In April 2026, Klaviyo announced Custom Skills for Customer Agent. It was the kind of announcement that gets a press release and a blog post, gets bookmarked by a few hundred operators, and then gets quietly ignored by most brands for another six months. That gap is an opportunity.
Klaviyo Customer Agent is not a chatbot bolted onto your helpdesk. It is built directly on your CRM, which means it knows your customer's full purchase history, their predicted LTV, their churn risk score, their email engagement, and every product they've ever bought. When it talks to your customer, it is not guessing. It is working from a profile that your marketing team spent years building.
This is the operator's guide to what it actually is, what it can do for CPG and wellness brands specifically, and how to configure it properly so it drives revenue rather than just deflecting tickets.
What Klaviyo Customer Agent Actually Is
Klaviyo Customer Agent is an AI that handles customer conversations in real time, across your website chat and SMS. It answers questions, makes product recommendations, resolves support queries, nudges reorders, and hands off to a human when the situation calls for it. You set the rules. It handles the volume.
What separates it from a generic chatbot is the data layer underneath. Every conversation is informed by the customer's actual profile. If they bought your collagen powder three times and haven't reordered in 45 days, the agent knows that. If they're a predicted high-LTV customer who has never bought your bundle, the agent can surface it. If they're mid-churn-risk and have a support issue, the agent can prioritise the resolution and add a retention offer.
This is the difference between support that deflects tickets and support that generates revenue. The average Klaviyo Customer Agent deployment resolves 65% of questions autonomously. Naked Wardrobe, a fashion brand deploying it across both web and SMS, resolved 84% of conversations autonomously and saw a 28% increase in average order value. Happy Wax resolved over half of all conversations without any service team involvement.
The Three Layers: Skills, Tools, and Guidance
The April 2026 update introduced Custom Skills and built a proper three-layer architecture around how the agent operates. Understanding this is the difference between a basic deployment and one that actually moves the needle.
Custom Skills
Custom Skills let you define entirely new AI workflows for your specific business needs, in plain language, without engineers. You describe what you want the agent to do and it builds the workflow. A subscription brand might build a skill that detects a customer is approaching their reorder window based on usage patterns and initiates a nudge conversation. A CPG brand might build a skill that asks discovery questions and recommends the right product format based on answers. These are not pre-built templates. They are bespoke workflows you define for your brand's specific use cases.
Custom Tools
Custom Tools let you connect the agent to any external system in your tech stack. Your subscription management platform, your loyalty programme, your 3PL for order status, your reviews tool. Once connected, the agent can read from these systems and act on them mid-conversation. A customer asks about their delivery? The agent pulls live tracking data. A customer asks about their loyalty points? The agent surfaces their balance and the next reward tier. No copy-paste, no forwarding to another team.
Agent Guidance
Agent Guidance is where you set the behavioural rules. Your brand voice, your tone, your escalation triggers. When should the agent hand off to a human? What topics should it never handle alone? What's the exact language it should use for refund requests? You write these rules in plain language. The agent applies them consistently across every conversation, at any hour, across any channel.
What CPG, Wellness, and F&B Brands Should Actually Build
Most of the case studies Klaviyo publishes are fashion brands. If you're selling supplements, functional drinks, or skincare, the use cases look different. Here are the ones worth building first.
Reorder nudge conversations
CPG and wellness customers have predictable consumption rates. A customer buying a 30-serving protein powder every 35 days is one of the most forecastable buyers in ecommerce. Build a Custom Skill that initiates a proactive conversation around day 28 to 30: 'You're probably running low. Want us to send your next order now?' This is not an email. It is a real-time conversation that the customer can respond to immediately. Conversion rates on these are significantly higher than email nudges because the context is immediate and the friction is low.
Subscription upgrade and lock-in
If a customer is a repeat one-time buyer, the best moment to convert them to a subscription is mid-purchase conversation, not a pop-up at checkout. Build a skill that detects repeat one-time buyers and asks a single discovery question during a conversation: 'You've ordered three times in the last 90 days. Would a subscription save you 15% and make sure you never run out?' One question, one decision, done. Your human team would never have time to do this at scale. The agent can.
