You can build useful email segments with AI today, but the AI part is smaller than the marketing suggests. In Shopify, Klaviyo, Mailchimp and Omnisend, “AI segmentation” mostly means two things: a box where you describe an audience in plain English and the tool writes the segment rules for you, and predictive scores such as spend tier or lifetime value that only switch on once your store has enough order history. For a small store, the practical setup is five or six segments built from recency, order count and spend, with AI writing the first draft of each definition and you checking it before anything is sent.
Below: what each platform’s AI segment tools do and need, the data thresholds that decide which features your store can use yet, how Shopify’s RFM groups are calculated, a starter set of segments with ready-to-check definitions, and a review routine. Platform details come from Shopify, Klaviyo, Mailchimp and Omnisend help, pricing and announcement pages checked on November 8, 2026. We have not tested these tools in a live store; the store examples are hypothetical.
What “AI segmentation” means in the tools you already use
Every segment is a set of rules: who bought what, when, how often and for how much. AI helps in two different places.
- Natural-language segment builders. You type “customers who ordered more than once but not in the last 90 days” and the tool turns it into filter conditions. Nothing about the customer data is predicted. The AI is translating your sentence into the tool’s rule language.
- Predictive attributes. The platform scores each customer on something that has not happened yet, such as future spend or the date of the next order, and lets you filter on that score. These need history to work, and small stores often do not qualify.
A third approach sits outside your email tool: exporting customer data and asking a general AI assistant such as ChatGPT or Claude to analyze it. That can work, but it moves customer data into another service, which we cover further down.
What each platform offers, and what it needs
Shopify (Customers > Segments) Ask Sidekick to write a segment from a plain-language description; it shows how many customers match and you review and save Predicted spend tier (high, medium, low); RFM groups as a segment filter Predicted spend tier needs more than 100 sales; only customers who have bought are scored
Klaviyo Segments AI: type a phrase in the condition selector, review the generated conditions, then save Predicted CLV, predicted number of orders, churn risk, expected date of next order; RFM groups Segments AI needs a paid account and is not available on EU-hosted accounts. Predictions and RFM need 500+ customers with orders, 180 days of history, orders in the last 30 days and some customers with 3+ orders. RFM needs a separate paid subscription
Mailchimp AI Segment Builder announced May 28, 2026, in beta for select customers Customer lifetime value and likelihood to purchase again, ranked low, moderate or high Standard plan and a connected store “with enough data”; no minimum is published
Omnisend AI Segment Builder from plain-language prompts, listed on all plans including Free Customer lifecycle stages based on RFM, updated hourly, using orders from the last 365 days Most accurate with at least 100 returning customers (5 to 7% of contacts)
Built from Shopify Help Center, Klaviyo Help Center and pricing, Mailchimp pricing, feature, help and newsroom pages, and Omnisend Help Center and pricing pages, checked November 8, 2026.Two points stand out. First, the plain-language builders are the most widely available AI feature, and they are the least risky, because you can read the rules they produce. Second, the predictive features have real data floors. Klaviyo states its floor precisely; Mailchimp only says “enough data”.
Which AI features can your store use yet?
Before you plan segments around predicted value or churn, check whether your store clears the thresholds. Here are three hypothetical stores run against the published requirements.
Months selling 6 14 36
Total sales (orders) 80 420 3,100
Customers who ordered 70 360 1,900
Returning customers 6 60 420
Customers with 3+ orders 0 25 160
Shopify predicted spend tier (more than 100 sales) No Yes Yes
Klaviyo predictions and RFM (500 ordering customers, 180 days, recent orders, some 3+ order customers) No No (360 customers) Yes
Omnisend lifecycle stages at their most accurate (100+ returning customers) No No (60 returning) Yes
Hypothetical stores; all store figures are our assumptions. Thresholds from Shopify, Klaviyo and Omnisend help pages checked November 8, 2026. Checked twice in Python.Store B is where many small shops sit: busy enough to qualify for Shopify’s spend tier, but short of Klaviyo’s 500-customer floor. Omnisend still assigns lifecycle stages below 100 returning customers, but says the map is most accurate above that line. If you look like Store A or B, build segments from plain rules (recency, order count, spend) and use AI only to write them faster. That is not a lesser setup. Predictive scores are estimates built from the same three signals.
