Switch on your platform's built-in abandoned checkout email first (in Shopify it lives under Apps, then Messaging, then Automations, and it is free), set it to send about an hour after someone leaves, then add two follow-ups. Use AI to write the copy, subject lines and product-specific lines; let plain settings decide who is emailed and when.
One caveat up front: in many places a cart reminder counts as marketing, so who you email matters more than what the email says. And get the first email right before building the other two. It goes to everyone who enters the sequence, so any mistake in it is repeated more often than anything else you send.
Abandoned cart or abandoned checkout: who you're allowed to email
The terms get used interchangeably, but platforms treat them differently, and it changes who you can reach.
| Event | What happened | Can you email them? | Name in Shopify / Klaviyo |
|---|---|---|---|
| Abandoned checkout | Started checkout and entered an email address, but did not pay | Yes: you have their address | "Abandoned checkout emails by Shopify" / Started Checkout |
| Abandoned cart | Added items to the cart but never reached checkout | Only if the platform already knows who they are, such as a subscriber | "Recover abandoned cart" template / Added to Cart |
| Abandoned browse | Looked at products, added nothing | Same as above, known subscribers only | "Convert abandoned product browse" template |
Shopify defines an abandoned checkout as one left incomplete for more than ten minutes after the customer has given their email. Its cart and browse automations send only to customers subscribed to email marketing by default. For a small shop, the checkout email is where most of the recoverable sales are, so start there.
Step 1: switch on the checkout email and set its two key settings
In Shopify, go to Apps, then Messaging, then Automations, and open the abandoned checkout automation. Two settings decide how it behaves:
- Send to: whether the email goes only to people subscribed to your marketing emails or to everyone who abandons a checkout. Emailing everyone reaches more people; emailing subscribers only is the safer reading of consent rules in many countries. If you are not sure which applies to you, keep it on subscribers and ask an adviser before widening it.
- Send after: how long after abandonment the email goes out. The options run from about an hour to a day. For most small shops an hour works well, because the customer still remembers what they were buying.
Cost: Shopify Messaging includes 10,000 free emails a month, and Shopify says abandoned checkout automations are always free and do not count towards that limit. Before editing anything, abandon a checkout on your own store with your own email so you see exactly what customers receive now.
Step 2: plan three emails around how your customers buy
One reminder recovers some sales; a short sequence recovers more, as long as each email has a different job. Here is a filled-in plan for a florist, where timing is tied to delivery dates:
| When | Job | Florist version | |
|---|---|---|---|
| 1 | 1 hour | Remind and remove the likely doubt | Shows the bouquet; answers "will it arrive on the day?" with the cut-off time |
| 2 | 24 hours | Reassure | Freshness promise, substitution policy, two short reviews |
| 3 | 72 hours | Offer a way forward | A smaller size or a different bouquet at a lower price; no discount by default |
Shopify's built-in automation handles email 1. For emails 2 and 3 you either build a longer workflow in Shopify's automations or Shopify Flow, or move to a dedicated email platform (see below).
Step 3: write the emails with AI
Give the AI your facts and the job of each email, not "write an abandoned cart email", which produces the same urgent template everyone gets.
Write three abandoned checkout emails for a small florist.
Voice: warm, calm, local shop. British English. No exclamation
marks in subject lines. No fake urgency ("only 2 left!").
Facts:
- Same-day delivery if ordered before 1pm Monday to Saturday.
- Flowers are replaced free if they arrive damaged or wilt within
3 days of delivery (send a photo).
- We may substitute a stem of similar colour and value if a
flower is unavailable.
Email 1 (1 hour): remind, show the cart, answer "will it arrive
on the day?". 60-90 words.
Email 2 (24 hours): reassure with the replacement promise and
substitution policy. 60-90 words.
Email 3 (72 hours): suggest a smaller size or another bouquet.
No discount. 50-80 words.
For each: 3 subject line options under 45 characters and a
preview line under 90 characters.
Leave [CART ITEMS] and [CHECKOUT LINK] as placeholders.An illustrative extract of what comes back for email 1:
Subject options: Still thinking about these flowers? /
Your bouquet is waiting / Order by 1pm for today
Preview: Same-day delivery if you order before 1pm.
Hi there,
You left something lovely behind. [CART ITEMS]
If you'd like them delivered today, order before 1pm and we'll
hand-tie your bouquet this morning. Hurry, our flowers sell out
fast! [CHECKOUT LINK]What to fix: "Hurry, our flowers sell out fast!" breaks the no-urgency rule and is not true of a florist who makes bouquets to order. "Order by 1pm for today" is a good subject line only if the email goes out before 1pm; at 6pm it is wrong, so pick "Your bouquet is waiting" as the default. And "this morning" should be "on the day". Small fixes, but each would have gone to every customer.
