The Past, Present, and Future of AI in Email Marketing


Not sure if you’ve heard, but Artificial Intelligence (AI) is kind of a hot topic right now - in literally every industry and even among consumers. If you’ve attended any of our Thursday Only Influencers (OI) Members-only Live Zoom calls this year, you’ve probably noticed that AI comes up on pretty much every call. (Quick plug - if you aren’t attending the weekly OI Live calls, you should! It’s a benefit for members of OI.) It’s probably safe to say that 2023 is going to be known as the ‘Year of AI.’ Probably the first of several years that will make that claim. 

As with a lot of buzzwords, AI is getting thrown around a lot. But there’s actually a bit of nuance to the topic that tends to get lost in generic conversations about how AI just suddenly arrived and is already disrupting industries. The thing is a lot of what gets included in the AI bucket isn’t actually particularly new. Just a few years ago, some of the capabilities being credited to AI were known as Machine Learning or part of some automated optimization tools and were readily available to marketers through a variety of marketing platforms.

While the definitions of AI and Machine Learning differ, the line between them in common understanding is fairly blurry, and for practical purposes in terms of marketing, they may often both be relevant. While the long-term goal of AI is to create a ‘machine’ that mimics human intelligence, today it is largely more related to machine learning, which focuses on teaching a machine to perform specific tasks rapidly and more efficiently than a human is likely capable.

So, as we all hop on board the AI bandwagon, it’s useful to realize that there is very much a past, present and future to discuss when it comes to AI and its impact on email marketing.

The Past - Machine Learning and Personalization

One of the areas that email marketers focus on with AI is on analyzing large datasets to make decisions on campaign optimization, segmentation, and personalization. However, none of these concepts are particularly new for email marketers. Various platforms have used machine learning to analyze this data and provide insights and campaign recommendations for years.

We are all likely familiar with receiving emails from e-commerce sites like Amazon that provide product recommendations based on past purchases. These types of emails rely on algorithms using large amounts of customer data to identify patterns in purchase behavior suggesting that when someone buys a particular product, they are then more likely to purchase one or more additional products. Putting those next likely products in front of the consumer via email makes it even more likely that the next purchase will be made. With a vast product catalog, there is no individual who could sift through all the purchase data to make these kinds of connections - especially since they are always shifting as new products become available and consumer purchase behavior evolves.

Similarly, these types of machine learning tools have analyzed data on past campaign performance by the time of day they were sent, what subject lines drove higher open rates, which call-to-action drive the most clicks and conversions, etc. Each of these insights is then used to make decisions about how to optimize future campaigns to drive higher performance.

Based on this broader definition of AI, it’s clear that email marketers have been using it for years. Email recipients are entirely familiar with receiving this type of content in their inboxes, even if they wouldn’t necessarily have attributed an of it to an AI. This type of AI has already had a powerful impact on email marketing but until fairly recently, it was mainly focused on more traditional analytical processes.

That began to change in the last couple of years as various companies rolled out content creation tools based on AI. From my own experiments with these early content creation platforms, I found them useful for very specific types of content, but also not ready for primetime. That changed with the release of ChatGPT.

The Present - Generative AI

ChatGPT! MidJourney! Bard! The new Bing! We’re clearly living in a pretty exciting present when it comes to AI. For the most part, when people talk about the massive explosion of AI since the beginning of 2023, they are actually talking specifically about Generative AI.

Generative AI is a specific type of AI system capable of generating text, images, and other media (video, etc.) as directed by prompts provided by the user. Generative AI systems are trained on an existing set of data (often some selected portion of publicly available data on the Internet) and they then use that information as the foundation for their responses to user prompts and requests.

Generative AI varies from ‘traditional’ AI in that it focuses on creating new content based on an existing data set of previously existing content. More traditional AI is focused on using large data sets to identify patterns, provide analytics, and make decisions or suggestions based on this data. One simple way to describe the difference is that traditional AI might analyze data on a past email campaign’s performance and recommend a specific segmentation strategy to test, based on historical data. Generative AI would provide the content to be included in that email campaign.

Today, seemingly everyone is at least experimenting (ok, maybe just playing around with) generative AI tools like ChatGPT or DALL-E. Email marketers are certainly in the mix, using ChatGPT to write email copy and brainstorm subject lines or leveraging tools like MidJourney to create customized graphics. Results are predictably mixed since the output relies entirely on how well the user writes a prompt to request a certain type of content to be created.  In fact, prompt writing might just be the most sought-after ‘writing’ skill for marketers at the moment. The potential for these tools to make content creation faster, more efficient, and potentially more effective has everyone’s attention - consumers, marketers, and legislators.

As with any emergent technology, it raises almost as many questions and challenges as opportunities. Here’s an interesting question. Who owns the content created by generative AI? When you prompt ChatGPT to write copy for a new email campaign or use MidJourney to design a graphic for your email creative, who owns the rights to that content? You might want to familiarize yourself with the user terms and conditions for any generative AI platforms you use in your marketing programs. Currently, OpenAI’s Terms of Use assigns all the rights to the user related to the output they create. MidJourney also asserts that paid users own all assets they create through the AI platform, but also retains the right to use any content created on the platform for a variety of purposes at the company’s discretion. From a legal perspective, there are already multiple lawsuits in motion against various generative AI companies with regard to the data it references to generate content. The ultimate question of who owns the intellectual property rights to AI generated content is still unsettled.

So, our AI present is a predictably mixed bag of excitement, concern, and speculation about what comes next.

The Future - ?

Predicting the future is always a bit of a fool’s errand. How many articles have you read over recent decades that predicted the end of email as a communication and marketing channel? I lost count years ago. The point is that forecasting the future is hard. But let’s take a look at where AI (as a whole) may take email marketing in the years ahead.

It certainly seems feasible that the next generation of AI tools could combine aspects of traditional AI and generative AI to provide something even more full-featured. Imagine an email campaign that would first rely on AI to analyze all past campaign data to develop an overarching strategy, along with specific tactics. It matches products and offers to email recipients on a 1-to-1 basis, predicting what offer, product, and message should be sent to each member of your email list, along with when it should be sent to drive the highest ROI. Next, it dips into the generative AI arena to produce individualized email copy for each recipient and does all of this in real-time.

That all seems very much in the wheelhouse of next-generation AI tools. At the same time, just because something is possible, doesn’t mean it’s actually a good idea. Hyper-personalization is a great example. There’s logic behind a process of true 1-to-1 personalization, but will it actually hold up when performance data is analyzed? We already know it's possible to overdo it on personalization when it seems to veer into creepy territory and consumers start to feel like marketers know them just a bit too well. Will the ability of AI to make this type of intensive audience segmentation and personalization not only possible but potentially lead to marketers overstepping into the creepy zone? Probably.

It may be that we will learn that the potential of AI may go beyond what are actually the most effective marketing strategies and tactics. Only time will tell. One thing I can predict - it’s going to be an exciting ride.

wyron a GY38n9WKjQI unsplash 600Photo by Wyron A on Unsplash