Email marketing (and SMS) are your company’s ATM machine. They’re consumers and clients that gave you their information, have shopped already, and who are familiar with your brand, which is why these campaigns convert better than other channels. But when you over-email, use the wrong messaging, or the offers are not relevant to the recipients, your company gets marked as spam and customers unsubscribe. This is where AI may be able to help.
Instead of guessing what each subscriber wants, let your favorite AI system help you maximize your return on investment and increase your revenue with email marketing. Below you’ll find five ways AI can help sort through data and some prompts. You can use these starting today with ChatGPT, Claude, Perplexity, or whichever LLM system you like most.
- Improved subject lines by user and service provider
- Which body style to use
- The offer to send by who shops for what
- Which products to feature by which segment
- The products to promote based on what clients buy by region
Note: Before you begin, check with your lawyer to make sure your privacy policy and website terms of use allow you to use a third-party tool for sorting data and analytics.
Subject Lines and Service Providers
Each email service provider (ESP) has different character limits that display in the out-of-the-box setting for both desktop and mobile devices. On average, you’ll want to keep the subject lines under 50 characters regardless of Gmail, Yahoo Mail, iCloud, or Outlook, especially if most of your audience is using mobile phones to open and engage.
Where AI can help is by dividing up your email list by service provider and seeing which subject lines resulted in the best open rates based on commonalities like character count, words, questions vs. sales pitches vs. announcements, etc.
Start by uploading your email lists and asking the system to separate the email list into four groupings. Then export it into individual spreadsheets.
- Gmail
- Yahoo
- Hotmail
- Other
Next, upload the analytics from the last five email campaigns you sent and ask the same AI system to analyze what the subject lines had in common when it came to the ones with the best open rates and the most clicks by ESP. Open rates can be skewed as the person may not have actually opened, so adding a click as a secondary requirement can give better data.
You can tell it to include words, character counts, or the type of subject line. Now, share what your next email campaign is going to be and have it make a recommendation for each list based on what will get the most opens and clicks.
Here’s what the prompt series could look like.
Prompt 1:
Please take the attached spreadsheet called “email addresses” and break it into four separate spreadsheets: all emails ending in gmail.com in spreadsheet 1, yahoo.com in spreadsheet 2, hotmail.com in spreadsheet 3, and everything else in spreadsheet 4.
Prompt 2:
Now take the attached spreadsheet called “email campaigns,” where the columns are subject lines for five email campaigns we sent and the rows are the email addresses from the first spreadsheet that opened the email. Tell me what the subject lines have in common that got the most opens by email service provider — including character counts, specific words, tone, voice, and messaging that may have influenced open rates.
This is easy to modify with information like the most clicks by email provider or the most clicks with conversions, total unsubscribes and what to avoid sending by email provider, or whatever metric is important to you. Some email providers have audience demographic skews, and this will help you cater the subject line to that group.
Now that you have an idea of what to send by email service provider or as a whole, you can use the separated lists from the first request and test the subject lines to see if you get better results.
The Body Style of Your Emails
Another way to use AI for email marketing is to reduce wasted sending. You can split test the style of the body to know if a shopping grid, a list format with paragraphs, text only like a letter, or any number of combinations are best. While a split test can give you important information, you have to wait for results, and you lose 20% or 30% of your list from the test. AI can help predict what will work and by subscriber.
If you know the next promotion and are stuck choosing between one or two body styles, upload the last five or ten of both styles and the total number of clicks and sales that came from it. AI can tell you which style of email body drives more revenue, and depending on the version, it can export which to send to who. The prompt below will include the type or promotion if you aren’t sure what to send next and are looking for ideas.
The prompt can look like this.
Please look at the attached spreadsheet where the first column is the metrics $ in sales, total number of clicks, total number of opens, and the type of email we sent including bundle deals, discounts, product launches, and flash sales.
The next columns are the past 40 email campaigns we sent and they’re named for the style of the body (i.e., shopping grid, list style, large image with text, newsletter style). The cells show the total aggregate number of the metrics for the email campaign.
Please tell us which email template does best by the type of promotion, meaning bundle deals, discounts, product launches, and flash sales. Our goal is to know when we should use which template, or if a hybrid like a product grid with bundles makes the most sense.
The goal here is to get the AI system to tell you which template to use for the email type. If you have a sale coming up with a flat percentage off or tiered percentage off but no bundles, maybe a product grid can work. For a product release with a bundle, a large image with text below could be best.