Personalised bundle and gifting advisor
For brands with multiple SKUs, the gifting and bundling use case is underused. Build a skill that acts as a product advisor: it asks a few questions, understands the customer's goal or recipient, and recommends the right combination of products with a specific rationale. This is not a quiz funnel. It is a genuine conversation that responds to what the customer says rather than routing them through a decision tree. Brands using this approach see significantly higher AOV on first orders because the recommendation feels personalised rather than algorithmic.
Allergen and ingredient handling
This is specific to F&B and wellness, and it is where generic chatbots fall down badly. Build a knowledge layer with your full ingredient lists, allergen declarations, certification details, and formulation notes. Then configure the agent with strict guidance: specific questions about medical conditions or drug interactions should always escalate to a human. Handled correctly, this is one of the highest-trust use cases available. A customer who gets a clear, accurate, immediate answer to an allergen question converts and stays loyal at a much higher rate than one who emails and waits 48 hours.
VIP routing and retention recovery
Connect the agent to Klaviyo's predicted LTV data. If a high-LTV customer contacts support with a complaint or issue, the agent should handle this differently from a first-time buyer. Different tone, faster resolution, and a retention offer in the resolution. Build a Custom Tool that reads the customer's LTV and churn risk score, and a Custom Skill that adjusts the resolution path based on those values. This is retention economics applied at the conversation level.
How to Configure It Properly
Most brands set up Customer Agent, push it live, and wonder why the results are mediocre. The configuration is where the work is.
Build your knowledge base deliberately
Customer Agent auto-ingests your storefront when you connect it. But auto-ingestion is a starting point, not a finished product. Go through the knowledge base manually. Fill the gaps. Your most-asked questions are not always the ones your website answers clearly. Pull your last 90 days of support tickets, find the top 20 question types, and make sure each one has a clear, accurate answer in the knowledge base. This single step has more impact on autonomous resolution rate than any configuration setting.
Write your Agent Guidance like an operations manual
The most common mistake is writing vague guidance. 'Be friendly and helpful' is not guidance. 'Match the customer's energy: brief and direct when they're transactional, warmer and more descriptive when they're exploring or gifting' is guidance. Define your tone for three different conversation types: support, purchase, and general enquiry. Write escalation rules for every scenario where a human must take over. The agent will apply these rules perfectly and consistently. Your human team can't do that.
Connect your subscription platform before launching
If you're running subscriptions via Recharge, Loop, or Skio, connect it as a Custom Tool before you go live. The most common CPG support query is 'how do I skip, pause, or cancel my subscription.' If the agent can't answer and act on this immediately, you're burning the trust of your highest-value customers. Get the integration done first. It turns a potential churn moment into a save.
Start with web, then expand to SMS
Web chat is lower stakes: customers expect a degree of automation there. SMS is higher trust and higher intent. Get your web deployment dialled in first. Monitor your resolution rate, your escalation rate, and your customer satisfaction signals for four to six weeks. Then expand to SMS with a configuration that's already been tested and refined.
Review your escalation patterns weekly
The escalations tell you where the knowledge base is thin or where the guidance is wrong. Build a simple weekly review: what did the agent escalate, and why? For every escalation type that appears three or more times, that's a gap to close. Either add the knowledge, refine the skill, or adjust the guidance. Within six to eight weeks of consistent review, your autonomous resolution rate should be meaningfully above the 65% baseline.
A mid-market skincare brand we reviewed increased Klaviyo's share of total store revenue from 15% to 42% over six months, with repeat purchase rate going from 18% to 31% and lifetime value rising 55%. Not all of that is Customer Agent. But the compounding effect of having personalised, data-informed conversations at every post-purchase touchpoint is significant.