Email 2 usually needs heavier editing, because reassurance is where models reach for proof they don't have. An illustrative draft:
Subject: We guarantee you'll love them. Every bouquet is covered by our no-questions-asked freshness promise: if you're not happy, we'll replace it. Don't just take our word for it: "Absolutely stunning, arrived perfect!" and "Best florist in town, five stars."
Three problems. "No questions asked" contradicts the actual promise, which needs a photo and applies within three days, so a customer could reasonably hold you to a policy you never offered. The two reviews were invented: the prompt said "two short reviews" and gave none, so the model wrote some. And "we guarantee you'll love them" is a claim no florist can keep. The corrected version states the real policy in one sentence, then quotes two genuine reviews you've picked (with the reviewers' permission if you use their names). The lesson for the prompt: never ask for content, such as reviews, that you haven't pasted in.
Email 3 has a different trap: it names products. An illustrative draft suggested "our Petite Posy from $29 or the Seasonal Jar at $35". Both sounded right, and neither existed; the florist's smaller options were a "Small Hand-tied" at $38 and a "Letterbox Bouquet" at $32. Models fill product names from the pattern of the prompt when you haven't supplied them. Either paste the exact names, prices and links of the two alternatives you want to suggest, or use the platform's product-recommendation block and let the AI write only the sentence around it ("If that one's a bit much, these two are smaller and still made up on the day").
Where AI personalises, and where plain rules do better
The personal part of a cart email, the product photo, name and price, comes from the platform's dynamic cart block, not from AI. What AI adds is a set of variants you write once and check once:
- A line per product type. Bouquets get the delivery cut-off; vases get "it arrives boxed, with a breakage guarantee"; plants get care information.
- Tone exceptions. Sympathy arrangements get a separate, quiet version with no sales language at all. Use the platform's conditions (product tag or collection) to send it. A version that works: subject "Your order with us"; body "You started an order for [CART ITEMS] but it wasn't completed. If you'd still like us to prepare it, it's saved here: [CHECKOUT LINK]. If you'd rather talk to us about the flowers or the card message, just reply to this email." No subject-line hook, no follow-ups, and a direct route to a person.
- Subject line options to test, one change at a time.
What should stay as rules, not AI judgement: who receives the emails, how many, when the sequence stops (as soon as they buy, or if they buy something else), and whether a discount is ever offered. One edge case to cover in those rules: the same customer abandoning twice in a day, say a bouquet at lunchtime and a different one that evening. Without a condition such as "not already in this sequence in the last 7 days", they get two parallel sets of reminders and six emails in three days. Most platforms have a re-entry or "skip if already in flow" setting; switch it on before launch. Our tutorial on segmenting even a small email list explains how to build those conditions.
A realistic mistake: the reminder that arrived after the occasion
A florist's sequence sent email 3, "Still time to make their day", three days after a customer abandoned a checkout for flowers to be delivered on a named date. The date had passed. The customer replied, understandably annoyed, that the birthday was over. The fix: add a condition that stops the sequence once the chosen delivery date has passed, and for peak days such as Mother's Day, send a single reminder only, a few hours after abandonment.
Discounts: the setting that teaches customers to abandon
If every abandoned checkout earns a 10% code, regular customers learn to leave their carts and wait. Three rules keep discounts from eating your margin:
- No discount in emails 1 or 2. Most recovered orders come from the reminder and the reassurance.
- If you use one in email 3, make it a single-use code with a short expiry, and only for first-time customers.
- Exclude anyone who received a recovery discount in the last 90 days.
Run the numbers before switching a discount on. On a $55 average order with a 40% gross margin, a 10% discount takes $5.50 of the $22 margin. It has to recover a lot of orders that would not have come back anyway to pay for itself. Put numbers on it: if email 3 brings back 4 orders a month without a code, adding a 10% code costs $22 of margin on those 4 alone. The code has to produce at least one more order every month (worth $22 of margin) just to break even, and more once you count the regulars who start waiting for it.
When to move from the built-in email to Klaviyo or similar
Shopify's built-in tools cover a small shop's first year. Consider a dedicated platform when you want several emails with different conditions, cart and checkout flows that avoid duplicate messages, and reporting per email. Klaviyo's free plan covers up to 250 active profiles and 500 email sends a month; its paid plans scale with the size of your list, so use its price estimator. Klaviyo recommends two flows, one triggered by Started Checkout and one by Added to Cart, with a filter on the cart flow so someone who reaches checkout does not get both. Klaviyo's AI marketing agent, now called Composer, can draft emails and flows across email, SMS and other channels; treat its output exactly like the prompt output above, and read every email before the flow goes live. The tutorial on the best AI email marketing tools for small businesses compares the alternatives.
