Twenty Seconds on the Wall

Palms Down on the Pit Wall

Mara: Brazil, 2019. Lap 52, and a safety car comes out with Max Verstappen leading. On Red Bull's pit wall, the trackside stand where engineers and strategists make the calls, they'd been talking it over, and nobody had committed either way. Then Hannah Schmitz, the team's strategist, steps in. As Females in Motorsport tells it, she says, "Right," and then, "I think we should pit." Verstappen comes in from the lead for fresh soft tyres. He hands over first place, and he goes on to win.

Eli: And that's the world the listener's question lands in. Oracle and Oracle Red Bull Racing have developed an AI-powered strategy agent that helps to make race-day decisions. The listener wants examples, and how they do it. Today's angle is how Red Bull uses an AI strategy agent on race day.

Mara: We'll go from that pit wall to the factory and the cloud behind it, then to the agent itself. Who actually makes a call is the part that stays open. And Schmitz is the place to start. Her first job at Red Bull was modelling and simulation, before she moved over to strategy. She'd also just come back to work after her first child. So Brazil was, in her words, "quite a big thing for me, to prove I was still here and could do the job well."

Eli: And she's called that day "the pinnacle of my career."

Mara: So the call that turned that race came out of a room where nobody had committed, until one strategist did. What was feeding her the numbers is Eli's part of this.

Billions of Permutations in the Cloud

Eli: And a lot of that feeding happens nowhere near the circuit.

Mara: Right. Schmitz alternates her race weekends: trackside on the pit wall, or heading the Operations Room at the factory in Milton Keynes. From there, she's said, "We can be listening to every team's radio, we can be watching every team's on-board footage, we can be looking at all the numbers in detail and have that passed onto the pit wall in seconds." After a call like Brazil's, she describes maybe twenty seconds of waiting to see if it paid off. In a race, she says, that can feel like a lifetime.

Eli: So half the strategy team is watching the race on screens at the factory.

Mara: Watching, and crunching. The numbers they pass on have to come from somewhere.

Eli: They come from Oracle's cloud. Oracle says Red Bull runs billions of simulations on Oracle Cloud Infrastructure, before and during each race. Red Bull's Jack Harington put one figure on it: "We run four billion simulations before we arrive at the track." About one for every two people on the planet. The method is called Monte Carlo. Play the same race over and over, rolling the dice differently each time on tyre wear and weather, and see which plans come out ahead most often. Miles Jackson describes building tyre-degradation and full-race models for exactly that.

Mara: So by Sunday, the team has already watched billions of versions of the race.

Eli: And it isn't the kind of AI people tend to picture. Oracle describes time-based models, refined over years of racing. In Oracle's own words, "These are not massive generative AI workloads." They're precision calculations, built for throughput and consistency.

Mara: Which is the same craft Schmitz started her career in: modelling and simulation.

Eli: Billions of runs, then, and a person on the wall at the end of them. Whether all that arithmetic has ever settled a championship is a separate matter.

From Crunching Scenarios to Agentic AI

Mara: Schmitz has a line on that herself: "Every race-winning strategy that we've made since 2021 has been made on Oracle."

Eli: And 2021 is where it starts. That's when Oracle began working with Red Bull, and those billions of simulations got their first title test that same season. Red Bull's team principal at the time put it plainly: "OCI enabled us to make race-day decisions that helped Max Verstappen win the 2021 Drivers' Championship." A year later, Oracle's name went on the car as title partner. The team counts three Drivers' titles and two Constructors' titles since then.

Mara: Which puts Brazil in a new light. That call came two years before Oracle arrived.

Eli: It does. Then comes the next layer. In 2025 Oracle announced what it called a groundbreaking pilot, bringing generative AI to the pit wall. In 2026, PlanetF1 reported that Red Bull will develop an AI-powered strategy agent with Oracle, designed to work alongside the race engineers. Jack Harington describes it as agentic AI. Plain version: software that doesn't stop at finding information. It recommends an action. Oracle hasn't spelled out which models it runs on or what data goes in. What it suggests with a real race in front of it is the part Mara has been chasing.

The Call at Curva Grande

Mara: It comes from Monza, the 2025 Italian Grand Prix, in Jack Harington's account as Superpower Daily reported it. That's the pilot, in the middle of a race. Verstappen is defending his position, and he cuts a corner. Keeping a place by cutting the track is an advantage he isn't allowed, and it can draw a penalty. Handing the place back cancels the gain before that can happen. The agent recommended exactly that: give it back. Harington said the advice went against the team's competitive instincts. Red Bull followed it anyway. And Verstappen went on to pass his rival again.

Eli: So its showcase moment wasn't a pit stop. It was a place given away.

Mara: Given away against instinct. And that's where people read it differently.

Eli: One reading, held by Carlos Astilleros and by Harington, is that race windows, the target laps to pit, live telemetry and safety-car periods move so fast that teams and drivers have to stick closely to the machine's recommendations. Monza fits that: instinct said keep the place, the agent said hand it back. The other reading runs through Harington too, and through Schmitz and Verstappen. The models guide, and people keep the final say. Harington says the AI is there to enhance engineers and strategists and let them focus on their strengths. And Oracle doesn't describe its pit-wall assistant as making strategy decisions on its own.

Mara: Verstappen puts it bluntly: "The driver is still such an important factor." And: "At the end of the day, I think the driver is still the most important sensor in the car." It's the chain Schmitz described, the factory room feeding the pit wall, with people at both ends.

Eli: Either way, Monza was a place on track, not a pit stop.

Mara: The pit stops on record are still Schmitz's. Qatar, 2025, lap seven, and the field is crawling behind a safety car. When everyone slows right down like that, a stop costs far less time than it would at racing speed. McLaren stays out. Everyone on the wall is asking whether she's sure. She told Dutch broadcaster Viaplay she answered, "Yes, really!" She was sure it was the right thing to do.

Eli: So a stop is cheapest exactly when everyone else is crawling.

Mara: Cheapest, if someone commits to it. Verstappen won, his 70th Formula 1 victory, and Schmitz stood on the podium beside him, six years after Brazil. In between came Hungary in 2022, with undercuts, pitting before rivals to jump them on fresh tyres. And Monaco that year, a three-stop plan that won it for the other Red Bull car. The team's adviser said afterwards, "if we won, it was mainly due to Hannah."

Eli: So, for the listener: all those runs end up in front of people, and strategists like Schmitz turn the odds into a call. The agent's one named live example is Monza, telling the team to hand a place back. The marketing leans on AI, but the race-winning pit calls on record were made by engineers like Schmitz working from cloud-hosted Monte Carlo models, not by an agent acting on its own.

Mara: Three things would change that picture. Whether Oracle ever publishes the model behind the agent. Which races in 2026 run it trackside, and which only test it offline. And whether Red Bull ever credits a pit stop, an undercut or a tyre choice to the agent itself. Until then, the "I think we should pit" from Brazil to Qatar has come from a strategist.

Eli: One thing to read up on: agentic AI.

Mara: Follow technology with us: how should people weigh AI recommendations?

Eli: Thanks for spending this time with us. I'm Eli.

Mara: And I'm Mara. Until next time, this has been Angle.