Making Sense of Martech

The episode delivers solid, practical insights about AI's actual limitations in email development rather than hype. Megan provides specific constraints (liquid templating works, email code doesn't; dark mode handling, accessibility gaps in ESPs) and honest takes on where AI fails. However, much of the conversation revisits familiar frameworks (designer-developer tension, enterprise vs. SMB needs) and includes stretches of meta-discussion about the podcast itself and Megan's volunteer work that dilute density.

“I am not using it with email code at all. I find that you have to be so prescriptive with the prompts to use it for email code, that I might as well just code it myself”

“What I think people or developers are going to need to do is broaden their knowledge a bit more. Maybe start learning some JavaScript. Start learning systems thinking”

Megan's contrarian stance - "AI Was a Waste of Time, Until It Wasn't" - and her nuanced take that AI is useful for specific tasks (liquid, JavaScript debugging) but not a generalist tool is fresher than typical "AI will replace everyone" or "AI is useless" narratives. Her ghost tables vs. tables debate and the observation that developers need to become strategists rather than specialists is solid. However, the framing still echoes common martech wisdom about cross-functional skills and the importance of fundamentals.

“Well, for sure. It reminds me I was catching up with an industry friend who hired a designer developer that built out their modular email design and system for their entire company. I asked, like, how is it working like it is? Is it looking good? Is it working well? And he's like, yeah, it's great, but we don't use it. I was like, what do you mean?”

“You can still hire the professional photographer for the things that matter. So for the things that matter, email developers will still be needed without a doubt”

Megan is exactly the right guest for this conversation: a development lead at Inbox Army with deep, multi-company infrastructure experience (Gmail, Jen, Email on Acid), real hands-on expertise in email design systems, and demonstrated willingness to challenge prevailing narratives (she said AI was useless two years ago and reassessed). She's also actively building systems, writing, and contributing to industry standards conversations. She is a genuine practitioner, not a talking-head consultant.

“Megan Bush Howson is the development lead at Inbox Army and one of the most respected voices in email development today. A graphic designer by training, she made the pivot into code and she never looked back, overseeing email development at gmail, jet and email on acid before landing her current role.”

“I've been working on this really big email design system that's just getting more complex by the day”

Megan provides concrete examples: Liquid templating for dynamic products, Claude Code routines for calendar integration and project management, the ghost tables + divs approach to email development, Masal for component frameworks, and specific client failures (e.g., Hebrew rendering, custom fonts in Figma). She names tools (Gemini, Claude Code, Visual Studio Code, Figma, Mailchimp) and describes actual implementation details. However, the episode lacks hard metrics on productivity gains, specific client outcomes, or time savings.

“I needed some robust liquid to be written, and I like knew 90% of the way there what I needed to do, but I needed that extra 10% of help. So this was I started with Gemini before moving to Cloud Code, and I went back and forth with Gemini to get what I needed for it to work, because in this situation I had to dynamically display products, but the products were not being housed in a CRM”

“I've been using Claude code, just released routines, and they just change it so that you don't have to hook it up to a repository, a GitHub repository to use it. I actually started using that to kind of start turning Claude code into a sort of pseudo executive assistant to me.”

Jacqueline asks solid follow-ups ("What broke first, your patience or the Outlook rendering engine?", "Give me one thing that makes you say nope, doing it myself") and steers toward practical over abstract. She presses on real use cases and calls out when answers drift. However, she misses opportunities to push back - Megan's statement that AI is "getting better" and "learning from other people" isn't challenged; the political email tangent feels indulgent; and some questions feel softball ("What was your first martech tool?"). The host also spends air time on her own anecdotes rather than deepening Megan's insights.

“I mean, for me it's really it does come down to the prompting where I'm like, if I wanted to hit all the accessibility stuff properly, if I want it to code the way I want it to code, it's just you have to be so prescriptive with it that I'm like, why am I wasting the time being this prescriptive when I can do it myself?”

“I would say it's dangerously oversimplified because it's getting better, but unless you write that prompt correctly and you're extremely prescriptive, it's not going to give you what you need.”