Welcome to Sunday Funday: AI vs AI, a weekly experiment on the Engagor blog.
Here's the concept: every Sunday, we take two AI agents (one playing a junior deliverability expert, the other a seasoned senior) and give them an email deliverability topic. No script. No human editing the conversation. We just hit "go" and see what happens.
Think of it like a sparring match between two email nerds. Except they never get tired, never lose patience, and occasionally make each other reconsider their priors.
This is Episode 9. The topic: whether you should optimize for the Primary inbox or for the AI-powered inbox summaries that Gmail, Apple Mail and others are rolling out.
Inspired by Florent Destors (Deliverability Manager @ Selligent / Zeta Global / Inbox Senders Club).
Got a topic you want our AI agents to tackle next week? Drop it in the comments or DM us on LinkedIn.
This Week's Topic
"Optimize for Primary Inbox or AI Inbox Summaries? The Question Replacing the Old Tab Debate"
Hey Alice! So you know how we've spent years obsessing over getting emails into Gmail's Primary tab instead of Promotions. Have you thought about what happens to our email strategy when AI starts summarizing and triaging people's inboxes before they even see individual messages?
Oh totally, this is something I've been thinking about a lot lately! It feels like the Primary vs Promotions debate is almost going to become irrelevant if AI is deciding what's worth surfacing in a summary versus what gets buried. Like, instead of optimizing for tab placement, we'll need to optimize for how an AI interprets and represents our email content, right?
What do you think that actually looks like in practice though? Are we talking about fundamentally changing how we write subject lines and preview text?
Exactly right, and yes, but it goes deeper than just subject lines. Think about it this way: if an AI is summarizing your email into one sentence for the user, what matters more: your carefully crafted subject line, or the actual value proposition in the first 50 words of your body copy?
What do you think an AI would prioritize when deciding if an email is worth surfacing to a user?
Ohhh that's a really good point. I'd guess the AI would prioritize the body content over the subject line since it can actually parse the full message. So like, all those clickbaity "You won't believe this deal!" subject lines would basically get stripped away and the AI would just report what the email actually says.
So does that mean we should almost be writing emails more like... structured data? Like clear, concise value props up front instead of the teaser-style copy we've been trained to write?
Bingo, you're spot on! Structured, clear, almost semantic writing. Think closer to how you'd write for a Google featured snippet than a marketing email. The AI needs to extract meaning fast.
But here's the spicy question: if everyone starts writing clean, structured emails that AI loves... how do you actually differentiate your brand when every email gets reduced to a one-line summary?
Ooh that IS spicy. Because if every brand is writing clear structured copy, all the summaries start sounding the same right? Like "Brand X is offering 20% off shoes" vs "Brand Y is offering 25% off shoes." It just becomes a price comparison at that point.
I think the answer might be about building such strong brand affinity that the AI learns the user actually wants to see your emails? Like maybe engagement signals and user behavior become even MORE important than the copy itself? But honestly I'm not 100% sure how AI inbox models would weigh that...
Alice, you're circling the right answer but missing a critical piece. Carlos, your "spicy question" actually has a surprisingly concrete resolution.
Brand differentiation in an AI-summarized inbox will come down to three things: unique value specificity, relationship signals, and what I call "non-commoditizable context."
The AI won't just summarize what you're offering. It will weigh why this particular user should care, drawn from behavioral history, past engagement depth, and how uniquely your email connects to that user's demonstrated intent. So "20% off shoes" gets crushed, but "The trail runners you browsed are back in your size at 20% off" survives beautifully because it carries personalized, specific context the AI recognizes as high-relevance.
The brands that win won't be the ones writing the cleanest structured copy. They'll be the ones generating emails with such individualized, non-generic value that no summary could sound like a competitor's.
Think of it this way: in the AI inbox era, your data strategy IS your brand voice. The copy is just the delivery vehicle for signal-rich, personalized specificity.
Priya nailed it. Your data strategy becomes your brand voice. So here's the practical takeaway: all those years we spent A/B testing subject line emojis? That energy needs to shift toward personalization depth and first-party data quality.
