Big bets are being made in the customer service space right now. From acquisitions to consumers expecting more from AI, brands are starting to ask what good customer service looks like in the agentic era.
My team at Klaviyo has been thinking about this for a long time. Here’s one of the mental exercises we did to understand the mechanics behind great customer service.
Think about a customer service moment that stayed with you. Maybe a hotel greeted you by name and already knew you wanted a high floor and a late check-out, before you asked. Maybe a brand called to say your delivery was running late. Maybe a shop assistant remembered the gift you bought 6 months ago and asked how it went.
What made those moments remarkable wasn't speed or simple problem solving. It was that someone knew you, and that made you feel like more than a ticket number.
That's the standard your customers now hold your AI customer agent to. On your site, in a chat window, over text, they bring the same thing to the conversation: know me.
AI agents should be held to a higher standard than resolution rates
Most brands have launched an agent, or they're about to. That part is straightforward. The harder part is understanding and implementing the difference between AI agent capability and customer experience.
Capability is what your agent can do. Experience is what your customer feels when it does it, and is where the relationship is won or lost.
- Does the agent know their history (purchases, open complaints, browsing behaviors, loyalty status, etc)?
- Does it sound like your brand when something goes wrong?
- Does it close the loop, or hand them to a human and make them start over?
A capable agent resolves a ticket, and it's why the dominant conversation in customer service AI right now is resolution rates. But those rates are the floor, not the ceiling. They tell you what happened, but they don’t tell you whether it mattered.
A good AI customer agent experience is one that makes the customer feel known while it resolves. So, what does your AI agent actually know about each customer it talks to? That’s a data architecture question.
Built in beats bolted on
Most customer service AI is bolted on top of a helpdesk. And a helpdesk manages tickets, not customers. There's no purchase history, no loyalty status, no sense of what the person was browsing 10 minutes ago. So the conversation starts cold each time, whether a human or an agent is running it.
But when service AI runs natively on the same customer data platform that powers marketing (not just a helpdesk), every service interaction with a customer begins differently. The AI agent already knows who it's talking to, and every conversation feeds information back to the same customer profile, so each interaction is smarter than the last.
That architectural decision shapes everything downstream: what the AI agent can know, what it can do with that knowledge, and whether service compounds into something valuable for the customer relationship over time or just closes tickets and resets.
Deflection is table stakes. Revenue is the real opportunity.
When an agent knows your customer, the interaction becomes genuinely useful, and useful is how it becomes revenue. Instead of forced, clueless up-selling, AI customer service conversations look like surfacing the right offer at the right moment, recommending the product that actually fits, and closing out a complaint in a way that sends the customer back to the site rather than away from it.
Women’s shapewear brand Naked Wardrobe is a good example of what this looks like in practice. Since using Customer Agent, their customers describe shopping with them as similar to being in a luxury boutique. The agent handles post-purchase questions, but it also has the product knowledge of a skilled salesperson: recommending the right fit, answering pre-purchase questions in real time, and giving customers the kind of 1:1 consultative experience that's usually only possible in-store. Revenue followed, not because the agent closed tickets faster, but because the AI agent knew enough about each customer to be genuinely helpful, not just responsive.
To get the most out of AI customer service brands need to prioritize every relationship and ensure they are building a data foundation that makes more of those interactions possible.
So, what's your AI agent trained on? It's a question worth asking right now, before the gap between you and your closest competitor gets far harder to close.
Related content
