How Generative AI and Email Work Together

December 29, 2023
Jump to...

    The AI bandwagon is a crowded one. Programs like OpenAI’s ChatGPT, DALL-E, and Google’s Bard have captured the world’s attention and spurred every type of company–from cosmetics giants to recipe apps to news sites–to implement generative AI, in some form or another.

    Email marketing–the oldest digital marketing channel, and a tried-and-true one at that–is due to hop on, too. But the ways a decades-old channel with a proven ROI will benefit from a technology that is, by all measures, still in a stage of experimentation and trial and error, has yet to be fully seen. 

    Here’s how generative AI differs from the type of AI already common in targeted email advertising–machine learning–and how email marketing might make use of generative AI, beyond just as an inspiration-and-headline-creating copy tool.

    With an adoption rate steeper than smartphones or tablets, generative AI has become one of the fastest-adopted technologies in recent years. It also set the record for the fastest-growing user base in the two months following its launch. Some are comparing it to what the early development of the web was like–and consider it similar in its potential to revolutionize the way we live, work, and play.

    Since ChatGPT’s launch in November 2022, its user base has jumped to over 180 million–that’s an 80% increase over the preceding eight months, and it continues to bring in new users drawn by the surrounding buzz. Generating 1.6 billion views on the website in June 2023 alone, ChatGPT is just one of many examples of AI out in the field, as companies continue to develop their own chatbots and iterations of the technology.

    Adoption of generative AI in the workplace has been incredibly quick, too, finding uses in the IT, sales and customer service, and marketing and communications departments within companies, according to the Capgemini Research Institute. Within professional settings, it’s become a relevant tool to help with research and creativity.

    As companies continue to experiment with internal and external tools making use of generative AI, governments seek to figure out how to regulate the new technology, and individuals continue to play with the tool and test its abilities and boundaries, email marketing is bound to witness its own development from a semi-personalized approach to advertising based on heaps of data to an even more fine-tuned, tailored, and effective channel.

    Generative AI is pushing marketing close to the dream of one-to-one personalization and faster, more efficient campaigns. And with the global number of email users set to reach 4.6 billion–over half the world population–by 2025, the relationship between email and generative AI is set to be an impactful one.

    Machine learning has been used in email marketing since at least the early 2000s (initially as a way to mark emails as spam or not-spam), while generative AI has found its footing in the medium primarily in 2023, firstly as a tool to help generate headlines and copy useful for multivariate and A/B testing–though always with the oversight of a team of humans. 

    Machine learning and generative AI are both subsets of AI (Artificial Intelligence), yet they have distinct purposes and capabilities. Machine learning, for its part, has been around a lot longer, though its primary purpose is to parse data and help with the segmentation and targeting parts of email marketing.

    Here’s how machine learning formed the basis for today’s email marketing: after achieving sophisticated levels of spam-filtering in the mid-2000s, marketers began employing machine learning to segment audiences and with that information, to  create rudimentary targeted campaigns sent to consumers based on their preferences and demographics. 

    Further developments in machine learning and predictive analytics allowed marketers to optimize send times and frequency in the late 2000s–and now, marketers can also predict conversion rates and make suggestions for how to personalize content.

    Natural language processing tools (like the kind that underpin current generative AI engines) began complementing email marketing in a rudimentary form in the 2010s, and they continue to be used–and improved upon–to facilitate personalized email marketing.

    Machine learning, focused on reading data sets and making predictions to provide the insight needed to personalize emails, is all that data that is underlying what tech-forward marketers hope to achieve–auto-generated, one-to-one personalization. That is, in combination with generative AI–which will ideally create moving, novel, and creative copy that resembles human talent. 

    While machine learning’s main focus is making predictions based on existing data and applying this knowledge to various ends and make recommendations, generative AI, new to the game, is the use of algorithms and machine learning to generate entirely new content, solutions, and ideas that resemble human creativity. 

    Generative AI can understand and mimic human behavior–and as email seeks to create more personalized content, it can fasttrack one-to-one email personalization at scale. 

    Between 2022 and 2023, almost twice as many marketers worldwide have noted AI as forming a critical part of their email marketing campaigns. According to a study from Insider Intelligence, “enhanced personalization algorithms and recommendation engines” are of interest to 51% of worldwide marketers surveyed, followed by “automated content creation and copywriting” at 47%.

    The future of generative AI promises the ultimate in marketing: one-to-one email personalization. This goes far beyond the practice of personalizing emails with a subscriber’s name–and even beyond including personalized product recommendations.

    Personalization means a lot of things: it could be location-based targeting, the creation of user-specific products, and even certain images used within an email. The hope is that generative AI can create personalized copy and design according to an individual’s preferences, based on their data. This would be done at scale, so human oversight would be nearly impossible to implement. Campaigns would have to rely on generative AI exclusively and hope that emails are error-free, correct tone-wise, and generally appropriate for each customer.

    But while machine learning, in its current form in use in email marketing, provides nearly all the insight necessary to be able to achieve mass amounts of personalization–even able to tweak copy and headlines as needed for specific populations–it can’t write perfect copy for every person. 

    While multivariate testing is still manageable with oversight (and additionally allows campaigns to lean toward the copy that is more effective), one-to-one personalization can’t be tested beforehand, since the copy would ideally be different for every single individual.

    Generative AI is still in its development stages. There is still a way to go to see how AI will be used and regulated in the future. Beyond the inability to individually oversee the copy of every email, there are other reasons we can’t reach one-to-one personalization yet:

    The disappearance of cookies, increasing state regulations, and further tracking challenges are making gathering sufficient data on an individual increasingly difficult. Before even considering a one-to-one approach to personalization, marketers need to figure out how to collect privacy-compliant consumer data in a way that consumers are comfortable with, that complies with regulations like the GDPR in Europe and state regulations in the United States, and that browsers allow.

    There’s a balance between the right amount of personalization and targeting, and an amount that is too much

    According to a Twilio survey, 46% of business leaders surveyed felt like brands were doing an “excellent job at providing personalization” in 2022, while only 15% of consumers felt the same way.

    While consumers may be comfortable with some level of personalization, there’s a threshold past which they may have data privacy concerns.

    Not only does extreme personalization create lots of extra work, but small tweaks to a campaign can also be effective in increasing ROI, too. 

    With one-to-one personalization, omnichannel attribution becomes nearly impossible–as there’s nothing to measure a single set of email copy against. 

    We have to face it–generative AI isn’t perfect, yet. A McKinsey study found that over half of those surveyed said the inaccuracy of generative AI “was a risk relevant to their company.” However, only a third believed that “their company was working to mitigate the risk.”

    Today, generative AI functions wonderfully as a tool for creative inspiration to generate headlines, body content, A/B testing, and more. It can cut down on work time for copywriters and designers, and can even be used to generate images with human oversight.

    Companies like Phrasee and Mailchimp are experimenting (or already have) tools that implement generative AI in their arsenals, but they are still programs that require a human to oversee them. 

    Beyond a research and brainstorming tool and an effective assistant in creating first drafts of copy, generative AI still has a lot of development ahead of it. And for a lot of tasks that involve branding guidelines, style sheets and certain tones associated with a company, as well as coding and planning tasks, it’s much more effective to employ a human instead of a machine.

    Digital marketing’s oldest channel and its newest tool still have a promising relationship to explore. While there are high hopes for the two and a future of one-to-one personalization potentially on the horizon, for now, there is still a lot to test and explore.