A two-email version for an independent bookshop
A bookshop's carts are small and often mixed: a paperback, a card, a tote bag. Discounts make little sense on thin book margins, so its sequence is two emails, not three. Email 1 at one hour lists the items and the free-delivery threshold ("add $6 more for free delivery"). Email 2 at 48 hours is only sent for signed or pre-order titles, and it states honestly how many signed copies remain, because that scarcity is real. AI's job is writing a short, true line about each signed edition, checked by the bookseller who ordered it.
A florist's recovery sequence in numbers
Run the numbers for a made-up but typical florist whose Shopify store sees about 120 abandoned checkouts a month with an average cart of $55. It sets up the three emails above, sending to subscribers only, with no discount.
- Setup time: about two hours to write and check the emails with AI, one hour to build the conditions and test.
- Cost: $0 on Shopify's built-in automation for email 1; emails 2 and 3 within the free monthly allowance.
- If 70 of the 120 are subscribers and 8 of those complete their order after an email, that is about $440 a month recovered.
Your recovery rate could be higher or lower; the point is to measure it rather than trust a benchmark.
The per-email split is where the decisions come from. Say the 8 recovered orders break down as 5 after email 1, 2 after email 2 and 1 after email 3, and email 3 also drew 3 unsubscribes. The recovery rate is 8 out of 70, about 11%. Email 3 brought in $55 and cost three people who'd agreed to hear from you. Before dropping it, try one change: rewrite it around a smaller-size option for carts over $60, where the price is the likely doubt. If a month later it still recovers one order and loses three subscribers, switch it off and run a two-email sequence.
How to check it is working, and not annoying people
- Recovery rate: orders attributed to the automation divided by checkouts that entered it. Check monthly.
- Per-email performance: if email 3 recovers almost nothing and draws unsubscribes, drop it.
- Replies and complaints: read every reply to a recovery email for the first month. They reveal wrong timing and wrong tone fast.
- Where the emails land: send test abandonments to a couple of personal addresses at different email providers and check they arrive in the inbox rather than spam or a promotions folder. A recovery email nobody sees recovers nothing.
- That the sequence stops: buy after email 1 on a test order and confirm emails 2 and 3 never arrive. Being chased for something you have already bought is the complaint that damages trust fastest.
- A test abandonment every quarter on your own store, to catch broken links and outdated delivery promises.
Before any change goes live, run it through an email QA checklist, and keep the trust basics in mind: the mistakes in online shop AI mistakes that hurt trust apply to automated emails as much as to product pages.
Abandoned cart emails: questions small shops ask
Do abandoned cart emails need the customer's consent?
In many places a reminder that encourages someone to complete a purchase is treated as marketing, so consent rules can apply even though the person started buying. The rules differ by country and depend on how you collected the address. Shopify's own automation sends only to email subscribers unless you change it. If you are unsure, keep that default and ask a data-protection adviser.
Can AI write a different email for every customer?
It can, but a small shop rarely needs it. The cart contents already make each email personal, because the product, image and price are pulled in automatically. Where AI helps is writing a few versions by product type, such as one line for bouquets and another for vases, which you check once and reuse, rather than unchecked text generated per customer.
Should I send cart reminders by text message too?
Only to customers who have agreed to marketing texts, and only after the email sequence is working. Texts feel more intrusive than emails, rules on marketing texts are often stricter, and each message costs money. Shopify Messaging includes SMS, and so do most email platforms, so it is easy to switch on and easy to overdo.
Further reads
- How to Win Back Lapsed Customers With AI-Personalised Emails — The next automation after cart recovery: bringing back past buyers.
- How Much Does It Cost to Add AI to a Shopify Store? — What the email and AI tools around a store add up to.
- How Florists Can Prepare for Valentine's and Mother's Day With AI — Timing emails and stock around a florist's busiest days.
- What Can AI Do for Your Email Marketing? 10 Practical Uses — Ten other ways AI helps a small shop's email.
- How to Use Shopify Sidekick to Run Your Store Faster — Ask Sidekick to set up and report on automations.
- Your First 30 Days of AI Customer Support for an Online Shop — Handle the questions that reminder emails bring in.
- How Long Does AI Marketing Automation Take to Set Up and Pay Off? — Setup times and payback windows by automation type, a four-number payback formula and a delicatessen's first seven months, month by month.
- How to Use AI to Find Out Why Website Visitors Don't Buy — Combine your analytics funnel, free session recordings and exit-survey answers, then have AI rank why visitors leave, with an online bookshop worked through.
- How to Build an Upsell and Cross-Sell Email Sequence With AI — Find what customers really buy next, plan a short triggered sequence, draft it with AI from a proper brief, and stop it the moment someone buys.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Shopify help pages on abandoned checkouts, Shopify Messaging automations and Shopify Messaging email pricing; Klaviyo pricing page and help articles on abandoned cart flows.