For marketers, generative artificial intelligence is a paradox. The technology enables us to customize messages for each customer, at a scale and precision never seen before. It provides the simulation of empathetic human interactions using the most advanced technology tools, like large language models (LLMs), artificial intelligence (AI), and machine learning (ML).
This is its promise and its danger. The benefit to the marketer—and, in the best case, to the customer—is the personal, appropriate message that delights and convinces.
However, the potential downsides and points of failure are staggering—and alarming. We could damage our brand by using AI-generated conclusions that people find offensive or that are completely wrong. We could establish an emotional connection with customers that backfires when they find out their conversation is AI-generated. The pros and cons of this technology also change from person to person. As Slava Polonski writes, “One user’s idea of personalization is another’s idea of intrusion.”
How do we balance the pros and cons, and take advantage of this powerful technology while keeping it in check? I propose a simple metric for any AI communications initiative: We ask the question, “Who’s getting the most long-term value from this communication—our business, or the customer?”
If we use AI to focus on the long-term value to customers above all, I maintain that we’re on the right track. What we’re looking for is the direct opposite of the common trend: the all-too-common regression from serving the customer to eventually serving only themselves. As firebrand Cory Doctorow explains, “Here is how platforms die: First, they are good to their users; then they abuse their users to make things better for their business customers; finally, they abuse those business customers to claw back all the value for themselves. Then, they die.”
IN PRACTICE
For personal interactions, such as email messaging and contact center dialogs, start by selectively curating the data you use to generate AI insights. While AI can produce deeply revealing information about a person by cross-referencing first-party data from the business with other publicly available information, and data from LLMs, these actions may be perceived as “intrusive and overreaching” if this content is used when you’re communicating with a customer or prospect. Yes, you may be able to discern their political leaning, education level, and so on, but sending communications that use that information is off-putting to many and may expose you to blowback.
It’s a better bet to use only the data you’ve collected—your web and email interactions, for example. In other words, if you’re looking to build trust with customers, use only the most trusted data you can: your own.
You can also provide guidance using look-alike or synthetic data that is free of personally identifying information (PII). This is the “customers like you” line of advice and is less likely to be seen as exploiting individual data.
The goal, of course, is to give customers the personal touch and empower them by providing relevant and helpful information so they can make the best decisions for themselves. If your advice hints at steering or manipulating the customer, rethink it.
Regardless of the information you send customers, AI is also useful in shaping how you say it. LLMs can be used to check messages against your own style book or your library of existing content, so all your communications are in your brand’s voice.
AI CAUTIONS
While people might like a personal touch or being sent loyalty coupons, nobody likes being exploited, and when AI is used to optimize revenue on an individual level or in real time based on external events, it can be taken that way.
In 2014, in Sydney, Australia, a gunman held people hostage at a café. Uber’s surge pricing algorithms, seeing the spike in ride requests from people fleeing the area, sent fares skyrocketing to 4x typical rates before Uber put the brakes on this algorithmic exploitation. The near-term reputational damage took time and money to overcome, with Uber ultimately having to pay out more for free rides than if they had simply done the right thing at the start—and could have been mitigated by having better monitoring and governance of their systems and algorithms.
Airlines have been applying dynamic pricing based on seasonal demand, inventory, and other factors for years. It’s technically possible to adjust pricing per person based on their logged behavior. This is an effective way to improve profitability and sell-through. However, it’s also a good way to alienate customers and is the opposite of rewarding a customer’s loyalty.
Another example of a misfire by the airlines was the recent CrowdStrike outage. Some airlines relied heavily on their technology systems in place and in doing so, completely lost all empathy. Here again, human oversight (guidance) in this extraordinary situation would have served everyone well.
All AI optimizations are based on data. The more trusted data you have, the more effective the tools can be. But to add value for the customer, I recommend that you provide clear information on what you are gathering and allow customers to opt out. Emerging regulations may demand this, and you must track these laws, which may vary by region.
TRUSTED DATA IS EVERYTHING
AI can be a strong asset for improving customer service and reinforcing brand voice. The most critical resource you can feed it for that purpose is trusted data—about your customers at the individual level, about your products, and from external factors that affect your business. We learned from a recent survey conducted for Teradata by NewtonX that while many executives trust the potential of AI, a large percentage lack confidence in their company’s data readiness to ensure its reliability. In fact, 4 in 10 executives don’t trust their data to generate accurate AI outputs.
Putting AI to work on all this information is powerful, but demands accountability and sensitivity. You can and should use AI to improve the customer experience, which will have long-term benefits for your business. Play the long game.