13 June, 2026
AI email marketing: 3 ways to personalize and keep past customers longer

After over 14,000 projects, MONA has met many shops that already have a customer list but still pay for ads every month to reach those same people again. AI email marketing helps you use the data you already have to write the right content, segment the right groups and send at the right time, before you think about buying more reach.
Email is worth doing first because these customers chose to leave their address, have viewed a product or have already bought. You can absolutely set up a basic flow in about 2 hours using a spreadsheet, an email sending tool and AI to write the content. Most sending tools today give you the first 500 addresses free, as long as you follow the order in this article. Use your old data first.
AI email marketing turns an old list into a cheap channel for nurturing customers
A shop that comes to MONA usually already has 3 places that hold emails: the offer sign-up form on the website, order data, and files saved by the sales team on their own. While those 3 sources sit idle, the shop pays to store the data and keeps buying ads to bring past customers back to the site. Email is cheap in distribution because you do not have to bid again for every click with an ad platform. The main costs are the sending tool, the time to standardize data and the effort to create content relevant enough for people to open it.
AI email marketing works best once the shop has clean data. An email address with no name, no purchase history and no recent interaction date only lets you send mass mail; add those 3 fields and the system starts to know what to say to each person. We usually advise shop owners to measure 4 numbers before sending the first campaign: valid emails, open rate, click rate and the number of orders tied to email. Measure again after 30 days and you will know which change actually sold products.
| Channel | Pay to reach them again | Data the shop owns | Best used for |
|---|---|---|---|
| Retargeting | Have | Partly | Bringing customers back fast |
| Social media messages | Depends on the platform | Partly | Short reminders that need a quick reply |
| Low once you have data | Higher | Nurturing, educating, selling again |
If your list is under 500 customers, go ahead and do it yourself and learn from real feedback. Do not buy external email lists, because recipients have not agreed to hear from your shop, and the spam complaint rate will damage your whole sending domain. Do not buy strange data.
AI changes three stages: writing content, segmenting lists, choosing send times
In a manual campaign, the writer usually drafts one version, selects the whole list and schedules one common time for everyone. That is quick when the list is small, but a new subscriber, a recent buyer and a customer who has been silent for 90 days all get the same greeting. AI changes the writing stage by producing multiple drafts based on the products, behavior and brand voice the shop provides. The person in charge still reviews prices, offer conditions and links before sending, and drafting time drops noticeably because nobody has to face a blank page for each group.
The second stage is segmenting the list. AI email marketing reads the data fields to suggest groups with similar behavior, for example people who viewed a product twice without buying, steady repeat buyers, and people who have not returned in 60 days. The third stage is send time. Instead of forcing 10,000 recipients to receive mail at 8 a.m., the system uses open history to place each cluster in the time window when they usually read, then feeds the numbers back for the next send.
| Stage | By hand | With AI support | What the person in charge still decides |
|---|---|---|---|
| Content writing | One version for the whole list | Multiple versions by group | Prices, offers, brand voice |
| Segmenting | Spreadsheet filtering | Group suggestions based on behavior | Conditions for entering and leaving a group |
| Send time | One common time | Learns from open history | Quiet hours, frequency |
We do not recommend handing sending rights straight to AI in the first week. Let AI produce drafts and suggestions, while one accountable person reviews the first 20 to 30 emails to lock down voice, facts and promotion limits. Lock sending rights first.
Prepare your data before running AI email marketing
At a fashion shop selling both online and in store, we once received an export file of over 20,000 rows, but emails were duplicated, returned orders and repurchase dates were recorded in 3 different formats. We spent 1 session consolidating it into 7 required columns, flagged 2,300 duplicate or wrongly marked entries separately, and only then built the sending groups. That was slower than opening AI to write subject lines, but for the whole following month the shop never sent repurchase content for an order that had been returned. Clean data first, then write.
Step 1: Collect data and keep proof of consent
Export data from the website, sales software and forms into CSV files, then merge by email address. Every row needs a collection source and the date the customer agreed to receive messages, so operations can handle it correctly when a customer asks to stop. Do not add people who messaged to ask a question but never signed up for email to the marketing list. Put an unsubscribe link in every email and test it with an internal address before the real send date.
Step 2: Clean up before giving it to AI
Remove duplicate addresses, fix extra spaces, drop emails that repeatedly bounce and standardize date formats. For canceled or refunded orders, add a separate label so AI does not put those customers in the group that just bought successfully.
A clean 2,000-row file is more useful than a 20,000-row file mixed with dead addresses and missing data. We are quite firm at this step because wrong segmentation leads to wrong content, wrong timing and wrong revenue reports. Bad data, stop sending.
Step 3: Write the group rules in plain Vietnamese
You do not need a complex model yet. Write rules clear enough to read, like “viewed the product at least 2 times in 14 days, no order yet” or “bought, has not returned in 45 days.” Conditions readable by eye let the operator check any 10 customers before turning the flow on. If all 10 records have the right reason for being in the group, the data is good enough to move on to content.
How to use AI to write sales emails while keeping your shop's voice
Many shop owners open an AI tool, type “write a good sales email,” and are disappointed because the result sounds like every other brand. We handle it with a short brief that includes the recipient, their behavior, the product, the reason for sending, the offer limits and how the shop usually talks. AI email marketing needs more specific data than praise for the product, because the line “the customer viewed the blue shirt 2 times in 7 days” gives a more useful draft than a long paragraph about a professional, dedicated, high-quality brand. The brief must be real.
Use this prompt template right away, replacing the parts in square brackets with real data:
Write an email for the group [viewed the product twice but did not buy]. The product is [product name], priced at [real price], offer [real conditions]. Friendly voice, short sentences mixed with long ones, subject line under 45 characters. Do not add benefits, prices or deadlines on your own. End with one button to view the product again.
