Key Takeaways
- Babylist grew from 5 million to 1 billion emails annually while expanding from one CRM channel to eight.
- To keep pace, the CRM team adopted agentic AI workflows powered by Claude and Iterable’s MCP.
- One workflow reduced email production time from roughly one hour to about 15 minutes.
- Another automated campaign QA, identifying broken links, personalization issues, and outdated tracking parameters in seconds.
- Babylist is on track to reduce marketing email production time by 50% while creating more capacity for strategy and testing.
Most CRM teams face the same challenge.
Customer communications continue to grow across channels, but team size rarely grows at the same rate. As a result, marketers spend more time building, checking, and deploying campaigns, leaving less time for strategy, experimentation, and optimization.
That challenge was the focus of an Activate Summit session led by Natasha Flynn, Senior Manager of CRM at Babylist. During the session, Flynn shared how her team uses agentic AI, Claude, and Iterable’s Model Context Protocol (MCP) to automate production workflows and reclaim time for higher-value work.
The story is particularly relevant because of Babylist’s explosive growth, which posed Flynn’s team with a choice: continue scaling production manually or find a smarter way to work.
Their answer was agentic AI.
In this article, we’ll break down the workflows Babylist built, how Iterable MCP fits into the process, and the lessons other CRM teams can apply to their own operations.
The Scale Problem Babylist Had to Solve
Babylist didn’t start its AI journey looking for a faster way to write emails. When Flynn joined Babylist in 2022, CRM was primarily an email program. Since then, the company has expanded into multiple new channels while continuing to grow audience size and message volume.
A few numbers illustrate the challenge:
- Annual email volume grew from 5 million sends to 1 billion.
- Push notification volume increased 220% after launch.
- CRM expanded from one channel to eight.
As communication volume increased, manual production became increasingly difficult to sustain. Rather than adding more operational complexity, the team focused on a different question: which parts of the workflow truly require human expertise, and which parts could be automated?
That distinction became the foundation for Babylist’s AI strategy. The goal wasn’t to replace marketers. It was to eliminate repetitive production work so the team could spend more time on strategy, testing, and roadmap execution.
The first opportunity they identified was email creation itself.
Use Case #1: Automating Email Template Creation
The first workflow Babylist tackled was email production.
For many CRM teams, building an email involves a surprising amount of manual coordination. Copy lives in one system. Templates live in another. Campaign settings are configured separately. Even simple sends can require multiple handoffs before an email is ready to launch.
Babylist wanted to streamline that process.
The solution was a workflow that combines Notion, Claude, and Iterable’s Model Context Protocol (MCP) to automate much of the production work that previously happened inside the email platform.
1. Building an “Iterable Command Center”
Babylist created what Flynn described as an “Iterable Command Center”, a centralized library of email templates and campaign workflows. Instead of starting from scratch, team members select the appropriate template and generate a preconfigured copy document.
That document includes many of the settings that would normally require manual configuration, including:
- Template IDs
- Message types
- From names and email addresses
- Audience-specific navigation settings
- UTM guidance and examples
The result is a standardized workflow that allows contributors outside the CRM team to participate without needing deep knowledge of Iterable.
2. From Copy Document to Live Email
Once the copy document is complete, the process becomes remarkably simple. The completed file is uploaded into Claude with a straightforward instruction to build the email. Using Iterable’s MCP, Claude can access the necessary template assets and create the email automatically inside Iterable.
After the build is complete, Claude returns the template ID. The marketer pastes that ID into Iterable, and the email is ready for review.
What previously required roughly an hour of manual production work can now be completed in about 15 minutes. In some cases, the actual build process takes less than a minute.
3. Proving the Workflow Actually Saved Time
One detail from Flynn’s presentation stood out. The team didn’t assume automation was valuable simply because it was new. Instead, they challenged themselves to answer a practical question: was this genuinely saving time or simply introducing a different way to build emails?
After measuring the workflow, the answer became clear. The new process reduced production effort enough to justify continued investment, while still leaving room for future optimization.
Use Case #2: Automated Email QA Across Channels
Building emails faster only solves part of the problem. Every campaign still needs to be reviewed before it goes live.
For growing CRM programs, QA can become one of the most time-consuming parts of the production process. Teams must check links, personalization logic, tracking parameters, copy, formatting, and dozens of other details. As volume increases, so does the risk of human error.
Babylist saw an opportunity to automate that process as well.
Teaching Claude How to QA Like the CRM Team
Rather than relying on generic prompts, the team trained Claude using its existing QA standards.
The process started with an internal spreadsheet containing all the criteria the CRM team reviews before launching a campaign. That checklist became the foundation for a dedicated QA skill inside Claude.
Now the workflow is simple. A marketer copies the campaign ID from Iterable, pastes it into Claude, and asks the system to review the campaign. Claude then generates a detailed report covering everything from personalization logic to tracking parameters.
Going Beyond Basic Link Checking
One of the most interesting aspects of the workflow is that it evaluates context, not just functionality.
