Transforming sports storytelling with generative AI on AWS

Sports broadcasters face an ongoing challenge: how to deliver compelling stories while managing vast amounts of real-time and historical data during live events. For mountain bike racing commentators, this means processing information about hundreds of athletes, multiple race configurations, and constantly changing conditions—all while keeping viewers engaged in real-time.

We’re excited to share how Warner Bros. Discovery (WBD) Sports Europe and Amazon Web Services (AWS) are transforming live sports broadcasting with Cycling Central Intelligence (CCI), a new generative AI-powered platform that enhances the art of sports storytelling.

The evolution of sports storytelling

Traditional sports commentary relies heavily on the expertise and preparation of broadcasting teams. For the WHOOP UCI Mountain Bike World Series, commentators previously spent hours researching across multiple sources before and during races—from in-race statistics and athlete profiles, to historical performance data and course information. This manual process limited their ability to deliver the rich, contextual storytelling that today’s viewers expect.

Moreover, with events spanning multiple countries and languages, accessing comprehensive information in real-time presented significant challenges. Commentators needed a solution that could help them focus on what they do best: bringing the excitement and drama of mountain biking to life for viewers around the world.

Meeting the challenge with CCI

CCI represents a new approach to sports broadcasting, born from the AWS Working Backwards process. The AWS Digital Innovation team collaborated closely with WBD’s production teams to understand their specific challenges and vision for the future of sports storytelling. This deep, customer-focused engagement helped shape a solution that seamlessly integrates with WBD’s existing production workflows.

Built by AWS Professional Services and integrated into WBD’s Production Assist application stack, CCI combines human expertise with the power of generative AI. The platform serves as an intelligent research assistant, providing commentators with instant access to relevant information through natural language queries. This integration with existing production tools verifies there is minimal disruption to established workflows while maximizing the benefits of AI-powered assistance.

Key benefits include:

• Instant access to comprehensive rider profiles and statistics
• Real-time translation of international content
• Contextual insights based on historical data
• Seamless integration with existing broadcasting workflows

How CCI works with Amazon Bedrock

At its core, CCI leverages Amazon Bedrock and Anthropic Claude to create a unified knowledge base. The AWS Professional Services team designed the platform to process and analyze hundreds of documents and dozens of statistical datasets, combining structured and unstructured data into readily accessible insights. By building on WBD’s pre-existing Cable News Network (CNN) Assist infrastructure to create the multi-org Production Assist application, the solution verifies familiar user experiences for production teams while adding powerful new capabilities.

Reference architecture

Figure 1: High-level architecture.

This architecture enables production teams to seamlessly access unified, AI-enriched knowledge extracted from unstructured sources, transforming legacy document management with minimal operational disruption. By leveraging the managed Retrieval-Augmented Generation (RAG) workflow of Amazon Bedrock and the advanced reasoning capabilities of Anthropic Claude, the CCI platform delivers accurate, context-aware insights at scale, streamlining information retrieval and decision-making for complex production environment.

Key components of the solution include:

Data ingestion pipeline

• Event documents-such as press releases, course maps, and rundowns are replicated from a legacy Document Management System (Cyberdocs) into an Amazon Simple Storage Solution (Amazon S3) staging bucket.
• Metadata for each document is synchronized with Amazon Relational Database Service (Amazon RDS), confirming both content and contextual information are preserved for downstream processing.

Automated processing

• Ingestion workflows are triggered by Amazon S3 events, supporting both bulk onboarding and incremental updates as new or modified files arrive.
• An Amazon Simple Queue Service (Amazon SQS) queue manages event flow, invoking AWS Lambda functions to orchestrate processing tasks.
• Lambda functions extract text using Amazon Textract, enrich metadata, and transform documents for ingestion into the knowledge base.

Content transformation

• While Amazon Bedrock Knowledge Bases can natively ingest most rich-text formats, PDF files with embedded images or markup are processed using the AWS Textract-Textractor solution from our GitHub repository for efficient text extraction.
• Each document is paired with a metadata JSON file, following Amazon Bedrock Knowledge Bases best practices for metadata management. This normalization verifies consistent, high-quality data is available for semantic search and retrieval.

