Artificial intelligence is constantly evolving. But what is it evolving into and how will this constant evolution affect the relationship between customers and brands? In a lively master class session at The Information’s Women in Tech, Media and Finance Summit, three panelists well positioned to see the future of AI explored this topic.

Moderated by Laura Mandaro, managing editor of news and talent for The Information, the panelists discussed what’s working and what’s not, how AI-assisted customer service chats can help influence next-gen products, how to use data to create better experiences and more.

What’s Working Right Now in AI for Customer Relations?

Lauren Wiener, managing director and partner of the Boston Consulting Group, fielded this important question. Lauren has a deep background in AI, having been the CEO of an ad tech company focused on AI and big data. Now she advises some of the biggest companies in the world on how to create meaningful relationships with customers through AI.

Wiener brought up Starbucks and how it uses AI to build customer loyalty through deep personalization.

“I think Starbucks is really one of the best in class. They created a personalization capability many years ago. They wanted to recreate on their email and on their app the personal experiences people would have in stores. It was a multiyear journey with lots of variables and lots of data science. They’re able to pinpoint—using thousands of data points in real time—what people really want, and to create a loyalty program around that. At this point, around 50% of their sales are coming from people in that loyalty program, which means that for you as a loyalty member they can guess what kind of an offer is really valuable to you every time.”

And What’s Not Working?

Wiener pointed out a mistake many companies make in confusing personalization with segmentation, and how that can damage a brand.

“On the worst side, there are a lot of companies out there whose idea of personalization is segmentation. So let’s say you’re interested in hockey and sign up for a sports newsletter. But then you get a newsletter about any kind of sport. But because I like hockey doesn’t mean that I like football or baseball. It’s not at all personalized to me. The segmentation can be inaccurate and not relevant.”

The Human Touch Is Still Important

So what’s the solution to this problem? Wiener believes it’s not just having better data and using it more intelligently, but also integrating human judgment. She gave the example of holiday newsletters. While many people sign up for them, they can inadvertently touch on sensitive topics. If the customer’s mother recently died, for instance, getting a newsletter celebrating Mother’s Day merely serves to open that wound again.

Wiener suggests extending opt-out forms and having a human scan them. “Is this going to affect anybody? Is this something that’s going to be a trigger for someone? I think it’s really important, even as you escalate the data.”

Lakshmi Sharma, chief product and strategy officer of Fastly, agreed: “So it’s validation, supervision to some extent, and putting some kind of regulations and standards in where you can really share the information in a better way.”

Building Products That Reflect What the Customer Really Wants

Cassandra Johnson, vice president of the customer care and vendor management office for Google consumer hardware and services, talked about making the customer front and center in terms of gathering data. “We have almost 700 million interactions a year that we have access to. You have to synthesize that data and not look at it from a single perspective but holistically. You need to be very thoughtful and say, ‘What are those products and services that really resonate with people?’” Now these AI-enabled customer interactions are driving the design of new products at Google.

Building Diversity In at the Start

To use AI as a tool to build trust, Johnson said, you have to think about your audience as a whole—and build that right into your hiring strategy.

“The mindset used to be that I have to hire someone with a Ph.D. But I think you have to couple the data with the context of real-life experience. Some people have amazing real-life experience, but they don’t even have a degree.…I think bringing in that dimensional aspect brings value to your data. It’s just a matter of making sure you have diverse perspectives and backgrounds from all over the world,” Johnson said.

Making the Value Exchange More Valuable

Privacy—and potential abuses of privacy—have been top of mind lately, not in small part due to Apple’s decision to allow iPhone users to opt out of advertising tracking. As a result, first-party data has become even more critical for companies moving forward. But how do you get customers to feel comfortable giving up data in a world where so much conversation in the news revolves around privacy risks?

Wiener from BCG believes you need to truly make exchanging that data worthwhile for the customer, and that means personalizing its value. “One of the things we talk about a lot at BCG is what’s the value exchange to get a customer to feel good about giving their data. It can be about saving time or saving money. It can be relevant to me.” But there’s also the trust factor. “Companies need to be transparent about what they’re doing with the data both in their privacy policy and in the way that they’re collecting information, but also in the way that they’re collecting it to the customers, that they’re not selling their data or using it for evil.” And most importantly, she said, “The offering has to follow up and deliver on what the promise was.”

How Can Companies Really Leverage the Power of AI?

The first thing, everyone agreed, is to bring multifunctional groups together to solve business problems. Engineering needs to have a seat at the table where companies are making the biggest decisions. As Wiener said, “Historically data, data science and engineering have been pushed down in an organization. But this really needs to be someone who reports to the CEO.” And on the flip side, data scientists and engineers need to be educated and prepared to deal with business problems.