Latest / Marketing Automation with Fexingo: HubSpot, Marketo, and Automated Customer Journeys / Why Your Apriori Email Sequences Still Outperform AI
Transcript
- Lucas: There's this 2026 Gartner survey that marketing automation vendors really don't want you to read. It says 68 percent of measurable ROI from automated campaigns still comes from old-fashioned, rules-based, trigger and action workflows — not from ai generated touches. Luna: Wait — you're saying the gpt powered 'smart' campaigns underperform a basic if this then that email drip? Lucas: That's exactly what I'm saying. And I know it sounds counterintuitive when every product launch touts AI. But the data is pretty clear. And look — if this kind of concrete, counterintuitive marketing insight is useful to you, the reason these shows stay ad-free and focused is listener support. You can help keep that going at buy me a coffee dot com slash fexingo. It's a small thing that makes a big difference for us. Luna: Yeah, I think that's worth calling out because we really do dig into the stuff that gets buried. And it's listener-funded, so it stays honest. Lucas: Exactly. So back to that survey — the number that jumped out at me was this: when Gartner asked marketers to attribute revenue to specific automation types, the top-performing category was what they called 'pre-scheduled nurture sequences.' Those are the drips you write ahead of time, set a cadence, and let run. Luna: Which is basically the opposite of adaptive AI that rewrites subject lines on the fly. Lucas: Right. And there's a real case behind this. I spoke with the head of demand gen at a company called CyberShield — they're a $12 million ARR B2B SaaS in endpoint security. They ran an A/B test for four months last year. One arm was a fixed 6-touch email sequence — perfectly boring, human-written, scheduled to the hour. The other was an ai powered adaptive sequence that changed timing, content, even sender name based on engagement signals. Luna: And the boring one won. Lucas: Forty-one percent higher meeting conversion. That's not a rounding error. That's the difference between hitting pipeline targets and missing them. Luna: So what was the AI doing wrong? Was it the content, the timing, the personalization gone creepy? Lucas: It was mostly timing, actually. The AI would see a prospect open an email at 2 PM on a Tuesday, so it would bump the next send to 2 PM the next day. But CyberShield's audience — IT security buyers — they're creatures of habit. They check email in the morning, before stand-up meetings. The fixed sequence sent at 7 AM local time and got opened consistently. The AI kept shifting sends to weird afternoon slots and engagement dropped. Luna: So the algorithm was optimizing for recency, not for rhythm. That's a classic overfitting problem. Lucas: Exactly. And this is where the psychology gets interesting. Predictability in email actually builds trust. When a prospect knows 'every Tuesday at 7 AM I get a case study from CyberShield,' that becomes a signal of reliability. The AI was optimizing for novelty — different send times, different subject line formulas — and it eroded the routine. Luna: I've seen this in consumer marketing too. There's a concept called 'apriori email' — where the sequence is designed before any engagement data exists. It assumes the buyer will follow a path, and it works because the path is logical, not because it's personalized. Lucas: That's the term I was trying to recall. Apriori sequences. They're built on the assumption that the buying process is somewhat predictable — problem awareness, evaluation, comparison, decision. You write one email per stage. And it still outperforms the adaptive stuff in many B2B contexts. Luna: But isn't that exactly what HubSpot's new Sequence Intelligence tool is going after? I read their release notes — they're adding AI to suggest send times and subject lines, but the core sequence is still human-designed. Lucas: That's the pivot. They realized that marketers want AI as an assistant, not as the architect. So Sequence Intelligence lets you set a manual 5-touch sequence, and the AI optimizes within your guardrails — 'don't send between 8 PM and 6 AM,' 'keep subject lines under 40 characters.' It's a compromise. Luna: It's almost like the market is rediscovering the discipline of the old-school marketing sequence. Remember when Marketo had 'Smart Campaigns' that were just glorified if-then logic trees? Lucas: And they worked. The problem is, in the last two years, every vendor raced to add 'AI' to every feature. And in doing so, they forgot that a human's understanding of context still matters. CyberShield's demand gen lead told me something that stuck: 'The AI didn't know that our buyers love the Tuesday morning case study because it makes them look smart in their team stand-up.' Luna: That's a context an algorithm can't infer from open rates. It requires understanding workplace rituals. Lucas: So I think the lesson for marketers right now is: don't hand the entire customer journey to the AI. Use it for testing variants, for analyzing which touch scored the meeting, for writing a subject line draft. But keep the structure, the flow, the logic of your sequence firmly in human hands. Luna: And maybe go back and audit your current active sequences. How many of them are fully automated with no human review? Lucas: That's the homework. Take your best-performing sequence from last year — the one you wrote yourself — and compare it against your latest ai optimized campaign. You might be surprised which one has the higher conversion rate. Luna: I'm definitely going to do that for our own nurture flow. I suspect the one I wrote in 2024 still beats the one the AI assembled last month. Lucas: And if it does, you're not alone. That Gartner survey — 68 percent — it's a reminder that the fundamentals still matter, even in an ai first world. Luna: So what about the other 32 percent? Where does ai driven automation actually win? Lucas: Great question. That 32 percent — it's mostly in lower-funnel, transactional scenarios. Think abandoned cart recovery, re-engagement for dormant users, or event-triggered follow-ups like 'you downloaded this white paper.' In those cases, the AI's ability to personalize at scale — like changing a discount code based on browsing history — genuinely lifts conversion. Luna: So the rule of thumb might be: for complex B2B nurture, stay apriori. For simple B2C triggers, let the AI run. Lucas: That's a solid framework. And I'd add one more thing: measure sequence fatigue. One of the hidden findings in that CyberShield test was that the AI sequence caused higher unsubscribe rates — 2.3 percent versus 1.1 percent for the fixed sequence. The constant variation felt like noise, not value. Luna: Because the fixed sequence became a trusted rhythm. The AI sequence felt like a stranger changing the conversation every time. Lucas: Exactly. So I'd encourage listeners to look at their own unsubscribe rates by sequence type. If your ai optimized campaigns have significantly higher opt-outs, that's a red flag that you're sacrificing relationship for recency. Luna: Great advice. And I think we've given people a concrete action item: audit one sequence this week, compare the old rules-based version with any ai enhanced variant, and see which one actually moves the needle. Lucas: And if you find that the old version wins, you're in good company. The tools are getting smarter, but the buyer hasn't changed that much.