Porsche Cup Brasil uses Azure AI to keep more drivers competing on race day

There is a moment in motorsport that no one plans for but everyone prepares for. A car goes into a corner just slightly too fast. Another car drifts a few centimeters too wide. And suddenly, physics takes over. 

The impact is immediate. The consequences are not. Because in a tightly run racing series like Porsche Cup Brasil, the crash is not the real problem. What happens next is.

Porsche Cup Brasil operates at a scale and level of control that is unusual, even within motorsport. Eighty-two identical cars, all maintained and serviced through a single centralized team. Multiple races across a single event weekend. Hundreds of people working across logistics, parts, and repair. Every car prepared to the same standard, every driver competing on equal terms.

It is a system designed not just for performance, but for consistency. COO Enzo Morrone puts it plainly, “The schedule of an event is really tight, so you cannot miss one minute.” This is not just a racing operation. It is a precision business system under constant stress.

And the constraint, the thing that quietly governs everything, is time.

What makes this constraint meaningful is not just how little time there is, but how much depends on understanding it correctly. A damaged car can often be repaired in time to return to competition the same day. That possibility is what keeps drivers in contention and preserves the integrity of the event. But it only works if the team can quickly and accurately determine what needs to be fixed and how long it will take. 

For the people managing the operation, the problem is not simply speed. It is predictability. Drivers depend on it to understand whether they will make the next race. Teams depend on it to plan their work. And leadership depends on it to ensure that a tightly scheduled system continues to function without disruption.

Improving what already works—without slowing anything down

By the time Porsche Cup Brasil began exploring AI, the organization already had a system that worked. Cars were being turned around in roughly two hours. For an operation of this scale, that is not a trivial achievement. It is the result of experience, process discipline, and a deep understanding of how to manage complexity under time pressure. Which made the next question more difficult. 

What do you improve—and what do you leave alone? Because in a system like this, stability matters as much as speed.

Dener Pires, the CEO, had built the series around that principle. Every driver runs under the same conditions. Same cars. Same support. Same rules. “We created a formula that is identical for everyone,” says Pires. That formula is what makes the competition credible. And it is also what makes change risky. He describes the opportunity in practical terms: “We already have a strong operation. The opportunity is giving our team better tools—so they can move faster, make decisions with more confidence, and focus on the work that really matters.”

When Microsoft and Kumulus began working with Porsche Cup Brasil, the team approached the problem with that in mind. There was no assumption that technology should reshape the system. Instead, the work began with understanding where time was spent, where uncertainty appeared, and where even a well-functioning process could be made more consistent without introducing disruption.

If you stand in the garage on race day, one part of the process stands out. The repair doesn’t begin immediately. There is a pause first. A damaged car arrives. Photos are taken. Mechanics begin to lay out each part and evaluate the damage. They are building a list—every component that needs to be repaired or replaced. Only once that list is complete can anything else move forward. This step is careful, necessary—it is also where time accumulates in a way that is difficult to control.

Morrone explains, “Sometimes we spend 20 minutes, 30 minutes just listing the parts.” That time is not wasted. It is essential. But it introduces something the system struggles to absorb: variability. Some incidents are straightforward, others are not. And until the full extent of the damage is understood, it is difficult to predict how long the repair will take. That uncertainty carries forward into everything that follows—planning, coordination, and ultimately the ability to return a car to the track in time.

Defining the handoff that sets the pace

This is where the focus of the project settled. Not on the repair itself. But on the decision that makes the repair possible. The solution was designed to accelerate that exact handoff in a way that could operate inside a live production system without disrupting the one already in place. 

What made this activity such a strong candidate for technology was that it sat at the intersection of diagnosis, inventory, and repair planning. A damaged car arrives in the garage and the first task is visual: the team photographs the vehicle from multiple angles and begins to assess what has been compromised.

Before the new system, that meant analysts and mechanics manually studying those images and the car itself, then building a list of damaged components one by one. Only after that list was complete could the request move into the parts process, where the stock team could identify what was available, what needed to be pulled, and what had to be prepared so the repair could begin. Every downstream step—inventory, ordering, labor assignment, repair timing—was waiting on that first human assessment.