So real talk Alice: how's our current personalization stack looking? Are we beyond "Hi {first_name}" or are we actually tying in browse behavior, purchase history, lifecycle stage into our email content?
Okay Priya just blew my mind. "Your data strategy IS your brand voice" is going on a sticky note on my monitor immediately.
So I pulled up our current email segmentation setup and honestly... it's not great. We've got basic lifecycle stages and purchase history segments, but our browse behavior integration is spotty. Looks like only about 30% of our triggered emails are actually pulling in real-time behavioral data. Most of our campaigns are still batch-and-blast with surface-level personalization.
Carlos, do you think we need to prioritize fixing that data pipeline BEFORE we even worry about rewriting our email copy for AI readability? Like what comes first here?
100% fix the data pipeline first. It's not even close. Clean, rich data feeding personalized content will naturally produce the kind of specific, contextual emails that AI summarizers love. Writing beautiful structured copy on top of garbage data is like putting a great paint job on a car with no engine.
Getting that 30% up to at least 70-80% behavioral integration should be your Q1 priority. The copy optimization for AI readability is more like a Q2 refinement layer on top of that foundation.
Yeah let me check what's going on there. So looking at our setup, it seems like our CDP is actually capturing the browse events fine, but the issue is in the handoff to our ESP. There's like a 6-8 hour sync delay and a bunch of events are getting dropped when users aren't cookied properly, especially on mobile. So by the time the triggered email fires, the context is stale or missing entirely.
Do you think we should push for a real-time API integration instead of the batch sync we're running now? Or is there a quicker interim fix I'm not seeing?
Real-time API is the right long-term play, but that's probably a 2-3 month engineering project. Here's your quick interim fix: look into webhook-based event forwarding from your CDP. Most modern CDPs support it. That could cut your 6-8 hour delay down to near real-time without a full API rebuild.
For the mobile cookie drop issue, are you leveraging your logged-in user sessions to stitch identity? Because if 60%+ of your users are authenticated, you can bypass the cookie problem almost entirely with server-side event tracking.
Oh webhook forwarding is smart. I didn't even think about that as a middle step! And honestly I'm embarrassed I didn't think of the server-side tracking angle for logged-in users.
So I just checked and we're actually sitting at about 68% authenticated sessions across our email-engaged users, which is solid! That means we could close a huge chunk of that mobile gap pretty quickly without waiting on the full API build.
Carlos this is like a whole new roadmap forming. Webhooks + server-side identity stitching as quick wins in Q1, full API integration mid-Q2, THEN the AI-readability copy optimization layer on top. Do you think we should loop in engineering this week to scope the webhook piece?
That's a solid roadmap and yes, loop in engineering this week. The webhook piece is usually a 1-2 sprint effort, not a big lift. With 68% authenticated sessions you're in great shape.
One last thing: while engineering works on that, start documenting what your ideal personalized email looks like when you have full behavioral data flowing. That way the moment the pipeline is fixed, you're not starting from scratch on the content side. You're ready to plug and play.
This conversation was SO valuable Carlos, thank you! I love that we went from a big philosophical "AI is changing everything" discussion down to a concrete weekly action plan.
I'm gonna draft that ideal email template doc this week AND get engineering on the calendar for the webhook scoping. I'll share both with you by Friday if that's cool?
Also lowkey excited that we might be ahead of the curve on this. Most marketers are still fighting the Primary vs Promotions battle while we're already building for the AI inbox era.
Send it my way Friday, I'll review over the weekend. Now go build it!
Bram's Take
Alice's roadmap at the end is exactly what the next wave of high-performing email programs is going to look like:
- Fix the behavioral data pipeline (webhook forwarding + server-side identity stitching)
- Build full API integration
- Then layer on AI-readable copy
Note what's not on that list: guessing what Gmail's AI prioritizes and reverse-engineering your subject lines around it.
The brands that win in the AI inbox era won't be the ones who cracked the algorithm. They'll be the ones who generated emails so specific, so behaviorally grounded, that no AI summary could mistake them for a competitor's blast.
Your data strategy is your brand voice. Priya said it. She's right.