How Shopify’s RFM groups work
RFM stands for recency, frequency and monetary value. Shopify gives each customer a score from 1 to 5 on each dimension, where 5 means the top 20% of your store and 1 the bottom 20%. The scores are based only on your store’s data, not on industry benchmarks. Shopify then places each customer in a group using the recency score and FM, the average of the frequency and monetary scores.
The group names are available as a segment filter called rfm_group, with values such as CHAMPIONS, LOYAL, AT_RISK, NEEDS_ATTENTION, ALMOST_LOST and DORMANT. That makes RFM the quickest way to get a sensible set of segments in Shopify without writing rules from scratch.
The half-point rule changes who lands where
When F and M add up to an odd number, FM comes out as a half, such as 3.5. Shopify’s help page says the result is rounded down. We worked through all 125 possible score combinations: 28 of them fall on a half point, and rounding down moves each of those customers into the group one step more cautious.
5, 3, 5 4.0 4 Champions
4, 4, 4 4.0 4 Loyal
4, 2, 5 3.5 3 Active (Loyal needs FM above 3)
3, 5, 2 3.5 3 Needs attention
2, 1, 5 3.0 3 At risk
Our worked examples applying the rules on Shopify’s Customers reports help page, checked November 8, 2026. Reproduced with exact fractions and a second, separate implementation in Python.One caution: the same Shopify page includes a worked example (score 154, FM 4.5) that it places in Previously loyal, which only works if 4.5 is not rounded down. Under the rounding rule it would be At risk. Shopify also says individual RFM scores are not shown anywhere in the admin. So for customers near a boundary, do not assume. Open the RFM customer list report under Analytics > Reports and check which group your best customers actually sit in before you build a campaign on it.
A starter set of six segments
These six cover the jobs most small stores need email to do. Each one maps to Shopify filters that appear in Shopify’s segment reference; the same logic can be typed into Klaviyo’s or Omnisend’s AI builder as a sentence.
Best customers
Subscribed customers in the Champions or Loyal RFM groups
rfm_group = 'CHAMPIONS' AND email_subscription_status = 'SUBSCRIBED' (repeat with 'LOYAL')
Early access and new products, not discounts
Lapsing regulars
Subscribed customers with a strong order history who have not bought recently
rfm_group = 'AT_RISK' AND email_subscription_status = 'SUBSCRIBED'
A reminder of what they bought before and what is new
Recent repeat buyers
Subscribed customers with more than one order who bought in the last 30 days
number_of_orders > 1 AND last_order_date >= -30d AND email_subscription_status = 'SUBSCRIBED'
Replenishment, care tips, a review request
High predicted spend
Subscribed customers Shopify predicts will spend the most
predicted_spend_tier = 'HIGH' AND email_subscription_status = 'SUBSCRIBED' (more than 100 sales needed)
Bundles and higher-priced items
Category buyers
Subscribed customers who bought products with a given tag
products_purchased MATCHES (tag = 'your-tag') AND email_subscription_status = 'SUBSCRIBED'
Launches and restocks in that category only
Subscribers who never bought
Subscribed contacts with no orders
rfm_group = 'PROSPECTS' AND email_subscription_status = 'SUBSCRIBED'
Best sellers, delivery and returns answers
The “never bought” segment overlaps with your welcome series, so exclude anyone who joined in the last week if your welcome email series is still running for them. And keep the best-customer segment away from blanket discount codes; our guide to customer lifetime value from profit explains why giving margin to people who would buy anyway is the most expensive mistake in retention email.
Building a segment with AI, step by step
In Shopify with Sidekick
- Go to Customers > Segments in your Shopify admin.
- Ask Sidekick for the segment in plain language, for example: “Create a segment of subscribed customers who have placed more than one order and last ordered in the past 30 days.”
- Sidekick writes the segment as a ShopifyQL query and shows how many customers match. Open the result as a new segment.
- Read the query line by line against what you asked for, then save it.
- Send to it from Shopify Messaging by choosing the segment in the To field.
