Jimmy Kim and a co-speaker explain why AI email segmentation tools from Klaviyo and Omnisend are problematic. These tools automatically target customers and time sends, but marketers can't see how the AI decides who gets included. When segments underperform, there's no way to diagnose the problem.
AI-driven segmentation. I'm talking about like predictive segmentation, AI-powered audience selection. I think it's still pretty dangerous because it's kind of like a black box. And I want to explain this one first. We'll jump into an ESP, right? And they all have predictive analytics. The big Klaviyo, the Omnisend, everyone's got different things like that. They have AI segmentation. And these are various power, like AI-powered audience tools that promise the same thing, that AI will figure out who to send it to and when. Have you seen that?
Yeah, I have. What's like your gut take? Do they work? It kind of works. Sometimes they're great.
Sometimes they're not. It really just depends. And the problem is when they don't work, you just don't figure out why.
The issue, right, is like they're kind of this black box and I get why, right? But like the AI makes like this decision based on factors that you just can't see, right? You can't check or audit this segmentation logic. You can't really understand like why customer A was included and customer B was excluded. And then when something goes wrong, you then therefore can't fix it. You can really just turn on or off the feature, which isn't helpful. It's kind of a problem if you're trying to use it at scale.
That's a huge problem actually, if you think about it, because the better you are the retention program gets, the more important it is that you can debug those segments.
You can't optimize what you can't see.
