From weeks to minutes: How Formula 1® uses agentic AI on AWS to accelerate data operations

Formula 1 races every two weeks, but its data engineering ran on a much slower clock. Chris Roberts, F1's Director of IT, says the team had an 18-month backlog just to wire up 12 new data sources into Customer 360, the platform that tracks fans across ticketing, F1 TV, social, and merchandise. Each source meant 6 to 8 weeks of engineers hand-writing schema mappings, ingestion pipelines, and GDPR tagging. That's not a pace that matches a sport where sponsorship deals and fan campaigns need to move fast.

So F1 and AWS spent early 2026 building something called the Data Accelerator, running on Amazon Bedrock AgentCore. The setup is fairly clever: an engineer uploads a business requirements document to S3, and an agent reads it, generates schema mappings, infrastructure code, DBT transformations, and governance policies, then opens pull requests in GitHub with a linked Jira ticket. A human still reviews and approves, but the boilerplate is gone. F1 says onboarding time dropped from 6-8 weeks down to about 40 minutes of code generation plus a few hours of deployment, with AI handling 95% of the work.

The more interesting piece might be what happens after onboarding. Upstream data providers change column names or restructure payloads constantly, and it always seemed to happen mid race-weekend. Now the same agent architecture watches for those changes via EventBridge triggers, figures out which pipelines and downstream consumers are affected, drafts the fix, and files a ticket explaining exactly what broke and why. What used to take days of firefighting now takes hours.

F1 also rebuilt how teams actually see the data. Before, engineers, analysts, and data scientists worked in separate AWS accounts with no shared context, so answering a simple