Authored by: Aaron Kerzner, VP of Customer Innovation at Zuora

As we think about the nearly endless applications of artificial intelligence (AI) for business, it can quickly become overwhelming. Where should businesses begin and what are the most important focus areas? Strategic planning and practical implementation are required to harness the full potential of AI, while also navigating the inherent challenges. 

I recently had the privilege to speak with AI expert Michelle K. Lee, CEO, Founder and Principal Consultant at Obsidian Strategies, Inc. (“Architects for AI-Driven Strategies”), which helps companies formulate and implement their optimal AI plans and educates business leaders and boards on AI and AI governance.

Michelle, thank you for joining us to share some great insights about launching and monetizing a machine learning initiative. First, how do you think AI could be leveraged as it relates to customer retention, churn reduction, and customer expansion?

Thanks for having me, Aaron. AI is essential to the staying power of the customers. Even today, carefully applied AI is already helping companies accomplish the goal of customer centricity, future proofing, and top and bottom line impact.

Wayfair’s Decorify, Zelig’s Virtual Try-On—these are just two examples of how machine learning is really enhancing both the top and bottom line and creating amazing customer experiences.

74% of customers are actively seeking a highly interactive, highly personalized experience. And AI can provide both at scale, quickly and individually. This level of interactivity and personalization creates amazing customer experiences and retention drives revenue; a win all the way around.

And in addition, the data that you will gather from how your customer is interacting with your products and services, the richness of that, will allow you to improve what you offer, alter what you offer, alter your pricing—all of that will benefit your business goals.

What is the biggest thing companies overlook as part of this process?

I don’t know that they overlook it, but they fail to get it right, and I would say it’s related to data and data strategies. I mean, they often fail to understand all their data assets within their company. I see a lot of data silos and data hugging and it’s critical to break down those silos.

 What data don’t you have today but you have the opportunity to gather? And what are you doing to begin gathering that? Because that creates your AI opportunities, your machine learning opportunities. 

No data, no machine learning opportunity.

Don’t forget the opportunity to combine data from different parts of your organization, but also, don’t forget to think externally outside your organization, if there’s data that might be used to augment your internal data.

So just to give you a very simple example: Domino’s Pizza and their predictive online pizza ordering system. What if they went and looked at the calendar of events and they noticed that Super Bowl Sunday is on a certain Sunday in February and that their pizza orders are likely to spike on Super Bowl Sunday? Or, if the Domino’s Pizza location is near a school that has soccer practice that gets out at 11:30 they might then increase their supplies and their inventory to meet the customer demand.

So think within your organization across departments but also, powerfully, at outside third party data that can help drive the power of these solutions.

So it’s really a journey, right? And some are just getting started. Some companies have kind of achieved that data nirvana where you’re blending data from different sources, but it can be a little bit overwhelming. Every company talks about experimentation, fail fast, minimally viable products, etc.—but when the rubber meets the road, very few have really built a culture around it. So could you share practical tips on building and sustaining a culture of innovation and experimentation?

Sure, Aaron, I’ll share two examples that I have based upon my experiences. So I was at both Google and Amazon and both are incredibly innovative companies that experiment a lot.

And what I would say is: don’t just rely upon good intentions.

Everyone wants to be innovative and they want to allow their employees to experiment but you need to put in structural mechanisms that support that.

So in the case of Google, what they did was in the early days, they allowed their engineers to have 20% free time to work on projects that were unrelated to their day to day tasks. Also, what they did was in the allocation of resources, they allowed 70% of the resources to be focused on their core products, 20% to the next generation (5 years out), and 10% for the incredibly far out moonshot, pie in the sky, almost impossible ideas. So you’re building in the allocation of your resources.

In the case of Amazon, what they did was something completely different, but also very innovative, to encourage experimentation. They would ask all their employees to think of what product or service might benefit their customers or benefit the company internally. 

What we were asked to do is to write what we call a Press Release (PR) Frequently Asked Questions (FAQ). In this two-page document you would spell out: if you were to make the press release announcement of this feature or functionality, what would it say?

You start with that and you work backwards. We would review all of these together with our teams and the best ones would go to the top and get a limited amount of funding. After a certain period of time you check in. If it was progressing well, more resources. If it was not doing so well, you cut it.

So that’s fail fast, but also plant 1,000 seeds and see what flowers bloom.

And that’s how Amazon achieved the culture of innovation and experimentation. Again, not relying upon good intentions.

So you’re asking people to weigh in on topics that might be outside of their core function. It’s also important to have a culture of safety around failure, right? You can fail. And we want to embrace failure because it means that we’re experimenting and we’re trying.

That’s right!

So on the topic of monetization, any advice to our audience on how AI can drive net new revenue streams?

My advice on that, Aaron, is to follow the data. Think about where within your company you have proprietary data that offers your company a competitive advantage. Where is it and what is it?

If we look at the insurance company Mass Mutual, they have 150 years of data on mortality, morbidity, and the pricing of insurance premiums. To the extent that they are able to more accurately predict those items, it goes straight to the bottom line and they’re able to price it more effectively so they can sell more policies.

They use that model to improve their business, but they could also ask themselves: Do we need that model just for our use, or might other insurance companies be able to benefit from the use of that machine learning model? And we could, perhaps in the cloud, charge a fee for it and unleash the full value of it?

In essence, it’s a new product offering.

So, you can build the machine learning model, use it for yourself, and let others use it. But even if you don’t build the model, consider selling or licensing the data that you have to others for machine learning training—either exclusively or non exclusively. The demand for data driven by AI has created revenue opportunities for companies who have valuable proprietary data.

I’ll share with you two examples of this: Reddit and Getty Images.

Reddit recently announced its plans to license its user generated content to a large unnamed AI company. Do you know how much they get paid for that? $60 million per year.

That’s a new source of revenue for an asset that was just sitting there. Think of the possibilities!

Can you share examples of how companies are thinking about monetizing their GenAI investments?

Sure, so I mentioned that I work with a company called Zelig that does virtual online outfit try-on using an AI platform. We considered several monetization strategies. 

First, we considered a user-based fee related to the number of users who signed on to the platform. We also considered an activity-based fee for the number of requests for a total outfit. Then, we considered a combination of the two with an initial, rather modest, fee to acquire the user. If they are a power user above a certain number of outfits, we’d capture more there. And then ultimately, we considered an outcome-based fee which is a percentage of the sales that were converted on our platform.

The bottom line, Aaron, is that these are early days. Companies are experimenting with these applications. They’re going to want to try different monetization strategies and they’re going to have to pivot as needed.

I understand Zuora produces a pretty good tool to allow you to easily explore, track, and change your monetization strategies if needed?

That’s right. Fundamentally, we at Zuora are in the business of helping companies support and evolve their monetization strategies. Just sell how you want to sell, right? Subscriptions, usage, one-time, bundles—being able to handle the nuances of these models from the usage metering all the way through to revenue recognition.

I think that will be much needed as we try these new models.

Thanks so much, Michelle! Really appreciate all the great insights.