Shopify says Sidekick can recommend or create segments for an email campaign, and that you review each result before it takes effect. Two Messaging rules matter here: only customers who subscribed to your emails receive the campaign, so a segment can show more customers than the number who get the email, and a campaign sent to a segment with 0 subscribers reverts to a draft.
In Klaviyo with Segments AI
- Go to Audience > Lists & segments > Create > Segment.
- Type your description directly in the condition selector. Klaviyo’s own examples include “Someone who has viewed a product and not made a purchase in the last 60 days”.
- Review the generated conditions, including the metrics, timeframes and AND/OR connectors, and edit any that are wrong.
- If you see a partial generation message, look for missing conditions before you save.
Klaviyo states that you remain responsible for the final segment definition. That is the right way to think about every AI segment builder.
Five checks before you send to an AI-built segment
- AND versus OR. “Customers who bought shoes or socks and are subscribed” can be built so that “subscribed” only applies to socks. Read the connectors.
- Time windows. “Recent” has no fixed meaning. Check the AI picked the window you meant, and remember Shopify dates follow your store’s time zone and whole days.
- Subscription status. Make sure the segment filters on email consent, or that your sending tool does. In Shopify,
email_subscription_status = 'SUBSCRIBED'is the filter;PENDINGmeans the customer has not finished subscribing. - The count. Compare the matching number with what you expect. If “best customers” returns 40% of your list, the definition is too loose.
- Spot-check five profiles. Open five customers from the segment and confirm each one belongs. This takes two minutes and catches most logic errors.
Using ChatGPT or Claude on a customer export instead
If your email tool has no AI builder, or you want to work out RFM-style groups yourself, you can export customers and ask a general assistant to help. Keep it to aggregated or de-identified data. Delete names, email addresses, phone numbers and street addresses before uploading, and match customers back by ID inside your own systems.
Check how your plan treats uploads. OpenAI says it may use content from its services for individuals, such as ChatGPT, to train its models, and that it does not train on content from business offerings such as ChatGPT Team, ChatGPT Enterprise and the API Platform. Individual users can turn off “Improve the model for everyone” under Settings > Data Controls. Anthropic says that by default it does not use inputs or outputs from its commercial products, including Claude for Work and the API, to train models, while consumer Free, Pro and Max users choose whether their chats can be used to improve Claude in privacy settings. Our guide to analyzing your Shopify sales export with ChatGPT or Claude walks through which columns to delete and a prompt that reads the file correctly.
The better use of a general assistant is often to write the rules, not to process the data. Paste in Shopify’s filter names and ask it to draft a query for the segment you describe, then build it in your own admin where the count and the subscription status are checked for you.
Common mistakes with AI segments
- Building predictive segments too early. Shopify’s spend tier and Klaviyo’s predictions have stated minimums. Below them, use plain rules.
- Treating predicted values as facts. Klaviyo describes individual predicted order counts as estimates that are most reliable when grouped. Use predictions to rank customers, not to promise anything to one person.
- Too many segments. Six segments you email well beat twenty you never use. Each one needs its own message, or it is just your full list split up.
- Forgetting the EU hosting and plan limits. Klaviyo’s Segments AI is unavailable on EU-hosted accounts and on free accounts; Mailchimp’s AI Segment Builder was in limited beta when announced.
- Never revisiting. RFM groups move as customers buy or go quiet. Omnisend updates lifecycle stages hourly; Klaviyo refreshes RFM properties nightly. A segment you built in spring is not the same list in November.
If you want segments to trigger actions automatically, such as tagging a customer when they reach a spend level, our guide to Shopify Flow automations worth setting up first covers the triggers. For a wider view of what AI can run in a store without supervision, see AI agents for store operations.
What you should check next
- Count your customers who have ordered, your returning customers and your customers with three or more orders. Compare them with the thresholds table above.
- In Shopify, open the RFM customer list report and note how many customers sit in Champions, Loyal and At risk.
- Build the six starter segments, using AI to draft each one and the five checks before saving.
- Send one campaign to one segment and compare its orders and unsubscribes with your last send to the full list.
- Recheck your segment counts monthly, and revisit predictive features once you pass the data floors.
