Sport Direct: NFL is using GenAI to shorten the distance between fan and field
Vast media libraries are now only a natural language query away
NFL Deputy CIO Aaron Amendolia explains how the NFL has used generative AI to create a more immersive and engaging fan experience.
The NFL season concluded with the Super Bowl a couple of weeks ago, but whilst the league is currently in offseason, there is still plenty going in behind the scenes. In recent years the NFL has embraced generative AI and the cloud to connect players, teams and fans to enhance engagement.
Aaron Amendolia is deputy CIO at the NFL, and gives a whistle stop history of the organisation’s data driven progression.
“In 2019, we started the Digital Athlete Program (DAP), which was for player health and safety. We've used data from the Digital Athlete Program - which is looking at camera footage, computer vision, skeletal modelling of the players - and then determining outcomes or predictive analysis for injuries.”
This data-driven approach has already led to changes in both equipment and the rules of the game, helping to make football safer for athletes at all levels.
“Then you start to move into the modern era, where we're doing much more broadcast workflows with AWS, distributing video streams and live games for broadcasters, for fans, for internal officiating purposes.”
The League also began incorporating AI and machine learning into its digital ecosystem, creating a more interactive and data-rich experience for fans. Amendolia explains:
“We have vast content libraries of games, interviews, drafts, all different aspects of our sport. All that stuff has been manual in the past, meaning, going through all our vast archives and trying to search through them. That’s not sustainable or replicable.
“We had multiple systems with one group doing more of our NFL films - content which is documentary, historic. We had one type of system set up with metadata tags, or ways to search for data. They had to tag all these things manually. The next gen stats system was separate from the media asset management system. The systems wouldn't give insight, and you couldn't connect it all up. “
Playbook Pro
Playbook Pro is model built on Amazon Bedrock, which takes the NGS API and integrates it with NFL’s media libraries. It allows show hosts and producers to access detailed game statistics and video highlights all in one place.
Users can ask specific questions like, “Show me all snow games with six-yard rushes” and instantly retrieve relevant video clips and stats. Amendolia estimates that Playbook Pro has cut research time by up to 20 times.
“It’s very powerful for us, but not just for media research. We may potentially also use it in officiating. We use it for researching back into health and safety, because we're changing the interface from being a console or data query to natural language to ask questions. It’s connecting different silos of data into a conversational interface, and it’s been transformational for us.
“As far as production and turn around for showtime, it’s huge. We can use it to create more shows, more content, but really react also things faster and provide more insight. Because GenAI can look across more sources of data than a human can, it can give us different, richer insights about the data. More factors are aggregated together to produce a more engaging view.”
Amendolia expects the NFL’s use of generative AI to continue this evolution, with an emphasis on deeper insights, faster content delivery and an even more personalised fan experience.
“We have a lot of things during our broadcast where we're repeating. For example, ‘I think Josh Allen was the fourth player ever to throw a pass that he also caught back to score a touchdown.’ That’s a very surface level stat so let’s go deeper. Let's talk about the context around the games that happened, about quarterbacks before: Did it win the game? What did it do? I think Gen AI can do that, whereas we're only touching the surface in our current methods.
“We can tell more stories and create deeper engagement.”