For customers who have already bought, change the goal from closing the sale to usage guidance and suggesting add-on items:
Write an email to send seven days after an order. Mention exactly [the product purchased], give three usage tips found in this document [paste document], then suggest [add-on product]. Do not create discount codes. If facts are missing, mark the spot for the reviewer.
For the group that has not returned in a long time, use a shorter prompt and avoid pretending to be close:
Draft an email for customers who have not bought again in 60 days. Remind them of the category they once cared about, introduce three new products from the list [paste list], and keep the content under 180 words. Create two subject lines to test, and do not use pressuring language.
After receiving the draft, the reviewer reads 3 things in order: product facts, offer conditions, then voice. We usually read twice in 15 minutes, the first pass checking only numbers and the second fixing the voice. A wrong fact means stop; an off voice can still be fixed, but sending the wrong price to the whole list means handling complaints.
Run your first campaign yourself in seven steps
A cosmetics shop in District 10 once gave us a list of 8,000 emails that had been inactive for nearly 2 years. We cleaned the list, split groups by last purchase and re-ran the nurture flow for 3 weeks. 41 orders came back from that old group within 3 weeks, enough for the shop to drop its habit of blasting one discount email to the entire list at the end of each month, and keep this nurture flow as its default channel. Start with a small group.
Follow these seven steps:
- Pick one group with a clear condition, such as customers who bought 45 days ago and have not returned.
- Set one single goal: repurchase, product view or completed cart.
- Use AI to create two subject lines and one email body from a brief with locked facts.
- Send a test to Gmail, Outlook and shop-domain email addresses.
- Check how it displays on a phone, the buttons, the landing page and the unsubscribe link.
- Send to 10% of the list first, and watch bounces and replies for 24 hours.
- Send the rest when the data looks fine, then reconcile orders after three to seven days.
Do not change the subject line, content, offer and send time all at once. If 4 variables change together, you see results go up but do not know which factor to keep for the next campaign. We usually test 2 subject lines on 2 equal groups, keeping the body and send time unchanged. After one round, keep the subject line with the better open rate, and only then test the call-to-action button or the way the product is presented.
| The | What it tells you | If it is low, check first |
|---|---|---|
| Delivery rate | List quality | Wrong emails, full inboxes, sending domain |
| Open rate | Subject line and send time | The promise in the subject line, frequency |
| Click-through rate | Content and buttons | Product, price, link |
| Orders from email | Real ability to sell | Landing page, cart, checkout |
| Unsubscribes | Relevance | Segmentation, frequency, source of consent |
AI helps read response patterns quickly, but the decision to keep or drop a campaign must rest on orders and customer reactions. An email with a high open rate that produces no action still needs fixing. Change only one variable.
Use email to keep customers, and do not turn the inbox into an ad board
At the shops we have reviewed, the send calendar usually has only sale days, product launch days and month-end days. Customers keep receiving invitations to buy but lack usage guides, warranty reminders, tips for choosing products or information that helps them avoid buying wrong. An easy nurture schedule is 4 emails over 30 days: confirmation and guidance after purchase, a check-in on how it is going, suggested related content, and only then a repurchase offer. After 7 days, read the real responses before lining up the next email.
If you are not sure where the line falls between nurture email and retargeting ads, the article what is remarketing explains how a person is recognized again after visiting or interacting. Email is one channel in that process, alongside Zalo, Messenger and ads. We like email because the shop keeps its send history, content and response data inside its own system. The price is that the operations team has to look after domain quality, consent and steady frequency. Do not send only sales.
When doing it yourself starts to break: large lists and multiple channels
A spreadsheet still runs fine with a few hundred or a few thousand records if one person manages it and there are 2 customer groups. When the same person exports orders from the web, filters files, imports into the email tool, then reconciles Zalo and Messenger, the delay grows longer than the moment the customer wants to buy. AI email marketing cannot fix data that sits apart in 5 places. If order status lags by more than 24 hours or the same offer is sent twice, the shop needs to recognize one customer across email, phone number, orders and conversations, and stop every old reminder the moment that person buys.
Signs that you have outgrown doing it yourself include:
- A customer buys and still receives an abandoned cart email.
- One offer is sent repeatedly across three channels on the same day.
- The sales team does not know which email a customer just clicked.
- The open rate report has numbers, but the order report cannot trace back to the source.
- Every campaign run needs manual file merging by two or more people.
This is when you need to understand email marketing service will cover which part and which part the internal system must cover. We advise keeping your current tool if it still sends reliably, and first fixing the data connection and stopping the wrong scenarios. Connect the data first.
MONA connects AI marketing directly to your online store
MONA has built websites since 2016, across over 14,000 projects, and keeps 85% of clients coming back, so we usually look at email starting from order data before looking at the email template. The website must record product viewers, cart abandoners, buyers and payment status clearly enough for every scenario downstream to run correctly. With a system for proper online store development, data generated on the web flows straight into the customer record, AI drafts content under rules you have approved, and email stops by itself once the customer completes the order. You see one line from opens through to revenue instead of stitching reports together by hand at the end of the month.
We will also be blunt: a small shop with a compact list and only 1 sending channel does not need dedicated software yet. Run 3 to 5 campaigns yourself, note where the effort goes, and invest only when the repeated volume is large enough for a system to give real time back. If your list is already large, your data is scattered and customers receive mismatched content across several channels, AI email marketing needs to be reconnected from the online store. Call MONA at 1900 636 648 so we can review your current flow and agree on exactly what needs to be done. Small shops, keep running it yourself.
Related articles
Professional website
design services
Own a website with a beautiful, 100% exclusive design and a lifetime warranty, with the ability
to expand features flexibly as your business grows, starting today!































Free AI courses
AI tools at work
VI
EN


