70% of enterprises will abandon agentic AI built by vendor forward-deployed engineering (FDE) by 2028. Hone's homepage offers to build custom Engines with customers: "Tell us what you want an Engine to own, and we'll build it with you." Our reading: that forecast bears on any custom Engine, so the exit and ownership lines above matter more.

Why we might be wrong

Hone may publish strong customer results within a quarter. The backtest on past work is a real control, and many agent vendors skip it. A team that already runs clean routing and fast follow-up may see little lift; a team with slow response times may see a lot. Gartner's 2025 forecast covers agentic AI broadly and may say little about one well-funded team.

Sources

1. Hone, "Introducing Engines: AI with Motion," Oct 8, 2026: hone.com/blog/introducing-engines
2. Hone homepage, read Oct 11, 2026: hone.com
3. Index Ventures (Shardul Shah, Mark Xu), "AI That Owns Business Outcomes: Our Investment in Hone," Oct 8, 2026: indexventures.com
4. Bloomberg (via Bloomberg Law), "Former OpenAI, Cognition Staffers Want AI to Help Run a Business," Oct 8, 2026: news.bloomberglaw.com
5. Gartner press release, Sep 29, 2026: gartner.com
6. Gartner press release, Jun 25, 2025: gartner.com

Get the weekly read on where AI is moving revenue. Subscribe to the Report.