Editorial
The Gist
- Term clarity. Machine learning and generative AI in marketing often get conflated, leading to confusion for brands evaluating tools.
- User focus. Generative AI for marketers mainly offers time-saving benefits, while machine learning primarily enhances campaign performance.
- Email significance. Machine learning algorithms have revolutionized email marketing optimization, from send times to subject lines.
The most popular classification models have generative AI being a subset of machine learning, and machine learning being a subset of AI. While this may technically and architecturally be the correct way to think about these tools, it has led to significant confusion.
Let's take a look at machine learning and generative AI in marketing.
Considering Machine Learning and Generative AI in Marketing
First, because of our natural tendency to drop words to simplify and economize our language, this has led to many people referring to generative AI simply as AI. Muddling things further, because generative AI and machine learning are both lumped under AI as a big umbrella, all of these terms are being used somewhat interchangeably.
All of this makes it difficult for brands to evaluate these tools and understand how to use them to accomplish their goals. After all, most people aren’t building these tools. They’re using them.
So, let’s consider a more user-centric framework, one that focuses on what brands — and in particular, marketers — care about. Let’s also focus in on just machine learning and generative AI, which comprise most of the AI tools that marketers use. From a user standpoint, the biggest differences between machine learning and generative AI are twofold:
- Where the data comes from that fuels the tool’s output.
- The primary benefit of the tool.
Let’s first examine how machine learning looks when viewed through these two lenses…
Machine Learning and the Impact on Email Marketing
Used by marketers for around a decade, machine learning fuels a variety of common email marketing features and applications, including:
- Send time optimization, which determines the best time to send email, SMS, and other digital campaigns to individual subscribers based on their past engagement times.
- Subject line optimization, which helps determine the best words and other elements to include in the subject line of a particular campaign based on the historical performance of campaigns using those subject line words and elements.
- RFM modeling, which uses the recency of a subscriber’s last engagement, the frequency of their engagement and the monetary value of their engagements to classify subscribers into groups, helping to determine the best contact and content strategies to use with each group.
- Fatigue analysis, which determines which subscribers are the most at risk of opting out if they are included on your next send based on the historical engagement patterns of subscribers who opt out.
- Product and content recommendations, which determines the best products or content to include in a campaign based on each subscriber’s past engagement with products and content.
- Channel selections, which determines the best channel or channels to use to reach each subscriber for a particular campaign or journey.
Where the Data Comes From
Machine learning algorithms are fueled almost entirely by your company’s data — whether it’s the performance of your campaigns, the engagement of your subscribers or other aspects of your business.
However, there are instances where machine learning results may be informed by data from other brands that also use the machine learning tool. For example, sometimes subject line optimization tools prioritize your subject line data, but also use data from other users to supplement. This kind of data supplementation works best when limited to users who are similar to your brand, such as in your same industry.
The Primary Benefit of Machine Learning in Marketing
Marketers use machine learning because it improves the performance of their campaigns. The return on investment is seen in higher engagement and conversions, and in lower opt outs.
At its core, machine learning helps marketers fulfill our long-standing goal of sending the right message to the right person at the right time via the right channel — and doing that with precision at scale.
However, when used for things like product and content recommendations, you could argue that machine learning also saves time. Indeed, many of our clients have been excited to redesign their campaigns so that such recommendations have a steady presence — or even constitute the entirety of a recurring campaign — so they save time on writing, designing and coding. But for the most part, machine learning allows marketers to perform tasks that would be time-prohibitive otherwise.
Generative AI in Marketing
Bursting onto the scene in the fall of 2022 with the debut of ChatGPT, generative AI in the form of large language models (LLM) and image-generation engines can help marketers with:
Learning Opportunities
Webinar
May
9
Driving Customer-Centricity in the Financial Services Industry
Discover how financial organizations are leveraging generative AI technologies.
Register
Conference
May
9
An Irish Roundtable in the Valley: The Economic Impact of AI on the World of Work Stanford 2024
Exploring the Impact of AI On the World of Work with Irish diaspora leaders and tech business champions from the North of Ireland
Register
- Copy, including background research, ideation, copy drafting, content gap analysis, changes in copy tone or style, subject line and headline iteration, and more.
- Code, including everything from debugging, refactoring and documenting existing code to writing new pieces of code or even entire projects.
- Images, including everything from background changes, image manipulation, and image upscaling all the way up to images created from prompts alone.
Where the Data Comes From
Today, foundational generative AI models like ChatGPT, Bard, and Midjourney are built on massive amounts of external data. For example, ChatGPT is trained on the Common Crawl, Wikipedia, and other public sources. (A number of lawsuits by artists, authors and communities have also been brought claiming they’re trained on copyright-protected works as well.)
But internal brand data can also be fed into these models. The most obvious way is when prompting these tools. Prompts can include brand-specific content, as well as information on brand style and more. Many brands have already created prompt libraries so they can better standardize the results they get when their employees use generative AI.
Beyond prompts, a small percentage of vendors and brands have already started building their own specialized generative AI models, where they train a foundational model on extensive amounts of additional content. This allows a model to perform narrower tasks more effectively or with a much stronger brand or industry focus.
The Primary Benefit of Generative AI in Marketing
Marketers use generative AI primarily because it saves time compared to doing those tasks manually themselves. However, even with specialized generative AI models, the output isn’t typically better than what a skilled person could create themselves. So, human review and editing — sometimes extensive — is almost always needed to ensure that quality is up to a brand’s standards, and that the voice of the brand is present and appropriate.
This is probably the biggest point of confusion. Perhaps because of the term “artificial intelligence,” some marketers assume that these tools are smarter than they are and, therefore, can make better decisions. For example, some think that generative AI can write more effective subject lines than they can. Then they A/B test subjects lines they’ve written against generative AI-written ones and are shocked when theirs win.
But, of course, theirs would win, because they have a solid understanding of what their audience responds to based on past results, whereas generative AI doesn’t have the benefit of any of that knowledge. It just strings words together based on what words people tend to put next to each other, not on how effective any of it is.
That said, specialized generative AI models can yield some performance benefits, particularly for resource-challenged businesses.
Related Article: 7 Irresistible Email Subject Lines
In the Future: Machine Learning and Generative AI in Marketing
Hopefully I’ve illuminated the key ways that machine learning and generative AI in marketing are different today. But the key word is "today," because the overlap between machine learning and generative AI in marketing will steadily increase going forward.
This is because more technology vendors or brands will be building specialized generative AI models. We’ll also see more of the largest companies create their own foundational generative AI models akin to ChatGPT and Bard.
For example, Oracle has created an generative AI model that writes APEX code, which both accelerates app development and effectively eliminates human error. And Amazon just announced the Alexa LLM, which is optimized for voice communications. Over time, creating such models will become easier and cheaper, and continually training on fresh data and content will become simpler, too.
But even as this happens in the years ahead, brands should continually be clear about the benefits they’re getting from their machine learning and generative AI tools, as well as their other AI tools that don’t fall into either of those two classifications.
Are you increasing performance or saving time? Or some combination? Because depending on your answer, you may need more than one solution to get all the benefits you want.
Learn how you can join our contributor community.
