Limitations of AI in Marketing

You may have heard that AI is the Next Big Thing for marketers. AI does indeed have the potential to transform the way that marketers work. AI can help improve efficiency by allowing marketers to automate their more routine business processes. This frees up valuable employee time to allow them to focus on more creative, less mundane, endeavors.

However, it is important that marketers not rely too much on AI. There’s a tendency to think about AI in somewhat mystical terms. People think of it as a magic wand that they can simply wave and all their sales and marketing will be done for them. Anyone who’s actually used AI-based marketing can tell you that there is a lot more to it than that.

As powerful as AI is, it is not going to replace human marketers any time soon. Let me give you an example. Like anyone who runs a business, I get a lot of sales emails for products and services that I usually do not need. Typically, I scan them and then discard them. But the other day, I got one that caught my attention, but probably not in the way the sender had intended. There was something odd about the way the email was worded.

It wasn’t that the text sounded like it was written by a non-native English speaker, or that it had a bunch of typos. The content of the email was entirely coherent, but somehow, it was wrong. Each sentence on its own made sense, but taken together the sentences did not really seem to add up. The more I read the email, the more I was sure that it was not written by a salesperson at all. This was an AI-generated email. Needless to say, the company that sent the email did not convert me into a customer.

Spending on AI-based marketing technology will double by 2025, IDC estimates. It is vital for marketers to think critically about their use of AI, and approach it as they would any new tool. They need to know what it is good for and what it is not good for. If marketers are not careful, there is a real danger that AI could lead to an enormous and seemingly uncontrollable increase in sales and marketing communication.

Call it the “spam-pocalypse.” 

Limitations of AI

As marketers begin to explore AI tools, they will quickly notice that the technology, given enough information, is excellent at mimicking human speech. This is thanks to natural language processing (NLP). But research shows that NLP lacks common sense.

For example, researchers tested the ability of state-of-the-art NLP to write a sentence using the words “dog,” “frisbee,” “throw,” and “catch.” The NLP model generated the sentence, “Two dogs are throwing frisbees at each other." Now, that sentence is coherent, it’s grammatically correct, and it does not have any typos. But even a child could tell you that dogs do not throw frisbees.

Like the email I received, that sentence demonstrates the pitfalls of relying on AI as the solution to all of your sales and marketing challenges. According to AI researcher Bill Yuchen Lin, even the strongest NLP model, after training with a large dataset, at this point “can still make silly mistakes.”

It is tempting to use AI to write or personalize your content and indiscriminately blast out messages to prospects, customers, partners, publications, etc. But there can be significant ramifications to such a blanket approach. In my case, I was just mildly annoyed: I immediately deleted and disregarded the sales email. There was no long-term damage to the vendor who sent the email.

But left unchecked, AI can unintentionally create text that is discriminatory, even offensive. That’s going to do more damage to a brand than just a poorly targeted pitch. According to Wired, “the tech industry is scrambling to constrain the dark side of a technology that’s shown enormous potential but also can spread disinformation and perpetuate biases.” The AI-based language generator GPT-3 has certainly improved over its previous versions, but there is still a long way to go before AI-generated language is consistently reliable.

AI is a tool, not a panacea

These risks do not mean you should eschew AI altogether. Especially in the age of Big Data, AI can execute tasks more quickly and reliably than employees can. For example, AI can sift through piles of data to extract relevant trends. AI can help marketers improve customer segmentation, tighten up copy, create A/B test emails and subject lines, and even draft blog content. And despite its limitations, AI can still be an excellent tool for writing sales emails and managing customer service interactions.

Someday soon, AI will likely be a lot more reliable and hands-off. Currently, the number-one use for AI is chatbots that allow customers to get faster answers to their questions. Today, more than two-thirds of chatbot conversations are handled entirely by AI and do not require any human intervention. That is up from 20% in 2017.

Despite the limitations of AI, marketers should definitely look into how AI-based marketing tools can help them improve various business functions. In fact, those who do not embrace AI as a powerful marketing tool run the risk of being left behind the rest of the industry. If we as marketers can all approach AI with a strong sense of what the technologies can and cannot do, we might just collectively avert the spam-pocalypse.