Paul Devlin of Amazon Web Services
Artificial intelligence is moving from a niche tool to a central part of sports, betting and live-event production, Paul Devlin of Amazon Web Services said Thursday at the G2E Asia gaming expo. Speaking on the final day of the event, Devlin said the sector’s challenge is not simply adopting new technology, but using it to solve clear problems, improve customer experience and manage risk.
Amazon Web Services (AWS) is a subsidiary of Amazon that provides on-demand cloud computing platforms and application programming interfaces to individuals, companies and governments on a metered, pay-as-you-go basis.
AWS presents itself as the world’s most comprehensive and broadly adopted cloud platform, offering over 200 fully featured services from data centers globally. By providing computing power, database storage, and machine learning capabilities, the platform allows organizations to replace physical data centers with scalable, virtual infrastructure.
The company has established itself as a critical technology partner for major sports organizations, including the NFL, Formula 1, the NBA and the Bundesliga.
Through these partnerships, AWS processes massive datasets in real time to generate advanced analytics, such as predictive player performance, speed tracking and historical game insights. AWS data feeds are integrated directly into television broadcasts and digital fan experiences, deepening audience engagement and providing viewers with a more analytical perspective on gameplay.
Moreover, the platform’s machine learning and security tools allow operators to monitor for fraudulent betting patterns and maintain system integrity during high-traffic events. Devlin said AWS’s approach centers on “model choice,” which he described through Amazon Bedrock, a platform that gives customers access to a wide range of large language models.
He said users can choose the model that best fits a specific workload and replace it when needed. Essentially, AWS functions as the foundational layer that enables betting platforms to scale securely and reliably in a fast-moving, data-driven market. This very infrastructure then serves as the technical backbone for the betting and gaming industries.
Sports as a test case
Devlin said the company’s approach is to “think global and react local” by taking lessons from one market and adapting them to another. Devlin described AWS’s cloud infrastructure as a set of “Lego” blocks that companies can combine to build services at scale. He also pointed to Netflix as an example of a global business built on AWS.
For Devlin, the cloud allows businesses to test ideas without making long-term commitments, because failed experiments can be switched off quickly. “Invention requires two things: the ability to try a lot of experiments and then not living with the collateral damage or failed experiment,” he said. He also cited Jeff Bezos’s idea of “two-way doors,” meaning decisions that can be reversed if needed. “When you have 70% of the information you expect to be able to have on a two-way-door decision, make the decision,” Devlin said. “Speed of innovation: step through the door, try the experiment, learn, iterate, and if it doesn’t work, you can switch off those resources.”
AI and generative AI
Drawing a distinction between traditional artificial intelligence and generative AI, Devlin explained how AI is used to automate manual tasks, while generative AI is able to create new content and new forms of analysis.
Using the NFL as an example, Devlin said analysts once had to manually tag every player’s position during a game. Now, that process is automated, and machine-learning systems help identify the most effective plays. “Generative AI can do all of that,” Devlin said, “but generative AI can generate and create plays that you and I have never seen before.”
And AWS has been applying that thinking in sport for years, he said. Devlin detailed that these thought processes led the company to F1 Insights, which uses AI and machine learning to turn large volumes of car, pit and sensor data into race-strategy analysis, competitor comparisons and car-performance insights. “Take that stream of data from all the race cars, all of the data and insights that we’ve got in the past, and surface insights that can allow better storytelling, be it the broadcast or whoever the stakeholders are […] it’s quite simply text summarization.”
Tennis and fan experience
Then, shifting gears to tennis, Devlin pointed to AWS’s work with the Australian Open and Bolt6, which uses AI-enabled cameras across its courts to track players, rackets and the ball.
The project began with electronic line calling in 2024, designed to replace human line judges and reduce errors. Devlin even detailed that the project had to adapt the voice used for calls, because fans noted it not sounding authentically Australian. “
They actually found that the fans could detect if it was an Australian voice or not, so that they had to literally record the person who used to call them one of the judges to make it sound Australian, or to make it sound like a judge would sound, an unexpected part of innovation,” he said. The organization wanted to replace it for two reasons.
One, because the matches went on too late into the evening and their audience had to go to bed, and two, because humans made errors, and they couldn’t have errors, especially when it comes to the betting market. Then, by 2025 and 2026, Devlin stated, the system had become fully automated; and the next focus for AWS was on improving the viewing experience.
“How can we take fans to a level of understanding of the game that they can’t get from technology alone?” Devlin queried. He then described a setup in which cameras at the court’s baseline are stitched together in the cloud, allowing producers to switch angles and overlay real-time information such as ball speed and shot type.
“The key was (asking), how can we improve the fan experience to enable them to see the game from a different angle,” the betting, gaming and sports expert stated. Devlin also described a more ambitious project that puts viewers into the player’s perspective. The new idea, he said, was tested on YouTube using Nintendo-style characters before being sold to other countries.
Golf and betting data
Another major area where AI is influencing both broadcasting and betting, according to Devlin, is Golf. He pointed to the ShotLink scoring used by AWS on the PGA Tour, which tracks every shot by every player and generates a detailed data set that can be used for fans, broadcasters and betting content. He said betting fans want deeper research before placing wagers, but producing that content manually is slow and expensive.
AWS responded by using agentic AI through Amazon Bedrock, deploying six specialized agents to help generate articles and summaries.
“There are six separate agents built through Bedrock, and those six agents are fulfilling specific roles to be able to populate these articles,” he said while highlighting that the system generates “1200 articles generated per week using AI with human in the loop, but importantly, a 25 cent per article generated.” Devlin said that system represented a major reduction in cost and a far more efficient content operation. The same data can also support betting profiles and help operators better understand fan behavior.
AI agents and automation
AI is now moving beyond chatbots into autonomous agents that can complete more complex tasks with less human oversight.
“We can set an agent off to go and complete the task,” Devlin said, adding that more advanced systems can return with a finished result under the supervision of another agent. As an example, Devlin referred to the NFL. He said the league has used generative AI to cut onboarding time for new staff by 67 percent. While Formula One’s F1 TV service has also used agents to monitor feeds and reduce issue-resolution time, though such improvements are mostly invisible to the platform’s subscribers, he noted. Across Amazon’s wider business, Devlin said more than 10,000 applications have been migrated to AWS, saving thousands of developer years and more than 260 million dollars annually.
Data ownership
The session ended with questions about who controls the data flowing through sports betting systems. Devlin said AWS acts only as infrastructure provider and does not use customer data for its own purposes.
“We physically can’t,” he said when asked whether Amazon can access customer information. He said the data belongs entirely to clients and is protected by technical and contractual safeguards. That, he argued, is why major sports, government and financial-services customers trust the cloud.
No Comments