How do you determine what’s the right balance between personalizing content to keep a customer interested and wanting to engage more, and not making them uncomfortable by using information you know about them? 

Chakravarthy: The line there is something that you realize customer by customer, because not every customer is the same. Some people are like, ‘Oh, I’m fine with being reminded. I was trying to book a trip to go somewhere. I didn’t finish booking the trip, and they sent me a reminder to finish booking, and I actually like the reminder.’ Another person goes, ‘Wow, that was annoying. I don’t want to be reminded. I already went through a different channel and I already booked my travel, or I canceled. I’m not interested at all. I don’t want to get the reminder.’ There’s no perfect answer.

The only real answer is to be able to gather as much data—not only about the actions that the brand is taking in terms of putting out an offer, a message, a reminder or an ad—but also taking the reaction into account. If I put an ad for you and say, ‘Hey, reminder: You were trying to book a trip to New York,’ and you want to complete it, you go ahead and follow up on it, I know that it worked for you. That’s a data point that says that was a good thing to do. Same thing, if I did that for another customer, and the person completely ignored it, that is the other data point. 

AI is capable of learning both through positive reinforcement, as well as through negative reinforcement when an outcome that you’re expecting doesn’t happen. You can train the AI to get the right balance because ultimately, it’s in every brand’s interest to make sure that they’re putting in front of prospective customers something of interest to them so that they actually take the action. 

It costs them money to do that. It’s wasted if I’m trying to do that with somebody who’s not interested. If it upsets them, that’s a double whammy. Minimizing that is a key part of how we apply AI.

When using AI to enhance the customer journey and experience, are there any AI-related pitfalls that companies should look out for?

Yes. Start first with hallucination. When you’re using an AI model, the AI models tend to hallucinate. They may come up with text responses, images or video that are entirely generated, that never existed in real life. It’s very important to know where hallucination might be okay and where it’s not. If you are a pharmaceutical company using AI and you established a chatbot or a conversational interface with a customer who’s a potential patient looking for information, you don’t want there to be any [hallucination]. It depends on the context.

Second, depending on your industry, you need to know: If you provide some information to a customer or a patient or whoever, where did that information come from? You don’t necessarily find that through all the LLMs and AI tools, because they synthesize information from a lot of different places. They give you the synthesis, but they won’t say where it came from. Depending on your brand and depending on your product, it may matter a lot on where the information came from. You need to be able to trace it back. 

Third, when you are depending on your brand, you want to make sure that the AI you’re using is not misusing somebody else’s intellectual property. The AI might have been trained on your competitors’ ads and has ingested all of them. You go in and say, ‘Hey, I’m a real estate company. I’m trying to generate an ad for a home I’m selling.’ But you don't want that to come up with exactly the same ad that your competitor had. 

The fourth one is the same thing; the other side of the coin. You want to make sure that anything that’s your intellectual property, your trade secrets, are not used by the AI or misused by that AI. Say [you’re working with] Coca-Cola. You’re using AI to generate all these new images and videos and so on, which are true to the Coca-Cola brand, but you don’t want all those images to be used for some other brand somewhere else. That should be only for your use.

How much does AI raise the bar for what is considered good CX?

It raises the bar to the extent that it is fast becoming table stakes. Without it, you cannot really deliver the kind of customer experience that especially the next generation of customers want. This is the generation of customers that’s used to everything to be personalized. They can’t think of something where they go to a generic website. Going to the DMV, they feel like, ‘Wow, I can’t believe that this is how you provide service.’ They assume a number of things about what you know about them and what the level of experience needs to be. 

It is not just the personalization. It’s the kinds of things like: How do I understand and have empathy for you? How do I understand are you really in the market for something? When should I be selling to you? When should I be educating you? When should I be trying to guide you, and when should I just be quiet with you? Understanding that and applying that is raising the bar.

What advice would you give to a CMO about AI’s proper use for CX, and how to go forward with it?

You should be doing this. This is table stakes and it affects every industry, every brand. Ignoring it is not an option. You should be working with AI. 

Second, remember that it’s not just plugging in a new technology. AI is a catalyst for fundamentally changing your process, going from more of a waterfall-type process to an agile process. It’s a way of changing your operating model so that you’re going from a quarterly cadence or a monthly cadence, to a continuous operating cadence. You’re operating in minutes, days, hours.

It’s about your people and your organization and your skills, because there are many jobs that you have in your current organization that AI is going to do, but that doesn’t mean that those people can’t be re-skilled and are not interested in doing other things. They can do a lot more with AI. We always believe that AI amplifies the ability of the human. Instead of thinking from a marketer’s perspective, AI will replace the people I have, think how do I amplify what my people can do with AI? That’s what is going to get you the leapfrog effect.

Then, in selecting the technology partner that you want to work with, make sure that it is somebody who will invest and innovate with you for the long term, and who has the right business model so that they don’t have problems with hallucination or intellectual property. That their interests are aligned with yours.