Amazon Bedrock Knowledge Bases synchronization

• Processed files and metadata are ingested into Amazon Bedrock Knowledge Bases on a scheduled basis.
• Amazon Titan V2 generates vector embeddings for each document, powering semantic and hybrid search capabilities.
• Embeddings and source mappings are stored in Amazon OpenSearch Serverless, enabling fast, scalable, and relevant document retrieval.

Retrieval APIs

• Amazon API Gateway exposes endpoints for natural language queries from client applications.
• A Lambda function acts as the orchestrator, bridging client requests with Amazon Bedrock APIs and the underlying knowledge base.
• The Retrieval API retrieves relevant document chunks, from the Knowledgebase and return the chunks with meta data augmented with source document location.

Generative AI – Anthropic Claude and Amazon Titan V2 Model

• Amazon Bedrock APIs enables access to Anthropic Claude models for final summarization response generation.
• Amazon Bedrock APIs enables access to Amazon Titan V2 model as embedding model to vectorize unstructured data as part of Knowledgebase creation and Retrieval API execution for converting natural language query to vector.

Structured data CDF – Structured data and APIs

• WBD manages the live race results, rider’s profile and historical race results as structured data source in MySQL DB.
• WBD provides the set of APIs to expose structured data on top of structured data source in MySQL DB.

Production Assist – User experience and Retrieval Augmented Generation

• WBD manages Production Assist application as user experience layer for commentators to search the relevant information in natural language.
• Production Assist system manages the orchestration including identifying intent, call to structured data or unstructured data APIs, prompt management and calling the Amazon Bedrock API for final summarization of the commentator query response.

Infrastructure as Code
All relevant components are deployed and managed using Terraform as Infrastructure as Code. This modular approach ensures consistent, version-controlled, and repeatable deployments across all environments. By integrating AWS managed services with custom components, the architecture delivers robust, scalable, and maintainable solutions for data processing, AI-enhanced storage, and API-based data serving. The deployment stack consisted of:

• A data ingestion module that processes Cyberdocs into curated knowledge base documents.
• A knowledge base module that provisions the OpenSearch Serverless database and Amazon Bedrock resources.
• An API endpoint module that sets up the infrastructure for serving the knowledge base.

Real-world impact on broadcasting teams

“What makes CCI truly revolutionary is how it enhances and complements the human expertise that makes sports broadcasting special,” says Chris Ball, VP, Cycling Events at WBD Sports. “Our commentators bring unmatched levels of experience and passion for the sport. CCI ensures they can expertly craft the compelling stories that keep viewers engaged.”

The platform debuted at the 2025 WHOOP UCI Mountain Bike World Series season opener in Araxa, Brazil (April 3-6), where broadcasting teams used it to deliver enhanced coverage across WBD’s channels and platforms, including Eurosport, TNT Sports, Max, and discovery+.

Future possibilities for AI in sports coverage

The launch of CCI represents just the beginning of what’s possible when combining human expertise with generative AI in sports broadcasting. As the platform evolves throughout the 2025 season, WBD Sports will continue to explore new ways to enhance the viewing experience, including:

• Advanced performance analytics
• Deeper historical insights
• Enhanced multilingual capabilities
• Expanded race strategy analysis

Conclusion

Warner Bros. Discovery is revolutionizing sports broadcasting by combining human expertise with generative AI capabilities on AWS. Their Cycling Central Intelligence platform demonstrates the potential for thoughtfully applied AI technology to enhance, rather than replace, the human elements that make sports coverage compelling.

We invite you to follow the WHOOP UCI Mountain Bike World Series throughout the 2025 season to experience this innovation firsthand. For more information about AWS solutions in media and entertainment, visit AWS for Media & Entertainment.

We gratefully acknowledge the assistance of Dashiell Flynn, Pr. PS Cloud Architect, whose behind-the-scenes contributions helped shape both the project and this article.

Contact an AWS Representative to know how we can help accelerate your business.

Further reading

• Retrieve data and generate AI responses with Amazon Bedrock Knowledge Bases
• Amazon Bedrock Knowledge Bases now supports hybrid search
• Supported document formats and limits for knowledge base data
• Supported models and Regions for Amazon Bedrock knowledge bases
• Warner Bros. Discovery Launches Generative AI-Powered Cycling Central Intelligence Platform In Collaboration With AWS