Going Deep on AI: Lessons from the NFL's Chief Data Officer
As organizations go, the NFL is … complicated. The stakeholders are many, from franchise owners and coaches to referees and players. Then there are the league’s media partners, the American public (whose tax dollars help pay for stadiums), and, of course, the fans, who range from die-hard season-ticketholders to casual followers who like to catch the occasional highlights reel.
Oh, and there’s more than $18 billion in annual revenue at stake if one or more of these stakeholders are unsatisfied.
So perhaps it is no surprise that an organization this big and complex has been using artificial intelligence to aid its decision-making for a long time—long before generative AI became a buzzword.
But how exactly are its leaders thinking about AI? And what can other leaders and organizations learn from the NFL’s experience?
Joel Shapiro is a clinical associate professor of managerial economics and decision sciences at the Kellogg School of Management at Northwestern University, where he leads the Analytical Consulting Lab. Over the years, his students have worked closely with the NFL on various consulting projects, which is how he came to know Paul Ballew, the NFL’s Chief Data Officer.
Shapiro and Ballew recently sat down to discuss how the NFL is using data and AI today—and what its plans are for the future. The entirety of their conversation is available to stream above (a transcript is also provided below).
Here are four takeaways from their conversation that leaders in any industry might find helpful.
1. You’ve got to balance quick wins with long-term commitment.
For organizations that are new to investing in data and AI—or are interested in taking their investments much further—the question becomes how to bring others along.
“You know you have to have it,” says Ballew. “But then the question is: What do I do? What do I invest? How much do I spend? How do I justify it? What’s the ROI?”
His advice for anyone in this situation is to strike a balance between putting “points on the board early on” and making the kind of sustained commitment that will really be transformative.
One tip? Design those early wins such that you have optionality. “The optionality allows you to support the business use cases of today, but if you get it right, it affords you the opportunity to leverage it in ways you can’t even envision.”
2. Your organization is not a monolith. Your processes should reflect this.
Before the NFL or any of its 32 clubs can use data to make decisions, the data has to a) exist and b) be shared in the correct way with the correct party, who in turn must be correctly trained to use it.
This is where Ballew spends a lot of his efforts. “Over half my organization runs the data-engineering activities, because you’ve got to get the ingredients to bake the souffle,” he says.
For instance, Ballew has a dedicated data-acquisition team, whose only focus is to acquire data and build new datasets. Without the right data, “advanced technology is useless.”
But as you think through your processes around acquiring, managing, and using data, it’s important to remember that your organization is likely not homogeneous. Different parts of the organization will need different things from your data-science team at different times. Perhaps some parts are already quite AI savvy, for instance, while others will require more support. This is the position in which Ballew finds himself. “I think oftentimes people associate any sports league as this monolithic sort of entity. And it’s not,” says Ballew. “It’s 33 entities: 32 clubs and ourselves. And our job at the end of the day is to help them.”
Whatever those relationships with other parts of your business look like today, expect them to constantly shift and flex. For instance, sometimes his data team is serving clubs, other times it is working with them as partners, and still other times it is setting rules for the clubs to follow. “There are many hats that are worn all the time.”
3. AI is a great input into human decision-making.
On its own, AI doesn’t make decisions. It’s a tool that allows the NFL (or any other organization) to make more-informed decisions. And like all tools, AI can be used cynically, or it can be used to promote the organization’s larger goals.
This means that, yes, as some critics have complained, data and AI could be used to create a dull, perfectly optimized game—the Moneyball-ization of a beloved sport. But Ballew points out that they can just as easily be used to set the stage for more-exciting games—and the NFL is in the business of entertainment, so guess which one is in its best interest?
“We continue to enhance [the game] and modify it, and it continues to result in games that are close and competitive,” says Ballew, pointing to “teams like the Texans going from worst to first” and his “hometown Lions making it to an NFC Championship game for the first time in thirty years.”
On a similar note, data- and AI-enabled injury prediction could be used cynically to, say, trade away a player who might be at high risk of injury. But Ballew says that the league is currently using these tools to improve player safety. These include projects to reduce concussions (and detect them earlier), improve nutrition, and test equipment. Ballew also pointed to recent changes to the game’s kickoff rules that were designed to improve player safety.
“You don’t take the human out of the decision-making, but you bring the science to make the decision-making more accurate, more precise, and help the human beings make better decisions,” says Ballew.
4. Marketing isn’t about your customers anymore. It’s about a single customer.
Like many other organizations, the NFL has historically taken a somewhat blunt approach to marketing. “Buy a ticket; come to a game. We have a game. Here’s our schedule,” says Ballew.
In other words, here’s what we’ve got: take it or leave it. (Though we hope you’ll take it.)
But generative AI is enabling personalization on a scale that was previously unimaginable. For the NFL, this might mean connecting with an individual customer based on them being a Bears fan who travels to other midwestern stadiums once a year to catch an away game. Or it could mean providing helpful information to an individual fantasy football player.
“Now, it’s the opportunity to talk to you about what’s really relevant to you when you want to be spoken to about it and through whatever channel: we’re channel agnostic,” says Ballew, who estimates it will take the league three-to-five years to achieve that degree of personalization.
As much as a technical shift, then, this is a mindset shift. Because it’s not about “customers” anymore. It’s about individuals. “We’re in the entertainment business. And ultimately, we want to help individuals fully take advantage of the experience how they want to take advantage of the experience, not how operationally we think they should take advantage of the experience,” he says. “You can howl at the windmill, but you should embrace it.”
This article originally appeared in Kellogg Insight, a publication of the Kellogg School of Management at Northwestern University. It is used with permission.