Pro-tip: If you’re an advanced AI user, try segmenting the email template with the metrics and divide it out by email service provider. It may turn out Gmail users prefer list-style newsletters for discounts while outlook or iCloud users may prefer large images with text below them.
The Type of Offer to Send by Customer
AI may be able to help you determine which customers should get what type of offer. Some people respond to “20% off on purchases of $100+,” while others may respond to “save $20 on all orders over $100.” If the average order value doesn’t change and is right around the $100 mark, there’s really no difference other than catering to the individual’s preferences. AI can make this easy.
Upload a spreadsheet where you have the email addresses in column 1, then columns 2 – 6 have five email campaigns where you did a percent off and columns 7 – 11 feature where you did a dollar amount off. Now have AI sort the email addresses into two groups by who shopped more often with one or the other and add a third group where it doesn’t matter or where they’re unlikely to respond to either message.
The same sorting can be done for other deals including free shipping versus a percentage off, bundling where if you add a third product you save, or where you get a free gift with purchase. There’s no shortage of ways to help prevent sending email campaigns with a deal that does not match a user’s preference. You may also decide to ask for who not to email and exclude them if they’ve never engaged with the type of offer.
The prompt could be phrased like:
Please look at the spreadsheet I uploaded where the email addresses are in the first column, columns 2 through 6 are where we sent a percentage-off deal and columns 7 through 11 are where we offered a free gift with purchase. The cells are the amount of money each customer spent and, if it is empty, it means they did not shop. Please create three separate lists where each email address can only appear on one.
The first list is for customers that spend the most when the deal is a percentage off. The second list is for customers that spend more when there is a free gift with purchase. The third list should include anyone that did not shop more than twice.
Sending Offers and Products Based on User Preferences
A fourth way to use AI to optimize email campaigns is by uploading SKUs purchased with email addresses and conversion data from your email campaigns.
AI can do a quick analysis and sort the subscribers by people that (for example) purchase blue products but not green so they can be excluded from campaigns with green products, or that purchase above $200 but not under $50, or only when something purple is included.
The prompt here could be the following.
Please take the uploaded spreadsheets named “purchases by customer” and “SKU data” and analyze the SKUs purchased by email address.
The spreadsheet “purchases by customer” contains customer email addresses and the SKUs they purchased. The spreadsheet “SKU data” contains all SKUs from the first sheet, with column A as product type, column B as price point, column C as color, and column D as size.
I’m sending an email campaign featuring blue fitness products priced at $100, available only in small and large. Please create a list of all email addresses that have purchased products in this color, product type, and price range in either small or large sizes.
You can take this further by building multiple lists — for example, customers who shop purple under $50 in medium, or those who buy brown and green in small and large and blue in medium and large, consistently spending between $50 and $100.
What People Purchase by Region
If you’ve tried the above prompts, you’re now moving to an advanced level of prompting for email marketing. This is where it gets fun, and we can combine multiple prompts and strategies from above.
Take some of the spreadsheets and add information like shipping zip code, email address, SKUs, and the deal or offer the customer responds to. On top of this, add the month of the purchase or a range like Q3 with multiple purchases.
The goal here is to have AI determine which products as groups should be featured by region, especially with seasonal purchases. In the winter, it is cold in the Northeast but sunny and warm in Florida, Arizona, and Southern California. This would automatically make product mixes like t-shirts versus sweaters more relevant for those regions. There could also be trends. Let’s say people in the Pacific Northwest want free shipping, while the mid-Atlantic customers enjoy a percentage off.
Build a prompt that matches your spreadsheet and guides the AI to know what is in each column and row, what the data in the cells are, and what you expect from the output. Then add something like this into the prompt:
Using this data, build regional email lists and identify shopping trends that are meaningful and actionable. Break regions out by states matching the U.S. Census Bureau’s geographic divisions (reference: https://www2.census.gov/geo/pdfs/maps-data/maps/reference/us_regdiv.pdf — divisions are shown on page 2). For each region, recommend the offer type with the highest conversion rate, the best-performing product SKUs, and any additional insights the data supports.
If you’re cleared to use AI to divide and make recommendations on your email marketing strategy, it can help take the guesswork out of maximizing your ROI. With the right prompts you’ll be able to divide and segment lists, customize products and offers to the customers they resonate with, and know when and who to send a campaign to.
National Funding does not provide tax, legal or accounting advice. This material has been prepared for informational purposes only. You should consult your own tax, legal and accounting advisors.
