Like every conversation marketers have today, AI was omnipresent at K:BOS 2026, but one of the more practical conversations at Klaviyo’s flagship customer conference started somewhere more fundamental: you can’t personalize for a customer you can’t see.
K:BOS brought thousands of marketers, operators, and brand leaders to Boston in early September to discuss AI, customer data, personalization, and the future of customer engagement. On stage, our very own AVP of Sales & Partnerships; Jillian Dimoff joined Josh Bear, Founder and CEO of AMB Interactive, and Emily Britt Parker, Director of Lifecycle Marketing at LAKE, to explore what better personalization actually looks like in practice.
The discussion ranged from identity resolution and cross-device behavior to maternity journeys, signup optimization, and customer sensitivity. But four lessons stood out for marketers trying to turn more customer data into experiences people actually want.
Personalization tends to focus on what happens after a marketer knows who somebody is. The panel challenged that starting point.
As Jillian explained, “about 95% of your traffic is unknown.” That leaves a significant gap between the experience brands want to create and the customers they can actually recognize. Untitled document
And anonymous doesn’t necessarily mean new. A customer might browse on their phone before purchasing from a work laptop, they might have previously supplied an email address in-store, or they might return months after initially viewing a product.
Without identity connecting those interactions, personalization can quickly become impersonal. A customer completes an order on one device, for example, only to receive an abandonment message triggered by another.
Josh described the goal simply: “How do you send relevant, timely, but never an intrusive message?” That requires solving identity before optimizing the message.
It’s also where Wunderkind complements platforms such as Klaviyo. Wunderkind’s Autonomous Marketing Platform (AMP) uses identity resolution, behavioral intelligence, and real-time decisioning to recognize more customer intent, then activates those signals through the tools marketers already use. It’s an intelligence and activation layer, not another ESP that requires teams to move their existing programs elsewhere.
The most compelling example of the session came from LAKE. The sleepwear retailer had tripled the share of customer acquisition coming from maternity products, but retention within that cohort wasn't healthy. Traditional retention tactics weren't fixing it.
Instead of simply changing creative or launching another win-back flow, LAKE mapped the customer's maternity journey month by month. That exposed a fundamental problem with static segmentation. Someone three months pregnant and someone seven months pregnant might technically belong to the same “maternity customer” segment, but their circumstances, priorities, and product needs can be very different.
LAKE therefore began capturing a due date for relevant maternity purchases. Where customers don't provide one, the retailer can infer timing from purchase behavior. That gives the brand a useful reference point for adapting communication across early pregnancy, hospital preparation, and postpartum.
The wider lesson extends well beyond maternity. A customer profile tells you who someone has been. Intent can help tell you what matters to them now. Real-time behavior, repeat product views, cross-device browsing, an abandoned checkout or renewed interest in an old category can add the context needed to decide what the next interaction should be.
The LAKE example also surfaced an important counterweight to the industry's obsession with collecting more data.
Pregnancy is sensitive. Circumstances can change suddenly, making yesterday's perfectly relevant message deeply inappropriate today. LAKE therefore created an opt-out specifically for maternity communications rather than forcing customers to unsubscribe from the brand altogether. Retail and customer service teams can also apply that suppression when necessary.
It’s an important distinction. Good personalization isn't about using every available signal simply because you have it, it requires understanding when not to activate one.
The same principle came up when the panel discussed third-party data. Marketers should operate with caution when contacting people who aren't expecting an email and monitor engagement closely, because poor targeting can damage both the customer experience and deliverability.
For marketers, this creates a useful test: Does this data help us make the interaction more relevant for the customer, or merely give us another opportunity to contact them?
The distinction matters as AI makes it increasingly easy to generate and orchestrate communications. More automation shouldn't mean more noise.
Sometimes improving marketing performance means doing less. LAKE's signup experience offers a good example. Rather than immediately showing a popup on a visitor's first page view, an all-too-familiar eCommerce tactic, the retailer tested giving shoppers more time to experience the brand.
After testing different approaches, LAKE moved the signup unit to page two with a five-second delay. The result? A 38% increase in submit rate with purchases from new leads increased
The lesson isn't that every retailer should delay its signup unit by five seconds. In fact, the lesson is that customer experience and commercial performance don't have to be opposing forces.
Recognizing customers across sessions can make that experience smarter still. If you already have someone's email address, there's little value in repeatedly asking for it. If they're eligible for SMS but haven't subscribed, a different experience may be more useful. If they've already purchased, continuing to show a welcome offer may simply waste margin.
Real-time decisioning makes those distinctions possible. Instead of asking how to show every visitor the “best” experience, marketers can ask which experience makes sense for this customer, at this moment.
One of the clearest themes from K:BOS was the movement toward more autonomous, AI-assisted customer engagement. But smarter decisioning is only as useful as the customer context underneath it.
That brings the conversation back to Jillian’s closing point: “Personalization is really dependent on truly knowing your customers, especially the unknown.”
For eCommerce marketers, that means the next stage of personalization isn't simply another flow, segment, or AI-generated message. It starts by recognizing more customers, understanding their intent as it changes, and using those signals to decide when an interaction will genuinely add value.
Wunderkind AMP is designed to provide that intelligence while connecting with the marketing stack brands already use. Identity resolution expands who can be recognized; behavioral signals reveal intent; AI decisioning determines the appropriate next action; and native integrations allow those decisions to be activated through existing channels and platforms.
Because ultimately, the goal isn't to send more messages, it's to make every message more worth receiving.