Using cowork to personalize emails at scale: A 9x tutorial 

There's a gap most AI email workflows don't close. You ask Claude to write an email, it writes a good one, and then you're back where you started: copying text into your email tool, rebuilding the layout, checking that the footer is right, making sure the button style matches. The AI handled the writing. You're still handling everything else.

Jan Meinecke, co-founder at 9x, found a way to close that gap. His team built a Claude Cowork skill that takes a single prompt and creates a fully formatted, on-brand email directly inside Customer.io's Design Studio—ready to review, adjust, and send. The writing, the template, the brand styles, the footer—all of it, in a matter of seconds.

This post walks through how that workflow operates, why building inside the tool matters more than generating HTML, and what it means for teams where only a handful of people have historically known how to build a compliant branded email.

The email personalization problem most teams have isn't about writing quality. It's about the distance between a well-written draft and a sendable email.

Most AI workflows stop at the copy. You get good text, and then you still have to: open your email tool, find the right template, paste in the copy, check that the formatting survived, fix the spacing, make sure the footer is correct, add the right button, confirm the brand font is applied. That's 20–30 minutes of execution work on top of the writing time. And it requires knowing your tool well enough to do it correctly.

Jan's insight was that the bottleneck wasn't the writing—it was the last mile into the tool. If Claude can create the email inside Customer.io, formatted correctly and ready to review, the entire workflow collapses into a single prompt.

The workflow Jan built and demoed uses a custom Claude Cowork skill called "create marketing email." Here's what happens when you use it:

1. You write a prompt in Claude Cowork describing the email you want: who it's for, what it should do, any specific copy requirements. Something like: "Please draft an email we can send to students who just completed our AI Builder training program, congratulate them, and encourage them to download their certificate and add it to LinkedIn."
2. Claude picks up the skill and starts its task list. It recognizes this is an email creation task and connects to Customer.io using the integration.
3. Claude creates the email directly in Customer.io's Design Studio, applying the existing template structure—the font, the footer, the button style, the brand colors. It generates a template name, a subject line, and the full email body.
4. You receive a confirmation and a direct link to open the email in Customer.io. The email is there, formatted, complete, and ready for your review.

In Jan's demo, the total time from prompt to review-ready email was a matter of seconds. The example output: template name "AI Builder course completion – congratulations," subject line "You did it – your AI Builder certificate is ready," full email body formatted to 9x's exact template with the correct footer, font, and button.

This is the part of the workflow that's easy to underestimate. Claude is capable of writing HTML emails from scratch—but that's not what makes this approach useful. The reason creating inside Customer.io is better:

It's editable by anyone. A generated HTML file requires someone who understands HTML to adjust it. An email created in Customer.io's Design Studio can be edited by any team member—change a word, swap a button, adjust a headline—without touching code.

It respects your actual brand standards. Customer.io's Design Studio holds your templates, your global styles, your components. When Claude creates inside the tool, it applies those standards automatically rather than approximating them from scratch.

It's where you send from. There's no export, no import, no format conversion. The email exists in the tool you use to send—so it's already in the right place, with the right settings, connected to the right audience.

AI doesn't always get it right on the first attempt. Creating inside Customer.io means you can review and adjust directly. If Claude went in the wrong direction on the tone, you go back to Cowork and ask for changes. If a word choice is off, you edit it in Design Studio. The best-of-both-worlds setup: fast creation, full control.

The "create marketing email" skill is a custom Cowork skill that encodes three things:

1. The task logic: When someone asks to create a marketing email, connect to Customer.io, build the email in Design Studio using the existing template, and return a link to review it.
2. The brand context: Which template to use, what the brand guidelines are, what the standard footer and button configuration looks like. This is what ensures the output matches your existing standards.
3. The Customer.io connection: The skill uses Customer.io's MCP (Model Context Protocol) server to operate inside your workspace directly—creating content, applying styles, and returning the design studio link.

Setting this up requires Claude Cowork, a Customer.io account with MCP access enabled, and a skill file that encodes the above logic. The technical lift is modest. Once it's built, anyone on the team can use it with a plain-language prompt.

Jan's key observation from building this workflow: previously, only one or two people on his team used Customer.io. They were the bottleneck. Every time someone needed to send branded communication to students or customers, it flowed through whoever knew the tool.

With the "create marketing email" skill, that bottleneck disappears. Anyone on the team can write a prompt, get a fully formatted email back in seconds, make any final adjustments in Customer.io, and send it. The institutional knowledge that used to live in one person's head—what template to use, how to configure the footer, which button style to apply—is encoded in the skill.

For teams managing multiple customer segments, course completions, onboarding milestones, re-engagement campaigns, and ad hoc communications, this is a meaningful shift in speed. Communication that used to take an hour of back-and-forth and template work now takes a few minutes.

The same approach that works for course completion emails works for any recurring email type. Variations worth building:

Win-back sequence: "Create a three-part win-back sequence for customers who haven't logged in for 30 days. Use our standard retention template. Subject line options for each email."

Post-purchase follow-up: "Create a post-purchase email for customers who just completed their first order. Thank them, highlight what's next, and include a support resource link."

Onboarding milestone: "Create an email for users who just completed their profile setup. Congratulate them and point them to the next activation step."

Each of these follows the same pattern: a natural-language prompt, a skill that connects to Customer.io, and an output that's ready for review in Design Studio. The skill does the repetitive structural work. You handle the judgment calls.

What is Claude Cowork? Claude Cowork is an agentic AI tool that lets you connect Claude to external tools—like Customer.io—and give it the ability to take action inside those tools based on a prompt. It's designed for complex, multi-step workflows that go beyond single-turn conversations.

What's needed to connect Cowork to Customer.io? A Customer.io account, a Claude Cowork setup, and Customer.io's MCP server enabled. The MCP server gives Claude direct access to your Customer.io workspace, including your templates, segments, and Design Studio.

Does this work for teams without a developer? Yes. The skill setup requires a small amount of technical configuration, but once it's built, using it is entirely prompt-based. No code required for day-to-day use.

How is this different from using AI to write copy? Using AI to write copy gives you text. This workflow gives you a complete, formatted email inside your sending tool—with your template applied, your footer in place, and your button configured correctly. The output is execution-ready, not just draft-ready.

What if Claude creates something we don't like? You can go back to Cowork and ask for changes, or edit directly in Customer.io's Design Studio. The output is designed to be the starting point of the review process, not the final word.

Can we use this for SMS or push, not just email? The principle applies—Cowork skills can be built for other content types. Jan's workflow focuses on email because Customer.io's Design Studio is where the production work happens, but the same skill logic could be adapted for other message types within the platform.