AWS re:Invent 2025 - AI Agents Slash PGA TOUR's Content Cost While Boosting Coverage/Quality -SPF204

AWS re:Invent 2025 - AI Agents Slash PGA TOUR's Content Cost While Boosting Coverage/Quality

Overview of the PGA TOUR

• The PGA TOUR is a player-owned organization responsible for operating golf tournaments globally
• They have 4 different tours (PGA TOUR, Korn Ferry Tour, PGA TOUR Champions, PGA TOUR Latinoamérica) to provide playing opportunities for golfers
• The PGA TOUR events take place on large, 150+ acre courses with natural obstacles, making data collection a significant challenge

PGA TOUR's Digital Strategy

• The PGA TOUR's digital strategy is centered around:
  • A comprehensive website (pga.com) with in-depth golf content
  • Highly engaged mobile apps with features like leaderboards and video
  • Collaborating with other golf organizations to provide a consistent fan experience
• Data collection is crucial, with the PGA TOUR gathering petabytes of data from 140+ 4K cameras, radars, and walking scorers at each event

Challenges in Content Generation

• The PGA TOUR faces challenges in effectively telling the story of their events and engaging fans due to the complexity of golf:
  • 156 players, 31,000 shots per tournament
  • Variety of scoring formats (stroke play, match play, stableford)
  • Difficulty in making the sport easily understandable for casual fans

PGA TOUR's Journey with AI

• Initially, the PGA TOUR experimented with chatbots, but pivoted to focus on more impactful use cases
• In 2025, they launched an AI-powered shot commentary system that provides factual information and relevant context for each shot
• Parallel to this, the PGA TOUR developed an "agentic" AI system to generate content at scale, particularly for non-tournament days

Agentic AI Content Generation

• The PGA TOUR generates around 800 articles per week, covering various content types:
  • Betting profiles for 156 players
  • Tournament previews, recaps, and player recaps
  • Round-by-round updates
• Generating this content at scale with consistent quality and brand alignment is a significant challenge
• Key considerations:
  • Leveraging structured data from the PGA TOUR's APIs and unstructured data from media guides
  • Ensuring content accuracy and compliance with brand guidelines through validation
  • Generating content in various formats (long-form, social media, push notifications)

Technical Architecture

• The PGA TOUR's agentic AI content generation system is built on AWS, leveraging the following components:
  • Amazon DynamoDB for managing job requests
  • Amazon SQS for queueing content generation tasks
  • AWS Lambda and the Bedrock AI platform for executing the content generation workflows
  • Amazon S3 for storing the generated content
• By running the content generation on AWS Lambda and Bedrock, the PGA TOUR has achieved a 95% cost reduction compared to their previous approach

Business Impact

• The AI-generated content has resulted in billions of page views per year for the PGA TOUR
• The speed of content generation (within 5-10 minutes of a tournament ending) allows the PGA TOUR to be the first to market with relevant information, driving high engagement
• The cost savings from the agentic AI system have been significant, with each article costing only $0.25 to generate compared to the previous approach

Key Takeaways

• The PGA TOUR has successfully leveraged agentic AI to generate high-quality, on-brand content at scale, addressing their key challenges in fan engagement and operational efficiency
• By taking a product-first approach and prioritizing accuracy, brand alignment, and production-readiness, the PGA TOUR has been able to implement AI in a way that delivers tangible business value
• The technical architecture, leveraging AWS services and the Bedrock AI platform, has enabled significant cost savings and operational improvements
• The PGA TOUR's journey demonstrates the potential for agentic AI to transform content generation and delivery in the sports industry, with broader applications in other data-rich, fast-paced domains