Marketo's Co-Founder Wants AI to Pick the Next Marketing Touch

Jon Miller built Marketo around marketing automation rules. His new company, Phave, says the next version of the category will decide which rule-ready message each buyer should receive.

Phave came out of stealth on September 23. It reached general availability in August and starts at $36,000 a year. Miller founded it with Nick Bonfiglio, who previously ran product and engineering at Marketo.

The product’s core idea is Maestro, a system that creates a “Playlist” of touches for each person, account and buying group. Marketers still make and approve campaigns. The software then weighs what is already queued and puts the next touch into an order intended to reach the stated objective.

Miller’s argument is narrow and strong: “Rules are good at what is allowed, but bad at deciding what is best.” Consent, frequency caps and quiet hours should remain rules, he says. Choosing among several permissible emails is a decision model should make.

That change matters because most B2B teams don’t run one clean buyer journey. A single account may have several groups, at different stages, while one person receives messages from product marketing, demand generation and sales. Traditional nurture logic handles that by building exclusions, routing rules and a growing spreadsheet of ownership. Phave is betting a system can resolve the conflict at send time.

Servion Global Solutions is one of the named customers. Bruce Eidsvik, its chief growth officer, said the company has “more than doubled” marketing execution volume with Phave. That’s a meaningful operational claim, but it isn’t a pipeline claim. Phave hasn’t published the baseline, time period, cost or the effect on meetings, opportunities or revenue.

The company also scored itself against 481 requirements from 12 enterprise evaluations. It says Claude weighted the criteria and placed Phave at 2.97, above HubSpot at 2.46 and Marketo at 2.38. That’s a vendor-run comparison, not an independent benchmark. It describes product coverage, not whether a marketing organization made more money.

The feature to watch is governance. Phave says a team can run it through its own interface, an AI client, MCP or a REST API while keeping permissions and approvals consistent. That sounds mundane. It’s where the product either works or becomes another route around a marketing-ops team.

A 5–30 person B2B company doesn’t need an AI to invent a thousand new journeys. It needs a clear answer to a simpler question: when product, sales and lifecycle teams all want to contact the same account, who decides the next best touch and can anyone explain it later?

Phave gives that question a product form. The real test is whether customers can show a controlled lift in qualified pipeline or sales-cycle speed, with an audit trail that explains the decision. Until then, treat it as an interesting replacement for configuration work, not evidence that the marketing-automation category has been solved.

The operating model Phave has to prove

The launch announcement says Phave works at the person, account and buying-group level. That is the right unit of work for many B2B teams. One account can have a champion, a finance reviewer and an implementation team. A system that only scores a single lead can produce a clean dashboard while missing the buying committee.

The Phave product description argues that the platform replaces legacy MAP configuration with AI-native orchestration. That pitch will appeal to teams whose campaign calendar has become a queue behind one operations specialist. Phave’s homepage presents the same AI-native orchestration position. It will also create a new responsibility: someone needs to set the goal, inspect the logic and decide what a model is allowed to send.

The company’s $36,000 starting price frames the buyer. Phave is a platform-and-process replacement for teams routing signals between marketing, sales and customer success. The right evaluation has to include the time saved maintaining rules, the time added reviewing decisions, and the consequence when a model sends the wrong message to the wrong account.

That’s close to the question raised by HubSpot’s move to place CRM context inside more AI tools, and by the newer Amazon seller workflows. The interface can change. The governance problem stays.

Ask a vendor to replay a real account history. Give it an account with several contacts and competing campaign claims. Then compare its proposed sequence against the team’s existing rules. A useful system should show why it chose a touch, what it excluded and how a marketer can override it. That is the proof Phave needs before its thesis becomes more than a well-timed replacement pitch.