In a Microsoft Signal feature published October 16, Mercedes-AMG PETRONAS F1 Team lead race strategist Joey McMillan described how the team's 2026 partnership with Microsoft changed its race preparation. Using Azure Batch, the team runs complete race simulations - every lap, pit stop and overtake - and can scale to thousands or millions of runs, McMillan said. Work that once took up to an hour on a Saturday night now returns in seconds and can be run from the pit wall during a race, he said. McMillan also said AI coding tools let him write models and analysis much faster despite not being a professional coder. He stressed that strategy still depends on human judgment under pressure, citing a wet Miami Grand Prix where the team chose tire compounds and an undercut call in real time.
Our takeThis is a clearer example of AI at work than most agent launches: cloud batch simulation compresses a night's compute into a pit-wall decision window, and coding assistants widen who can build the models. But the final call remains human, and the article offers no error-rate or outcome data - just one race that went well. The real test is whether faster simulation improves decisions consistently, not whether it produces a good story.