The latest season opener of the science fiction show “Black Mirror” depicts personalized content as a uniquely modern form of torment. 

In the episode “Joan is Awful,” a tech executive discovers that, thanks to an advanced machine learning algorithm, a streaming company has adapted her life into its newest hit series. The streaming service’s AI tool has actually produced a dramatized version of everyone’s lives—captivating and infuriating viewers with a personalized retelling of their awful behavior on tech-altered display to the whole world. 

The horror in this episode comes about because AI delivers a personalized but imperfect vision to the user. In the real world, we’re still in a place where content personalization gone wrong is, for the most part, barely inconvenient and almost funny.

However, “Joan is Awful” raises valid questions for digital marketers. In today’s hyper-accelerated AI-driven moment, it’s not too soon to anticipate damage to a brand that is overly reliant on AI to deliver personalization to customers. Letting that unfold might just make your customers feel downright “awful.”

When it comes to adding AI to your personalization program, please heed these three pieces of advice:

Start slow and small

Customers prefer personalized content—it’s simply more effective and impactful. It also drives better business outcomes for brands. But at its current scale, personalization is still imperfect and employing it can also go south quickly. 

In 2019, years before the AI gold rush, a company called Mother’s Lounge, a wholesaler of mom and baby products, launched a marketing campaign in which women received postcard-style advertisements congratulating them on their pregnancies. The problem: Several dispatches were to recipients who were not pregnant. After the mishap, the chief strategy office said, “people are very concerned when they get this because they want to know what list they are on that says that they’re pregnant.” A fair question.

Worse in 2012, Target had assessed based on digital cookies that a teenage girl was pregnant—before her father knew. Target sent the girl coupons that would appeal to a pregnant woman, determined to use what it knew about a user to deliver a personalized recommendation. 

While these experiences are, in theory, personalized, they came off as creepy, unflattering and troubling brand behavior—and hardly enhanced the customer-brand relationship. 

Importantly, these missteps were limited to the bounds of human-built campaigns and were small in scope and reach. That won’t be the case with AI. Everything—the good stuff and the failures—will happen at scale. 

Start with a little bit of AI-powered personalization here and there before getting too specific. Gradually introduce tailored content or recommendations based on your customer’s actions to avoid overwhelming them with too much at once.

Test everything

AI will magnify accelerate missteps, magnifying everything a digital marketing team touches, from content production to customer communication and data management. With AI tools, teams can now realize hundreds of complete campaigns in a fraction of the time to which they’re accustomed. 

This revolution also means that with AI tools, teams can also realize hundreds of incorrect, poorly planned and inaccurate personalization campaigns in a fraction of the time they used to. AI will scale everything, including mistakes. Especially mistakes.  

One such example: In June, the National Eating Disorders Association presented an artificial intelligence tool, Tessa, as an online chatbot alternative to human counselors. Originally heralded as a cost-cutting move demonstrating AI’s promise, it was designed to address hundreds of requests at once, without the need for human staff. 

It did not take long for things to go wrong. The organization suspended the AI-managed helpline after experts and psychologists observed the chatbot sharing harmful weight loss tips, reflecting hundreds of users' own fears about their bodies. These users’ personalized counseling became AI-generated terror at a rapid and massive scale. 

Continuously test AI-enhanced personalization strategies and collect feedback. A/B testing is at the core of personalization, so test different AI tools and approaches as you would content. This helps in understanding what works best and allows for adjustments based on how your customers respond.

Add a human touch

Consumers increasingly look to their online experiences as an extension of their true selves—but are often met with a distorted reflection—the kind illuminated by shows such as “Black Mirror.” 

According to Vice, one in three Gen Zers say their online identity is the same as their most authentic self. Customers want to feel seen and acknowledged by the content brands serve them—it’s a validation of their deeper identities. When people see well-crafted content catering to them, it assures them their experience is valued. 

Inject a human touch to your personalization routine, even if machine intelligence is being used. Personalized interactions, especially related to sensitive subjects or for unique audiences (see the pregnancy story above) should be personal. While the recommendations might be algorithmically generated, ensuring that there’s a human element in customer service interactions can make the experience less cold and more successful. 

Leaders must tread carefully when applying AI processes to digital marketing operations—there’s too much at stake and so much that can go wrong. Employ AI but keep tabs on it. Whatever you do, don’t get caught up in the AI frenzy and risk letting the tech hijack your brand’s personalization program.