The age of reason, not rules: is this marketing automation’s second coming?

For twenty years, marketing automation has run on rules. Now Marketo co-founder Jon Miller is betting that reasoning AI can deliver what the category always promised but never quite achieved. Is this an idea that technology has finally caught up with?

It’s a long time since I’ve thought there was anything interesting to say about marketing automation, but that changed a couple of weeks ago, with the launch of Phave by Jon Miller (who interviewed a while back for the Trust & Influence in B2B podcast). My content schedule has just been too overloaded to post about it… until now.

Let me tell you a story: Once upon a time, marketing automation was the hottest thing in B2B. Arguably it was the first major tech category that was designed specifically for B2B (tell me if you disagree!) recognising the need to systemise, processize and formalise marketing to make it more accountable to revenue and joined up with sales (that was the theory!). As its lead protagonists, Eloqua and Marketo (which Jon co-founded) not only transformed marketing, they also blazed a trail for better B2B marketing themselves, leading by example with creative, human and often irreverent marketing. They were a multi-dimensional breath of fresh air.

But both were acquired a long time ago, Eloqua almost immediately subsumed in the sea of Oracle and Marketo has struggled on within Adobe, with varying levels of commitment and functional development. Meanwhile, lower cost email marketing systems incorporated some of the once groundbreaking functionality, and this once pioneering category sank into the mainstream. To one degree or other, everyone had marketing automation… whether they used it properly is a completely different question. And whether the focus on the individual buy was ever right, is another… but that’s another argument for another day.

Marketing automation: From pioneer to commodity

The category has been stagnating, with little new or interesting to say about it. According to those who know, Adobe and Salesforce seemed to be planning the slow dissolution of their respective marketing automation legacy platforms into something a more buyer group or account centric (although not without some resistance from die-hard users). And in the age of AI, with buyers completely opting out of conventional journeys altogether, the end was surely nigh for marketing automation. Wasn’t it?

One person didn’t agree. Jon Miller teased his interest in plotting the next generation of B2B martech in an interview for the Trust & Influence in B2B podcast that we did in January this year. Since leaving Marketo, he had joined Demandbase via its acquisition of his next startup Engagio, and consequently played a major part in the development of that company, and consequently all of our thinking about accounts and targeting.

But he left Demandbase in 2023, and as he said to me in January has been in ‘stealth mode’ since then, building his latest platform, called Phave, which launched at the beginning of September. So what’s different about Phave? How does it reinvent marketing automation for the AI era? And why might we need it?

B2B moved on. The marketing automation platforms didn’t.

Jon’s diagnosis is blunt. “I started Marketo 20 years ago. Now almost every company has one of these platforms. But the reality is B2B moved on and the platforms did not.” Anonymous research, post-sale marketing, buying groups and now AI have all passed them by.

Having spoken to more than 200 marketers while building Phave, he hears the same thing everywhere: nobody loves their Marketo any more, they tolerate it. He calls it marketing automation Stockholm syndrome. People haven’t left because “there hasn’t been anything worth migrating to.” So rather than build the classic startup wedge, an add-on you buy alongside Marketo, Jon spent two years in stealth building a full replacement for large, complex enterprises.

Why rules fail

Marketo, Eloqua and HubSpot are, at their core, if-this-then-that systems. As Jon puts it, “Rules are good at what must be true,” like not emailing someone without consent. “They’re not particularly good at handling ambiguity or judgement.”

B2B buying is full of ambiguity. Rules can’t cope with it, so most teams collapse their programmes into a handful of nurture tracks. And rules can each work perfectly yet fail collectively: a nurture email, an event invitation, a product announcement and a sales sequence, all in three days. Every rule fired as designed. Nobody was deciding on behalf of the person.

Traditional machine learning couldn’t fix this, Jon argues, because B2B data is too sparse and complex. Reasoning AI, particularly the models of the last 12 months, can take everything known about a person and plan the right decisions for them.df a fixed nurture track, each touch is drawn from approved campaigns and content, then ordered and timed for that individual. “Cleared” means it has passed the platform’s frequency checks. (Image: Phave)

I think that distinction matters. Generating fresh content for every buyer at scale is how we’ve ended up with the flood of generic AI outreach eroding trust in B2B. Recombining what your team has made and approved is a more disciplined use of the technology.

Leads are too narrow. Accounts are too broad.

Jon’s Engagio years show in the second big design decision: accounts and buying groups are first-class objects, each with its own score and journey. “Leads are too narrow, accounts are too broad. Buying groups are just right.” Yet the buying group has almost never existed as a genuine object in any platform, and the workarounds mean manually attaching contacts to personas.

Reasoning AI removes that step by judging, from someone’s title and engagement, how likely they are to fill a given role in a buying group. It also unlocks post-sale marketing, because one account can contain a buying group that is already a customer and another that is still a prospect.

Power without the complexity

Marketing automation has always forced a choice between Marketo’s power and HubSpot’s ease. Jon’s argument is that AI ends that trade-off. Anything you can do in Phave’s interface you can do through its agent, in plain language, inside Phave or from the AI assistant your team already uses.

The cleverer part is governance. The marketing ops team writes down how the organisation works, from exact UTM conventions to double opt-in for contacts in Germany, and the agent follows it. The expert stops being the one “touching the buttons” and becomes the one who sets the rules. Phave’s documentation adds that every AI connection acts as a named person with that person’s permissions, and only a signed-in human can approve a send. As AI acts on our behalf in front of customers, that kind of accountability will become a buying criterion.

The honesty test

What struck me most was how readily Jon volunteered the limits of his own case. To answer every buyer’s first question, what will I lose, Phave built a “super RFP” of 481 requirements from real RFPs and Forrester’s evaluation criteria, then had Claude score Phave and its rivals. Phave came top at 2.97, with HubSpot second at 2.43. But Jon flagged that the AI could read Phave’s source code but only competitors’ public documentation, and that a bonus score for differentiated features favoured Phave. Without it, Phave still scores 2.75. “It’s my analysis, take it with a grain of salt.”

Likewise on simplifying bloated stacks: “I think that’s true. I can’t prove it yet.” None of his development partners has removed adjacent tools so far. And at $36,000 a year to start, it’s aimed at companies of roughly 200 people and up, not one-person marketing teams.

There’s an independent counterpoint too. Scott Brinker and Frans Riemersma’s Martech for 2026 report sees a hybrid future, with AI augmenting rules-based systems rather than replacing them. Phave arguably agrees more than its framing suggests, keeping rules for what must be true. The real question is where that line sits.

An idea the technology has caught up with?

Marketing automation promised to make marketing accountable to revenue, joined up with sales and relevant to the individual buyer. It delivered the first two, at least in part, but never the third, because rules can’t scale judgement.

Which raises a tempting thought: perhaps the failings that pushed many of us away from the demand generation model (blunt nurture tracks, volume-driven scoring, a fixation on leads in a world of buying committees) lay in the tools rather than the idea. If so, this is an idea that technology has finally caught up with.

I wouldn’t let the technology take all the blame, though. The MQL was broken as much by pipeline pressure as by rules, and no platform fixes incentives. And with buyers researching privately, increasingly through AI, even a perfectly sequenced playlist still needs someone there to receive it.

It’s too early for a verdict, and as a commentator rather than a tech analyst, I’m not equipped to give a platform like this the in-depth evaluation it deserves. But someone is finally asking the right question about this category: not how to add AI to marketing automation, but what marketing automation should be now that AI exists.

You can listen to my podcast interview with Jon Miller (conducted whilst he was still in ‘stealth mode’) via the link below.