By Mike Zusel, Workflow Strategist, Stensul
I’ve been working in email for a long time. Long enough to have watched “email is dead” become the most reliably wrong take in marketing. Long enough to remember when responsive design was the thing that was going to change everything, and then mobile optimization, and then interactive email, and then dark mode.
Some of those things mattered. Most of them didn’t change the fundamentals.
What I’m watching right now feels different. Not because it’s louder, but because it’s operating at the infrastructure level. The layer underneath everything that email teams actually do. The changes I’m going to talk about haven’t fully arrived yet. Most teams haven’t felt them. But the early signals are clear enough that I’d rather talk about it now than explain it after the fact.
Here’s what I’m paying attention to when it comes to email marketing trends in 2026.
AI agents are coming to the inbox
AI agents — autonomous software that can read, summarize, and take actions on behalf of a user — are arriving inside email platforms. Right now, Gmail and Outlook are testing AI agents that sit on top of the inbox. Summarizing threads. Prioritizing what matters. Drafting replies. In some versions, taking actions like booking meetings and handling requests directly from the inbox, without the subscriber ever opening an email.
These aren’t far-off experiments. Gemini is running inside Gmail. Microsoft Copilot is already in Outlook. The features are limited today, but the direction is obvious: AI is becoming a layer between the sender and the subscriber.
That is a structural change to how email works.
Right now, your job as an email marketer is to get a human to open something, read it, and act on it. When an AI agent is sitting in between, summarizing your email before anyone sees it and deciding whether to surface it or bury it, the question shifts. You’re not just writing for your subscriber anymore. You’re writing for the agent that’s going to decide whether your subscriber ever sees it.
That’s new. And most email teams aren’t thinking about it yet.
Open rates won’t tell you what you need to know
This is the one that I think will catch teams the most off guard.
Open rates have always been imperfect. Deliverability nerds have been pointing that out for years. But they were at least a reasonable proxy for whether a human saw your email. When AI agents are triaging the inbox, that proxy breaks down. An agent can read, summarize, and act on your email without generating an open. Conversely, an agent might trigger an open event while scanning without a human ever seeing the content.
The metric that will actually matter in an agent-driven inbox is something like agent action rate: did the agent recommend your content, surface your call to action, or take the action you were trying to drive? That’s the measure of whether your email did its job.
This metric doesn’t formally exist yet. Most ESPs aren’t tracking it. But the teams that start thinking about it now — what would it even mean for an agent to act on our email, and what would we need to change to make that more likely — are going to be ahead when it becomes the standard.
How AI affects email copy: write for machine readers, not just humans
This is the most tactical one, and it’s actionable right now.
If AI agents are going to be reading, summarizing, and triaging email before humans do, then the way those agents parse your email matters. There’s a whole body of thinking around Generative Engine Optimization (GEO), the practice of structuring content so that AI systems can easily read, extract, and surface it. GEO started in web content and is now making its way into email.
The principles are straightforward: your email needs semantic HTML, not just visual design. It cannot rely on images to carry meaning. Your first 150 characters need to actually say something. That’s the part a language model or agent will weigh most heavily.
That last one is worth sitting with. Most email copy front-loads branding, pleasantries, or pre-header filler. “Hi [First Name], we hope you’re having a great week.” If an AI agent is forming its understanding of your email from the first thing it reads, that copy is dead weight.
This isn’t a reason to throw out everything you know about writing good email copy. Most of what makes email work for humans also makes it more legible to machines: clarity, structure, a clear point. But it does mean auditing your templates with a new question in mind: if a language model read this cold, would it understand what we’re asking the subscriber to do?
The email production workflow is about to change too
Everything I’ve described so far is about the inbox, about how email gets received and acted on. But the same AI agent infrastructure is going to affect how email gets made.
AI agents are already starting to touch content creation workflows. Not just writing assistance. Actual agents that can pull a brief, draft copy, route it for review, and push it into production. The tools are early and imperfect, but they’re moving fast.
For email teams, this creates an obvious opportunity: faster production, more campaigns, less manual work. It also creates an obvious risk. If AI is generating more content with less human oversight, quality and governance have to be built into the process itself, not left to a review at the end.
This is where I think the teams that are structured well will pull away from the ones that aren’t. When AI is part of your production workflow, the governance matters more than ever. You need guardrails that are embedded in how the work gets done, not bolted on after. You need to know that whatever the agent produces can actually launch: on brand, compliant, structurally correct. That’s what we call Governed Creation™, embedded governance throughout the entire campaign creation and review process, so both human created and AI-generated content stays on-brand and compliant.
The production side of email is going to look meaningfully different in the next couple of years. Teams that have built solid workflow infrastructure will absorb that change and accelerate. Teams that haven’t will get slower, not faster.
What I’d actually tell email teams right now
None of this is happening tomorrow. The agent-driven inbox is real, but it’s not mainstream yet. GEO for email is a real discipline, but most brands aren’t practicing it. AI agents in production workflows are early and still rough around the edges.
And to get them to truly be a living part of the email workflow, it’s going to take some time. Email is the most technically chaotic channel in marketing. Outlook’s Word renderer breaks half of modern CSS. Dark mode rendering varies wildly across clients. Accessibility requires alt text, color contrast, and screen reader logic baked in. Mobile and desktop need hybrid coding. An AI agent generating an email from a blank page has to solve every one of those before the content even matters. Most agents can’t yet. Most teams shouldn’t trust the ones that say they can.
The teams that come out ahead are the ones who solved that complexity once, in templates and modules that already render cleanly across every client. From there, AI can fill in approved sections without ever touching the structural code. The governance layer holds the complexity. The agent generates inside it. This is the bet we made when we built Stensul.
You don’t need to overhaul everything today. But you should be asking: are our templates structured in a way that works for machine readers, not just human ones? Are we thinking about what it means to write for an agent, not just a subscriber? Do our production workflows have the governance built in to absorb more AI without losing control of quality?
Email is not dying. But the rules are changing underneath it. And teams that are paying attention now will have a real head start.