Frequently asked questions
Do I need a paid plan to use AI segmentation?
It depends on the tool. Klaviyo’s Segments AI needs a paid account. Omnisend lists its AI Segment Builder on all plans, including Free. Mailchimp’s predictive segments need a Standard plan or higher. The Shopify help pages we checked do not state a plan requirement for Sidekick segments; predicted spend tier depends on having more than 100 sales.
How many customers do I need for predictive segments?
Shopify’s predicted spend tier needs more than 100 sales. Klaviyo’s predictions need at least 500 customers who have placed an order, 180 days of order history, orders in the last 30 days and some customers with three or more orders.
Is an RFM segment the same as an AI segment?
No. RFM groups are calculated from fixed rules on recency, frequency and spend, ranked against your own customers. AI comes in when a tool writes segment rules from your description or predicts future behavior.
Why does my segment show more customers than received the email?
In Shopify Messaging, only customers who subscribed to your emails receive a campaign, and Shopify also checks for valid email addresses. The segment can include customers who never subscribed.
Can AI pick the segments for me?
Shopify says Sidekick can recommend segments for an email campaign. Treat suggestions as a draft: you know which customers you want to reach and what you can offer them.
The takeaway
Start with six rule-based segments built on recency, order count, spend and subscription status, and let AI write the first draft of each one. Read every generated definition, check the count, and spot-check a few profiles before sending. Add predictive segments when your store passes the published data floors, not before.
Sources
- Shopify Help Center, “Creating customer segments”. help.shopify.com
- Shopify Help Center, “Managing customer segments”. help.shopify.com
- Shopify Help Center, “Predicted spend tier”. help.shopify.com
- Shopify Help Center, “Shopify-based customer segment filters”. help.shopify.com
- Shopify Help Center, “Customers reports” (RFM customer analysis). help.shopify.com
- Shopify Help Center, “Generating content with Sidekick”. help.shopify.com
- Shopify Help Center, “Create email marketing campaigns using Shopify Messaging”. help.shopify.com
- Shopify, “Segment query language reference”. shopify.dev
- Klaviyo Help Center, “How to define segments with AI”, updated January 20, 2026. help.klaviyo.com
- Klaviyo Help Center, “Understanding Klaviyo’s predictive analytics”, updated August 5, 2025. help.klaviyo.com
- Klaviyo Help Center, “Getting started with the recency, frequency, and monetary (RFM) analysis report”, updated November 18, 2025. help.klaviyo.com
- Klaviyo Help Center, “Understanding how free and paid plans differ”, updated October 20, 2026. help.klaviyo.com
- Mailchimp, “Intuit Mailchimp Launches Analytics AI and Expanded Data Integrations”, May 28, 2026. mailchimp.com
- Mailchimp, “Predictive E-Commerce Insights”. mailchimp.com
- Mailchimp, “Audience Segmentation Tools for Targeted Campaigns”. mailchimp.com
- Omnisend Help Center, “Get Started with Omnisend Segmentation”, updated April 9, 2026. support.omnisend.com
- Omnisend Help Center, “Understand Your Customer Lifecycle Map”. support.omnisend.com
- Omnisend, “Pricing”. omnisend.com
- OpenAI Help Center, “How your data is used to improve model performance”. help.openai.com
- OpenAI Help Center, “Data Controls FAQ”. help.openai.com
- Anthropic Privacy Center, “Is my data used for model training?” (commercial products), updated August 18, 2026. privacy.claude.com
- Anthropic Privacy Center, “Is my data used for model training?” (consumer plans), updated March 16, 2026. privacy.claude.com
Last reviewed: November 8, 2026. Feature names, admin paths, data thresholds, plan requirements and segment filter syntax match the vendor pages above as checked on that date.
How this article was made: researched and drafted with AI assistance, with every factual claim checked against the sources above. Stores A, B and C and their order and customer counts are hypothetical, the six-segment plan is our editorial recommendation, and the RFM rounding analysis is our own working of Shopify’s published rules.
Related guides
- Email welcome series: what to send in the first seven days
- Klaviyo vs Omnisend: pricing, automation and deliverability
- Customer lifetime value for small stores: an honest formula