During the session, Flynn demonstrated a campaign containing multiple intentional errors. Claude identified broken personalization syntax, spelling mistakes, incorrect URLs, and outdated tracking parameters. It also recognized when a destination page didn’t match the content being promoted, even if the link itself technically worked.
Traditional QA tools often verify whether a link resolves correctly. Babylist’s workflow evaluates whether the campaign itself makes sense. The result is a faster review process and an additional layer of protection against mistakes that could otherwise reach customers.
| See How Babylist Scales Customer Engagement Across Channels. Read the Babylist customer story to learn how the company uses Iterable to support millions of growing families while scaling personalized engagement across channels. |
Two Things That Make AI Adoption Work
The workflows Babylist demonstrated are impressive, but Flynn emphasized that the technology itself is only part of the story.
The bigger lesson is that successful AI adoption depends on both preparation and people. Teams that rush implementation often become frustrated by poor results, while teams that invest in the right foundation tend to see much stronger outcomes over time.
1. Build the Right Foundation
One of Flynn’s primary recommendations was to slow down at the beginning.
Agentic AI workflows require thoughtful setup. Prompts need to be refined. Processes need to be documented. Teams need to clearly define what “good” looks like before asking AI to automate it.
At Babylist, that preparation showed up in multiple places:
- Standardized email templates
- Clearly documented copy requirements
- Defined QA criteria
- Structured workflows that AI could reliably follow
Without that foundation, the outputs would have been inconsistent.
The lesson for marketers is straightforward: investing time up front often creates significantly larger time savings later. AI can accelerate a process, but it cannot compensate for a process that was poorly defined to begin with.
2. The Emotional Side of AI Adoption
Flynn also addressed a challenge that many organizations overlook: AI adoption is not purely a technical change. It’s a human one.
Teams often experience a range of reactions when AI enters their workflows. Some people are excited. Others are skeptical. Some worry about how their roles may change. Others are simply overwhelmed by the pace of new technology.
According to Flynn, those reactions are normal. The key is acknowledging them rather than ignoring them. Teams that approach AI with curiosity and an open mind are more likely to discover practical applications that improve their work.
Successful AI adoption requires more than introducing new tools. It requires helping teams understand how those tools fit into their existing workflows and where they create value.
The Results (and What’s Next)
For Babylist, the impact extends beyond faster email creation. By reducing time spent on manual production work, the CRM team has been able to increase capacity across the organization.
Some of the biggest benefits include:
- A projected 50% reduction in marketing email production time by the end of H1
- Faster turnaround on product and business requests
- More time for experimentation, strategy, and roadmap initiatives
- Greater ability to support a rapidly growing multi-channel CRM program
One example Flynn shared highlights the broader impact. Requests that previously might have taken months to schedule and execute can now move from concept to implementation in just a few days.
At the same time, Babylist views these workflows as a starting point rather than a finished product.
MCP is still relatively new, and the team continues to identify opportunities for optimization. Even within the email creation workflow, there are steps that remain partially manual today and may become automated in the future.
That’s an important reminder for teams evaluating agentic AI: The goal isn’t to build a perfect workflow on day one. The goal is to identify high-friction processes, automate what you can, and improve from there.
Agentic AI Works Best When It Removes Friction
One of the most valuable insights from Flynn’s session is that AI adoption doesn’t have to start with a complete transformation. Babylist didn’t begin by trying to reinvent its entire CRM operation.
Instead, the team focused on a practical question: which tasks consume time without requiring deep human judgment?
As you evaluate your own workflows, consider Flynn’s three recommendations:
- Approach AI with an open mind.
- Invest time in building the right foundation.
- Expect the technology to keep evolving.
The most successful teams won’t be the ones using the most AI. They’ll be the teams using AI to eliminate friction and create more space for meaningful work.
Frequently Asked Questions (FAQs)
1. What is agentic AI in email marketing?
Agentic AI refers to AI systems that can perform tasks on a marketer’s behalf rather than simply generating content. In Babylist’s case, Claude and Iterable’s MCP were used to build emails, retrieve assets, and perform campaign QA with minimal manual intervention.
2. What is Iterable MCP?
Iterable’s Model Context Protocol (MCP) enables AI tools such as Claude to interact directly with Iterable. This allows marketers to automate workflows like template creation, campaign QA, and other operational tasks.
How much time did Babylist save using AI?
Babylist reduced email production from roughly one hour to about 15 minutes for many campaigns and is on track to reduce overall marketing email production time by 50%.
Can AI improve email QA?
Yes. Babylist built a QA workflow that reviews campaigns for broken links, personalization errors, misspellings, tracking issues, and other common problems. The system can also identify contextual issues that traditional QA tools might miss.
What’s the best way to start using AI in marketing workflows?
Start with repetitive, time-consuming tasks that follow a defined process. Build a strong foundation, document expectations clearly, and focus on solving a specific workflow challenge before expanding into more advanced use cases.
Want to See Babylist’s Broader Customer Engagement Strategy?
Babylist’s AI workflows are just one part of a much larger customer engagement program. Read the Babylist customer story to see how the company uses Iterable to support millions of growing families, scale personalized communications, and grow across channels.