What Customer Agent Is Not
Worth being clear on this. Customer Agent is not a replacement for your Klaviyo flows. It operates in real-time, synchronous conversations. Your flows operate in asynchronous, time-based sequences. You need both. They serve different moments in the customer relationship.
It is also not a replacement for a human support team at the senior end. The agent handles volume: repeat questions, status queries, standard recommendations. The high-stakes conversations, relationship-building with wholesale partners, and edge cases with angry customers still benefit from a human on the other end.
The brands that see the best results treat Customer Agent as a layer that handles the 65 to 85% of conversations that don't need human judgement, freeing the team to focus on the 15 to 35% that do. That reallocation of time is where the real operational value sits.
The Competitive Window
Custom Skills for Customer Agent is in managed beta as of June 2026. That means the brands deploying it properly right now are six to twelve months ahead of the majority of their category. In CPG, wellness, and F&B, where repeat purchase rate is the single most important growth lever, having a retention and conversion layer that operates at the conversation level is a meaningful structural advantage.
Join the waitlist inside your Klaviyo account. Get familiar with the standard Customer Agent deployment first. Set it up properly, review your escalations, and build your knowledge base to a level where you're clearing 70% autonomous resolution before you start worrying about Custom Skills. The foundation is what determines the ceiling.
Growth Audit
Find Out If Your Klaviyo Setup Is Leaving Revenue on the Table
I'll audit your Klaviyo account, your retention flows, and your customer conversation data. Then I'll tell you exactly where the gaps are, what to build first, and what to stop wasting time on. No pitch deck. Just the numbers and the plan.
Book Your AuditFrequently asked questions
What is Klaviyo Customer Agent?
Klaviyo Customer Agent is an AI-powered assistant built directly into the Klaviyo CRM that handles customer conversations autonomously across web chat and SMS. It can answer product questions, resolve support queries, recommend products, handle reorders, and escalate to human agents based on your rules. Custom Skills, launched April 2026, let brands define entirely new AI workflows in plain language without engineers.
How much of customer support can Klaviyo Customer Agent handle automatically?
Klaviyo states that Customer Agent resolves 65% of questions autonomously on average. Naked Wardrobe achieved 84% autonomous resolution deploying it across web and SMS. Happy Wax resolved over 50% of conversations without service team involvement. Results depend on how well you configure the knowledge base, Agent Guidance, and custom skills.
What are Klaviyo Custom Skills for Customer Agent?
Custom Skills let brands define entirely new AI workflows for specific use cases in plain language, without engineers. You describe what you want the agent to do, connect external systems via Custom Tools, and set behavioural rules through Agent Guidance. Examples include post-purchase reorder nudges, VIP routing, personalised gifting advisors, and subscription upsell conversations.
Does Klaviyo Customer Agent work for CPG and F&B brands?
Yes. CPG, wellness, food, and beverage brands are strong use cases because their customers have high-frequency repeat purchase patterns, ingredient and allergen questions, subscription decisions, and gifting needs. An AI agent can handle reorder reminders mid-conversation, subscription upgrade nudges, ingredient FAQs, and personalised bundle recommendations based on CRM data.
How does Klaviyo Customer Agent compare to Gorgias?
Gorgias is a dedicated helpdesk for ticket management and human agent workflows. Klaviyo Customer Agent is built on your CRM data, meaning it personalises every conversation using purchase history, predicted LTV, churn risk, and email engagement. Gorgias manages support volume. Klaviyo Customer Agent uses support conversations as a revenue channel, driving recommendations, upsells, and reorders in the same thread.
Is Klaviyo Customer Agent available to all Klaviyo users?
Klaviyo Customer Agent is available now on eligible Klaviyo plans. Custom Skills for Customer Agent launched April 30 2026 and is currently in managed beta. Brands can join the waitlist directly inside the Klaviyo platform.
About the author
Caner Veli built Liquiproof from zero to 3,000+ global retailers in under 6 years. He now helps DTC and CPG brands fix broken growth engines and scale 2x-15x in 90 days.