New Zealand Rugby uses Agentic AI to transform performance analysis for teams

Despite sitting at the pinnacle of rugby, NZR operates in an environment where margins between teams are increasingly narrow. The All Blacks win 76 percent of matches and the Black Ferns 82 percent, but even small gains can make the difference between winning and losing.

As performance data and tools have rapidly expanded across the sport, teams are no longer short on information but need faster ways to identify what matters. With more than 1,000 data points collected per player per game, NZR needed to cut through the noise and surface insights that could support decision-making.

NZR set out to build a data foundation with AI capabilities so high-performance staff could validate hunches in shorter time frames. However, its existing architecture was fragmented, with no centralized view of player histories. Because workflows were largely manual, insights often took days to surface, limiting the ability to respond in the moment during matches and training.

“We needed a data environment that increased our nimbleness and gave insights that could support decision-making throughout the performance week,” says Jason Healy, delivery manager, Professional Rugby Platforms at New Zealand Rugby.

Working with AWS, NZR consolidated multiple niche sports analytics systems into a unified, cloud-based data platform. AWS Professional Services collaborated with NZR to design and build a scalable data foundation, addressing fragmentation and enabling consistent data management across teams.

NZR implemented Amazon Redshift as a centralized data warehouse to unify performance data. On top of this, Amazon SageMaker is used to develop and run machine-learning models that generate insights for coaches and analysts. Using Amazon Bedrock, NZR can analyze large volumes of performance data across multiple sources using natural language. The system helps identify relevant datasets and surface insights without requiring manual analysis, supporting new lines of inquiry.

This layered architecture allows NZR to move from data collection to data interpretation more efficiently. Rather than requiring staff to manually analyze datasets, the platform simplifies how data is accessed, understood, and applied in decision-making.

Alongside this, the AWS Prototyping team supported NZR data engineers in experimenting with new ideas and use cases. “AWS helped us get our data in order, creating the right foundation to start engaging generative AI,” says Healy.

Today, NZR operates an athlete-centric data ecosystem where data follows players throughout their careers across club and international levels. For example, if a player is rehabilitating from an injury at their club and joins the All Blacks or Black Ferns squad, staff can access a continuous view of their recovery and make informed decisions. High Performance teams can also access match data to assess player performance and support in-game decisions.

NZR has significantly improved how quickly high-performance staff can access and act on player insights. More than 150 teams surface data faster using Amazon Bedrock, eliminating delays that previously lasted days.

The platform surfaces blind spots and patterns that may otherwise go unnoticed, including moments in a match where unexpected events or changes in play occur. This helps uncover new lines of inquiry and supports more informed discussions around performance. Faster access to insights has also reduced the time spent on analysis, allowing more time to apply insights on the training field and refine performance strategies. “The flexibility of our AWS solution means we gain fast data that keeps us at peak performance and slow data that informs our training programs,” says Healy.

Centralizing player data has also improved continuity across teams, removing friction between club and national environments and reducing the need for repeated data collection, including repeated inputs from health and performance staff. With a flexible, AI-powered data foundation in place, NZR can now build on that continuity by testing and deploying new capabilities quickly, without rebuilding its infrastructure each time.