K:AI Customer Agent is more than a support tool
Generated by Klaviyo AI
Most brands deploy Klaviyo Customer Agent purely for support — resolving FAQs and order-tracking questions. Zak Cassady-Dorion of ECD Digital Strategy argues that's leaving revenue on the table, since post-purchase conversations are high-intent moments where the AI already knows a customer's order history and can act on it. The post walks through three custom skills ECD built with Klaviyo's Agent Builder that turn support interactions into sales.
- Order edits become upsell moments: A skill handling shipping corrections, item swaps, and gift notes lets Customer Agent recommend a complementary product right after fixing a customer's request — for example, after correcting a shipping address — with the item added to the existing order and invoiced through Shopify.
- Size guidance adds a sales layer: For a clothing brand, a sizing skill goes beyond a standard chart lookup: it asks the shopper questions to find their fit, then recommends products that pair with what they were already considering.
- Gift conversations build segmentation data: When a shopper tells Customer Agent they're buying a gift, the skill captures who it's for as a custom property on their Klaviyo profile, which the brand can then use to add them to relationship-specific flows (e.g., dad-focused content ahead of Father's Day). The post notes this only works if the custom property already exists in Klaviyo, since Customer Agent can't create one from scratch.
When Klaviyo launched K:AI Customer Agent, most brands did the obvious thing: they pointed it at the support queue. They had Customer Agent answer FAQs and handle “Where is my order?” (WISMO) questions.
Freeing up the human support team to do more strategic, creative work is a reasonable first move. Ticket deflection has real value, after all.
But that's not where the money is.
At ECD Digital Strategy, we've been thinking about Customer Agent differently from the start. In particular, we’re interested in what Customer Agent can do in the post-purchase stage of the customer journey, beyond resolving simple tracking queries.
Every customer who reaches out after purchasing has already done the hard part. They found you, they bought from you, and now they're in a conversation with you. That's one of the highest-intent moments in the entire post-purchase journey.Of course, you’re going to resolve their issue. But you shouldn’t let the interaction end there.
Reframe your AI customer service agent as a sales tool
Implementing Customer Agent purely for support treats every conversation as a cost to minimize. Someone reaches out, they get an answer, and the ticket closes. Fine.
But think about what you actually know about that customer in that moment: what they ordered, when they ordered it, what else they've bought, what their loyalty tier is. That's more context than most salespeople ever get. Customer Agent has all of it, and it can act on it instantly.
That's the shift: AI using the shopper’s answer as an opening.
3 custom skills we built with Klaviyo Agent Builder
We've now built and launched 10+ custom Claude skills using the Klaviyo x Claude integration for ~20 clients.These are 3 of them that move the needle the most.
1. Order editing as a sales engine
A customer who's editing an order is already committed. They're not browsing or comparing. They already bought. And now they have a reason to be in a conversation with you.
We built an order editing skill that handles the full range of post-purchase changes:
- Shipping address corrections
- Item swaps
- Adding forgotten products
- Gift notes
Customer Agent can handle everything a customer might need before their order ships, without getting a support rep involved.
Every one of those interactions is now an up-sell moment. The AI agent already knows exactly what's in the order. So when someone messages in to update their shipping address, the AI agent fixes it, then recommends a complementary product based on what they ordered. Keep in mind, this is a contextual product recommendation.
The friction is nearly zero: the customer confirms, and the product gets added to their existing order. An invoice goes out from Shopify, they pay, and everything ships together.
This came from a real client problem: a blue-collar brand where workers were constantly ordering with a company credit card but using different billing and shipping addresses. The volume of requests asking to update an address was significant. Once the skill was live, Customer Agent resolved those requests automatically.
After Customer Agent updates an address, the revenue opportunity is ripe: Customer Agent can make those contextual product recommendations.
2. Size guidance with a recommendation layer
For a clothing brand, we built a sizing skill that goes beyond a standard size chart lookup. The AI agent asks relevant questions, helps the customer find their fit, and then does what a good sales associate would do: recommends products that pair well with whatever they were considering. It's a useful experience that generates revenue.
3. Gift-giving conversations that build your segments
Every apparel brand with a gifting occasion hears some version of the same question at check-out: what should I get my dad? My wife? My best friend? For one apparel client, we built a skill that treats that question as first-party data, not just conversation.
When a shopper tells Customer Agent they're shopping for a gift, the AI agent asks who it's for. The answer becomes a custom property on their Klaviyo profile: gift giver, dad. From there, the brand can automatically add that shopper to the flows built for that relationship: dad-focused gift ideas ahead of Father's Day, broader gifting content around the winter holidays, and so on.
Customer Agent pro tip: Customer Agent can't create a brand-new custom property from scratch. The property has to already exist in Klaviyo for the skill to populate it, so this only works if you've built out that structure ahead of time.
It's a smaller lift than order editing, but the principle is identical. Customer Agent is watching the words a shopper actually uses, and turning them into first-party data that shows up in a relevant email months later, not just an answer that closes out the chat.
Where Customer Agent fits across the customer journey
Here’s how we think about implementation across the full journey, not just the post-purchase window:
- Pre-purchase: Customer Agent handles discovery: size questions, product fit, "What's right for me" conversations, etc. These are high-intent moments where a good answer converts and a non-answer loses the sale.
- Cart and check-out: Customer Agent catches the hesitations that cause abandonment. A customer who has a question and can't get an answer quickly is probably gone. One who gets an immediate, accurate response often isn't.
- Post-purchase: This is where we've focused most of our energy, because it's where existing tools tend to drop the ball: order editing, gift notes, add-on items, WISMO, etc. Customer Agent handles all of this automatically, and all of it’s an opportunity to sell.
- Subscription management: For subscription brands, we use Customer Agent to make subscriptions stickier and to grow average order value. When someone asks when their next order ships, the AI agent answers, then checks their one-off order history and asks if they'd like to add a product they've bought before. That's the kind of interaction that can turn a $50 subscription into a $65 one.
How conversations beat discounts on abandonment flows (a ~10% lift)
One of the most effective things we've done has nothing to do with the support queue.
We've been running a test on browse abandonment flows that's outperforming the standard approach. Instead of sending an abandonment email that drops a discount code, we send a plain-text email written to get a reply. It asks a simple question: “What stopped you?”
When the customer replies, Customer Agent picks up the conversation and actually addresses their concern. Instead of a template, the customer gets a relevant response based on what they said. We've seen around a 10% lift in conversion on those abandonment flows compared to the standard email-plus-discount approach.
This works because a discount tells the customer their hesitation doesn't matter. All it says is, “Here's money off.” A conversation actually resolves the hesitation. Those are very different things, and people respond to them very differently.
A lot of the time, hesitation has nothing to do with price. Maybe someone isn’t sure about sizing, or timing, or whether the product is right for them. By defaulting to a discount, we were solving the wrong problem.
Use Klaviyo Agent Builder to build custom skills, no code required
When we first started building custom skills, it required significant technical work and close collaboration with Klaviyo's implementation team. We were part of an early-access group of 3 agencies helping figure out what was possible.
That's changed. Now, Agent Builder lives directly inside Customer Agent. You describe what you want it to do, and it builds the skill for you. What used to require real technical lift is now accessible to any brand willing to think creatively about what the AI agent should do.
The skills and conversation flows you build determine how Customer Agent performs. Out of the box, it's a support tool. With the right thinking behind it, it's one of the better-performing revenue channels in your stack.
That's the gap we've been building into.
Related content
