The NFL has a playbook for AI

A version of this article originally appeared in Quartz’s members-only Weekend Brief newsletter. Quartz members get access to exclusive newsletters and more. Sign up here.

The Super Bowl is being played today in the San Francisco Bay Area, the global epicenter of artificial intelligence. But the NFL has been running its own AI operation for years now, and the results are showing up where it matters most to a $23 billion business: on the balance sheet.

The league's Digital Athlete program, built in partnership with Amazon $AMZN Web Services, uses AI to synthesize player tracking, video, injury, and practice data to flag injury risk before it becomes an actual injury. It helped teams avoid roughly 700 missed player-games in 2023.

That's 700 fewer instances of a franchise's multimillion-dollar investment sitting on the sideline. For a league where a single star player's absence can swing a game, a playoff berth, or a broadcast ratings window, that number represents serious money.

"We as fans want our favorite players on the field,” said Julie Souza, head of strategic partnerships at AWS. “Players want to be out there showing how exceptional they are. And clearly, clubs and ownership want those players on the field as well. The actions that teams are taking on this data are actually delivering real results and keeping players healthy."

The numbers game

Getting there required turning every NFL stadium into a data collection operation. There was no shortage of raw material. 

The NFL has been collecting player tracking data for a decade through its Next Gen Stats program, which puts chips in shoulder pads and footballs to measure speed, acceleration, and distance. That system generates about 500 million data points over the course of a season, tracking each player's position 10 times per second.

Digital Athlete, which layers in video tracking of 29 points on every player's body 60 times per second along with practice loads, injury records, and field conditions, generates that much data every week. Mackenzie Herzog, the NFL's vice president of player health and safety, said the system has turned stadiums into "virtual biomechanics laboratories."

That data helped the league redesign the kickoff. Digital Athlete simulated 10,000 seasons to model how reducing the space and speed of the play would affect both concussion rates and entertainment value. The result was the dynamic kickoff rule, which brought back more than 1,100 additional returns this season while bringing injury rates in line with standard passing and running plays. 

The league also used video analysis to identify the hip drop tackle as a common mechanism behind high ankle sprains. Banning the technique led to a roughly 25% reduction in lower extremity injuries.

But the league isn't the only beneficiary. If the technology can process data fast enough to predict injuries in a game where 22 players are colliding at full speed, it can handle most of what the business world throws at it. The NFL is, in effect, a proving ground for technology that gets sold to every other industry.

"How do you have data stored, secure, flexible, accessible, and then computed in a way that is actually going to help you derive insights?" Souza said. "If you distill a lot of these use cases down, it's the same fundamental.”

The other playbook

The same data infrastructure is also reshaping how the league reaches its fans. The NFL used AWS to clean up 90 billion rows of fan data scattered across more than 100 sources. Much of it was duplicate or incomplete. Once it was organized, the league went from being able to identify 12 million fans to 78 million.

Targeted campaigns built on that cleaner picture saw two to three times more people actually reading the emails. Meanwhile, Next Gen Stats data now powers everything from Amazon Prime's stats-heavy "Prime Vision" broadcasts to animated game presentations designed to hook younger and casual viewers.

The competitive implications are already creating a kind of arms race. Ryan Paganetti, hired by the Las Vegas Raiders last spring as essentially an AI coordinator, told The Athletic he believes a team will win a Super Bowl in the next few years "utilizing AI at a very high rate, significantly higher than it has ever been used before." An estimated 75% of NFL teams already use some form of AI in weekly preparation, though most are still at a basic level.

For the NFL, the calculus is straightforward. Players on the field generate revenue. Players on the injured list don't. And if the same technology that keeps a quarterback healthy can also sell that quarterback's jersey to a fan in Westchester who didn't know she wanted one, all the better.