Supported by

The way marketers build email keeps changing, and the direction is clear. More and more of the work is being done inside the AI chat box.

Litmus and Validity’s State of Email 2026 report shows how far this has already gone. Only 5% of teams are not using AI for email at all. Generative AI tools for copy and image generation are now the single most impactful AI use case in email marketing, ahead of personalization and performance analysis. And the report’s own read on where this is heading is blunt: smart teams are moving past using AI for content and toward AI-assisted email operations.

You can see it in the speed, too. In 2024, most teams took two weeks or more to send a single email. By 2025, most were down to a few days. Today, 76% of marketers produce and send an email within three days. Advanced AI adopters are deploying in under a day.

So the next step is obvious in hindsight. If the work keeps shifting into AI, why not build the email there too? That is exactly what is starting to happen. Production is moving into the chat box.

But there is a catch, and it is the thing most people miss when they get excited about this.

The chat box alone cannot build a real email

AI can write email copy. That part is solved, and it is good at it. Notice, though, what marketers are actually using AI for today: copy and images. The talking part of the email, not the building of it. That distinction is the whole story.

What AI cannot do, at least not yet, is build a real, production-ready email on its own. It could attempt to code one, but we have yet to see an AI generate email HTML from scratch that you would actually want to send. Email code is a strange, unforgiving craft, full of decades-old quirks that break the moment you look away.

And here is the deeper problem. Even if an AI could write perfect email code, it would still be guessing at everything that makes the email yours. How do your headings look? What are your brand fonts? How should a button render on mobile? What happens in dark mode? What is your spacing system, your color palette, your component library?

None of that lives inside a chat window. An AI working from a blank prompt has no way to know any of it. So you end up with something that looks plausible in the preview and falls apart in the inbox, off brand and broken in Outlook.

For an AI to build an email you would actually send, it needs a source of truth. It needs your email design system.

Your design system is the source of truth

This is the piece that makes everything else work.

Your design system is where your brand lives as real, reusable building blocks: your approved headers, buttons, image blocks, fonts, spacing, and colors, along with the rules for how they behave on mobile and in dark mode. It is not a style guide sitting in a PDF. It is the actual components an email gets built from.

With Email Love, your design system lives in Figma. The Email Love Figma plugin is the connection between Figma and your AI chat box. With it, your AI tool can build an on-brand, responsive email using pre-approved components that render correctly everywhere, on mobile, in dark mode, across every client that has ever mangled an email.

(Check out our free Ultimate Email Design System)

Give an AI access to that, and it stops guessing. It builds from components you already approved and renders them with an engine built specifically for email. That is the difference between a pretty draft and something you can ship.

How to build an email in Claude, start to finish

Here is how it comes together in practice. The connective tissue is MCP. Think of it as APIs for AI: it hooks an assistant like Claude up to your other tools so it can work with real data instead of whatever lives in the chat. With three connections, you can take an email from idea to ESP without leaving the conversation.

1. Start with real inspiration

The first connection is the Email Love inspiration MCP, which gives Claude access to our library of 400,000+ curated marketing emails. Ask it for strong DTC welcome emails and it pulls real-world examples to work from, then uses them to shape your design direction.

You are not asking the AI to invent a look from nothing. You are pointing it at real emails from real brands. You still bring the taste and the brand judgment, but now Claude has a much better sense of what you are going for, so you can iterate on your content and design far faster than starting from a blank page.

2. Build it on your design system

The second connection is the Email Love Figma Plugin MCP (Yes, we have two MCP connections), which links Claude to your design system in Figma. This is where the source of truth does its job.

With it, Claude assembles the email from your approved components and compiles it with the plugin’s engine into production-ready HTML. It is not trying to generate your email code on the fly. It is using your real components, rendering rules, and HTML, so the result is on-brand by default and built to survive the inbox: responsive, dark-mode-safe, and accessible.

By the end of this step, you have a finished, on-brand email inside of the Claude chat box. 

3. Push it to your ESP

The third connection is your ESP’s MCP. The last step used to mean rebuilding the email by hand inside your sending platform. Not anymore.

With your ESP connected, Claude pushes the finished HTML straight in as a template, ready to send as a one-off or drop into a flow. A growing number of sending platforms now offer an MCP, including Klaviyo, Customer.io, and Braze, so there is a good chance this works with the tool you already use.

That is the whole loop. You went from “I need a welcome email” to a built, on-brand, ready-to-send template, in one conversation.

Why this holds up as AI takes on more

It is easy to read this as a faster way to make an email, and it is one. But the more important point is about where things are heading.

As AI agents take on more of the production work, the tools that matter are the ones the agent actually needs. SaaStr’s Jason Lemkin put it well: software that is critical to an AI doing its job sees usage climb, and software the AI can route around quietly fades out. An AI does not need a clunky editor it has to click through. It does need a real source of truth to build on.

That is the role your design system plays, and it is the role the Email Love Figma Plugin is built for. It is the layer an AI reaches for when it builds your email. And because it is neutral by design, working with any assistant and any ESP, you are not betting on a single platform. You are giving whatever AI you use the email expertise it does not have on its own.

A lot of this is still speculation. I am sure people will be using drag-and-drop email builders in ten years, too. But I do think we are starting to see a new way that emails get built.

Have you tried building an email in Claude or another AI assistant? How was your experience? Do you think this could be the future of email production? Let me know: [email protected].

Much love,
Andy

Email: [email protected]
Twitter: @emaillove