How CMOs Are Adapting to the Machine Learning Age

Marketers at Elevate: AI discuss unconscious bias, trust issues and other challenges

Michelle Boockoff-Bajdek, chief marketing officer for IBM Watson, gave a keynote presentation at the event.
Sean T. Smith for Adweek

Breakthroughs in artificial intelligence over the past few years have fundamentally altered the ways brand marketers can efficiently reach masses of consumers with individually relevant messaging.

But making the changes necessary to capitalize on that potential—cleaning up data, upgrading systems, integrating disparate business functions—can be daunting for even the most tech-savvy marketing teams. At Adweek’s Elevate: AI event this week in New York, executives shared some of the challenges they are encountering as they adjust to a world where machines control more of the business levers than ever.

Their panel was followed by a frank look at some of the very human issues that can arise when AI doesn’t account for the unconscious biases of its creators, as well as why the early AI personas on the market all tend to be female.

The first step in implementing any AI is attaining and organizing the troves of data needed to power it, which can be much harder than it might sound. By some estimates, the average data scientist spends 60 percent of his or her time on the task.

Like most other brands attempting an AI strategy, TD Ameritrade is currently in the midst of that phase, says Christopher Kienle, managing director and head of marketing innovation and technology at the brokerage.

“One of the biggest projects that we’re engaged in right now is actually a data management program to really start to pull that data into a place where it can be cleaned up and made accessible,” Kienle said. “We still have a lot of legacy systems, and we are in the process of just upgrading even the basics. We are building a lot of AI tools to stand on top of these things, but if we don’t have the foundational basics in place, it doesn’t really matter what we do.”

That process also involves bringing together data from disparate parts of a given company, which may operate on different systems or exist in varying degrees of detail. While there are logistical hurdles involved, forging ahead can unlock new business potential beyond just marketing effectiveness, noted Jon Suarez, chief strategy officer for Salesforce’s Marketing Cloud.

“What’s key here is to engage not just in marketing channels, which are difficult in and of themselves—email, mobile, social—but across the entire journey,” Suarez said. “That’s really one specific area where we see AI, machine learning, the ability to know personalized engagement but try to bring that experience across the customer journey whether that’s marketing or service or commerce—all of the above.”

Marci Raible, vice president of global media and marketing services at Campbell’s Soup, said marketers must develop a degree of trust in technology. For instance, the brand recently let AI make decisions about which consumers are served which recipes, which Raible said felt disconcerting at first but ultimately yielded better results.

“Where we have more of a challenge is on the content and production side, where it’s changing from, ‘I produce this for these people,’ to ‘I’m going to be modular and let the technology actually determine who gets what message,’ and that’s where it’s a little scarier because it takes some of the control away,” she said.

But what happens when humans begin to cede control, only to find that the systems they’ve created are amplifying their harmful hidden biases? That was the subject of the next discussion, kicked off by futurist Amy Webb with an infamous example: a 2015 study finding that the first woman to appear in Google image-search results for the term “CEO” at the time—appearing below several rows of men’s names—was a CEO Barbie doll (actually a fake toy satirized in an Onion article).

“This was not done intentionally—it’s the result of a limited worldview and no-risk modeling and putting speed over longer-term implications,” Webb said.

That led to a conversation about why each of the tech giant’s digital voice assistants—Amazon’s Alexa, Apple’s Siri and Microsoft’s Cortana—have taken on a female persona, a fact that Webb attributes to social science research showing that people respond more positively to directives from a feminine voice. Some companies are pushing back against this norm by developing gender-fluid voices or ones that can be individually tailored to a customer.

Rooting out biases like these takes a commitment to diversity and inclusiveness in the development process, self-reflection and a willingness to acknowledge the limits of your own worldview, said Bruce Duncan, managing director at Terasem Movement Foundation, an organization committed to eventually uploading a person’s consciousness to an AI.

“When you’re thinking about bias, probably 90 percent of it is unconscious—so I’m biased, and I have no idea that my brain has an affinity for certain things based on something that I don’t even know is going on in my own brain,” Duncan said. “However, if you become aware and educated, you can pick up on some sort of sonar reflections of your bias.”

Recommended articles