To solve FootballCo’s challenge of automating content generation, Firemind developed a AI-driven solution using AWS services. The goal was to automate the creation of structured “Where to Watch” (W2W) articles while maintaining high factual accuracy. The solution was built using a combination of AWS Lambda, Amazon Step Functions, and Amazon Bedrock, all integrated to form a seamless pipeline for generating match previews.

The process began with AWS Lambda, which handled serverless compute tasks to scrape relevant match data from FootballCo’s GOAL.com platform. Lambda functions were responsible for invoking custom-built web scrapers that extracted essential information, such as match schedules, team rosters, and historical data, and uploading this information to Amazon S3 for storage.

Next, Amazon Step Functions orchestrated the entire workflow, coordinating each step of the content generation process. Step Functions managed the flow of data between different AWS services and Lambda functions, ensuring that the data was correctly processed, stored, and passed to the next stage of the pipeline. This orchestration was critical for automating tasks such as managing scraped data, triggering model invocations, and handling validation processes.

Once the data was prepared, it was passed to Amazon Bedrock, which powered the core of the AI-driven article generation. Using Large Language Models (LLMs), Amazon Bedrock transformed the raw data into structured W2W articles, following a predefined template. The model generated content based on prompt engineering, ensuring that the articles were formatted with key sections like kick-off times, how to watch the match, team news, and head-to-head comparisons. The prompts were customised to adhere to FootballCo’s editorial guidelines, and specific instructions were included to prevent hallucination and maintain factual accuracy.

The generated content was then processed for validation and review, ensuring it met FootballCo’s standards. By leveraging the combined capabilities of these AWS services, Firemind created a pipeline that automated article creation, maintained high factual accuracy (80-95%), and significantly reduced the time editors spent manually producing content, leading to an 80% reduction in production time.