Designing intelligence into an existing workflow

The Microsoft Azure AI platform, Microsoft Foundry, provided the foundation for improving that sequence without changing its logic. Rather than treating AI as a standalone tool, the system was built as part of the operational workflow itself: models, agents, and business data working together inside a controlled application environment. That made it possible to orchestrate the analysis step, connect it to the existing parts and stock process, and still preserve visibility into how the recommendations were being produced.

The process begins with computer vision. As soon as the car arrives, images are captured and analyzed. From there, a pipeline of specialized agents evaluates different sections of the car and compares what it sees against the known structure of the vehicle and its parts catalog. Instead of relying on a single model to interpret the entire incident, the system distributes the work across multiple focused agents, each responsible for a specific domain of the car, and then combines those outputs into a structured assessment. 

As Thiago Iacopini, CEO at Kumulus, explains, “We need a pipeline of agents. We need to create specialized agents for each piece of the car.” That multi-agent structure matters because the goal is not just to identify damage, but to produce an output that can be used operationally. The result is a proposed parts list—generated quickly, but not blindly trusted. It is reviewed by an analyst, who validates the recommendations, adjusts where necessary, and confirms the result before it moves forward.

From analysis to action, faster and earlier

This validation step is what preserves trust in the system. But once that happens, the important change begins: the confirmed parts list can move into the inventory and stock workflow much earlier than before. The stock team does not have to wait for a fully manual assessment to be completed before locating components, preparing them, or triggering the next step in the process. The information arrives sooner, in a more structured form, and that allows parts handling and repair planning to begin earlier as well.

Morrone describes the shift in simple terms: “With the system, we can take the pictures of the car and the AI already brings the list of parts. The analyst just needs to check if everything is correct.” It is not just that the list is generated faster. It is that the inventory workflow starts sooner, which means the parts can be checked, moved, and prepared sooner, and the repair team can organize its work with more clarity. What improves is not a single task, but the continuity between diagnosis, stock, and repair.

What used to delay the entire chain from damage assessment to parts handling is compressed significantly. But the more meaningful change is not the reduction itself. It is the consistency that comes with it. Morrone describes the intention clearly: “What used to take 20 or 30 minutes, now we can do in about five. We want to give more time to the employees so they can do their best work.” 

This is what changes for the team. Less time spent reconstructing what happened. More time executing the work that follows. Less pressure in the most uncertain part of the process. More clarity when it matters most. Because the parts list is validated earlier, the team can begin the inventory and ordering process sooner, giving them more control over how the repair unfolds. Labor can be organized earlier. Dependencies become clearer. And the repair itself starts with fewer unknowns.

For drivers, cutting repair time in half extends the impact beyond the garage. Pires says, “If we can repair the cars faster, we can bring more competition to race day and improve the show for the public.” What improves here is not just efficiency, but the opportunity to return to the track, to compete, to stay in contention. Dener’s perspective remains consistent with how the system was originally designed. The goal is not to change how the series operates, but to strengthen it—to reduce friction where it exists and allow the organization to perform at a higher level without compromising the principles it was built on.

Turning speed into predictability for teams and drivers

What makes this approach relevant beyond motorsport is not the domain, but the structure of the problem. In many organizations, the constraint is not execution, it is the time required to determine what execution should be. Whether in manufacturing, field service, healthcare, or logistics, teams often spend critical time assessing situations, identifying next steps, and aligning on what needs to happen before work can begin. 

For business and technology leaders, the lesson is not about racing. It is about where to apply technology when performance is already high. The instinct is often to look for transformation—to redesign systems, introduce new workflows, or automate entire processes. But in practice, the most effective improvements are often more precise. It begins by identifying where uncertainty enters the system—where time is spent understanding what needs to be done, where decisions are repeated, and where even experienced teams have to slow down to be certain they are right. 

In Porsche Cup Brasil, that point sits just before the repair begins. Not in the work itself. But in the step that defines it. Improving that step didn’t require changing how the system operates. It required understanding well enough to reduce variability, increase clarity, and allow the team to move forward with more confidence. In environments where performance is already high, the advantage rarely comes from doing everything differently. It comes from identifying the moments where uncertainty slows the system down and addressing those moments in a way that preserves everything else.

Discover more about Porsche Cup Brasil on Facebook, Instagram, LinkedIn, and YouTube.