Klaviyo AI: what it actually does, and what's worth using

What AI features does Klaviyo have?

Klaviyo's AI page lists a lot. The ones a Shopify brand is most likely to use:

• Composer. Klaviyo calls it its AI marketing agent. You tell it what you're trying to do, and it builds the audience, drafts the content and maps out the send strategy.
• Subject line generator. Suggests subject lines inside campaigns and flows.
• Segments AI and Flows AI. Generative tools for building segments and flows.
• Predictive analytics. Predictions like customer lifetime value, churn risk and expected date of next order.
• Customer Agent. An AI customer service agent that uses your customer data.

The same page also lists an AI photo editor, apps for Claude and ChatGPT, and an MCP server for connecting AI tools to your Klaviyo data. From Klaviyo's own FAQ: many of its AI features are optional, but its machine learning and predictive analytics can't be opted out of.

What does Klaviyo Composer do?

Per Klaviyo's help center, Composer can:

• Analyze your account and answer questions about account and flow performance
• Audit your work and return prioritized, plain-language findings
• Draft email, push, SMS and MMS content for you to review
• Build campaigns, flows and segments
• Run scheduled or recurring tasks

A few limits Klaviyo is upfront about. Drafts may not always be accurate. Composer works best with content built in Klaviyo's drag-and-drop editor and is weaker on image-heavy templates. For most audit findings, you make the fixes yourself. It starts on a free plan with limited capacity that resets each cycle, and some asks cost credits depending on how complex they are.

And the line I like most: "Composer does not send or publish campaigns, flows, or messages without your sign-off."

If you want to try it, start here. First drafts when you're staring at a blank email. Segment definitions you'd otherwise click together by hand. An audit as a starting checklist. If you want a human checklist to compare it against, here's my Klaviyo audit checklist.

Where I wouldn't lean on it alone. Deciding what you're selling this week, who it goes to, and whether the email gives people a reason to buy. More on that below.

Is Klaviyo's AI subject line generator any good?

It's useful for one thing: getting options fast. You describe the email in a sentence, it suggests 3 subject lines, and you can regenerate as many times as you want.

What it can't do is tell you which one will win on your list. My starting rules: under 50 characters, preview text 80 to 100 characters, no fake "Re:" or "Fwd:", and judge winners by clicks, not opens.

And even those rules flip by audience. Across 7 brands and 2,821 campaigns in my data, news beat brand voice and vague questions were the worst-performing pattern. Within one category of 4 brands, emoji in subject lines helped two brands and hurt two. A fitness tech brand I work with got more than 2x the clicks from subject lines of 20 characters or fewer than from 41 to 60 character lines. An AI tool trained on everyone can't know which of those is true for you. Your own tests can. More in my subject line best practices.

How does Klaviyo's predictive analytics work?

If you only look at one of these for a consumable brand, make it this one. Klaviyo's help center lists these predictions on customer profiles:

• Predicted customer lifetime value (spend in the next year), plus historic and total CLV
• Churn risk
• Average time between orders
• Expected date of next order

To get them, Klaviyo says you need at least 500 customers who've placed an order, an ecommerce integration like Shopify, at least 180 days of order history with orders in the last 30 days, and some customers with 3 or more orders. The model retrains at least once a week.

Klaviyo's own caveats are worth reading twice. The predictions are meant to be accurate on average, not exact for any one person. The expected date of next order ignores which products someone bought, so brands with different reorder cycles should use separate product-based flows. And if many of your customers reorder often, Klaviyo suggests using it mainly for one-time buyers.

That lines up with how I'd build it. Across four consumable brands in my data, replenishment flows earned about $2.23 per person, about 28x the best campaign. If your products run out at different speeds, I'd time reorder reminders by product. For VIPs, I build lists for clients by spend, with the threshold set per brand. Predicted CLV and churn risk can be an extra signal on top if your account qualifies. My replenishment email timing post walks through the timing side.

What AI do I actually use?

Not much inside Klaviyo. I don't really use its built-in AI features. I do use AI a lot, through Claude, Claude Design and ElevenLabs.

So this isn't an anti-AI post. My point is narrower: whichever tool does the drafting, the same few jobs stay with a person.

Where does a human still matter?

AI or not, here's what I'd keep with a person.

Deciding the offer. Content warms people up, but somebody has to ask for the sale. A wellness brand I work with sent a month of all-educational emails with no offer, and campaign revenue went from $3,146 to $110, down 96.5%.

Knowing what's going on in the business. When something sells out, the restock email is your biggest send of the month. A snack brand doing about $650K a month made $48,771 from one restock email, "We're back baby!", its best campaign ever. No AI tool knew that pallet was arriving. Someone on the team did.

Checking that the email works. Click every link before you hit send. Across the brands I work with, about 20 sends showed opens and revenue but zero clicks, which means tracking was broken. A fitness gear brand had a restock email with double its usual click rate that made $0. Clicks with no sales means something broke. AI drafted or not, somebody has to click.

Reading the results honestly. Judge by clicks, not opens. In my four-brand data, click rate tracked revenue at 0.65 and open rate at 0.32.

Where these numbers come from

The flow numbers and correlations are pooled from 4 consumable brands I work with: 2,897 campaigns and $17.7M in email revenue. The subject line patterns come from 7 brands and 2,821 campaigns. Both are small samples, so read them as directional. Single-brand numbers are from individual client accounts. Klaviyo feature details come from Klaviyo's own pages, listed in Sources.

FAQ

WRITTEN BY Jack White

Founder of Bardown Digital, a Chicago-based Klaviyo email and retention agency for Shopify brands. He's worked with brands doing $500K to $100M a year, with $100M+ in email-attributed revenue across them.

Instagram @jackintheinbox LinkedIn