People Behind AI: The Race to Better Insights Transcript
This transcript was generated using AI and reviewed by an editor.
AI is transforming business, but its greatest impact is on people. Welcome to People Behind AI, a series exploring how AI empowers people, solves real problems, and creates new possibilities in the way we work. In each episode, AI leaders share what they’re seeing, what they’re learning, and practical ways to put AI to work.
Pitlane IQ and Motorcycle Racing
Today, we’re with Jordan Chandler from BDO’s advisory AI team. Jordan has been doing some amazing things with Arch Motorcycle racing and their team. Jordan, welcome. Hey, Kirstie. Good to be here.
Great. And I know you just got back from the Mid Ohio race with the team. Exciting times. Tell us a little bit about what you’re doing with them.
Yeah. We had a good race at, yeah, Mid Ohio Sports Course. We were there to kind of help them with their bike data, using Pitlane IQ, which is BDO’s, AI race platform. Yeah. Just had a great weekend. Found a lot of issues that we could fix with the bike using the tool, and, it ultimately helped them, you know, gain some performance in the races.
Yes. You’re let’s just go back to the beginning of what you’re doing. You are helping motorcycle race teams, specifically Arch Motorcycle, deploy AI to make better decisions. We build a really great tool that allows us to take in that data and understand it. How would you explain the the benefits to the rider and the crew to someone that isn’t really familiar with racing?
Automated Data Agents for Performance
Yeah. So, you know, it’s it’s really this tool that allows them to, you know, really get another perspective of their data. So in this sense, it’s, you know, motorcycle, data that, you know, all these sensors on the bike. Right?
And they come off the track, and they wanna look at, you know, how was fuel pressure, how was oil pressure, you know, how did, air flow into the bike. And so this tool can really help them, you know, get a get another perspective and uncover things they may not have known just from a, you know, human analyst. Right? So it’s, you know, it’s all kind of this automated data agent that can really, you know, get a really find interesting things, you know, that’s going on with the motorcycles and hopefully lead to some performance gains. You know?
Yeah. It’s kinda like if it you had all the humans in the world looking at all the data that you had available and trying to identify every pattern, every correlation, what could you surface up. And you’re able to do that in real time very quickly at the track with the team, which is a huge benefit.
Yeah. You know, it’s it’s like speed is everything in motorsports. So, you know, that’s another kind of key benefit of this tool is that instead of, you know, hours looking at, you know, telemetry data, it’s really a matter of minutes now. You know, the the the process is much more efficient and quick.
You know, it’s it’s whole the whole thing is trained on, all of their, you know, healthy bike session data, and we kind of create this model that then helps them, you know, uncover anomalies that happen in the bike and also just answer questions, you know, that they have about performance related issues.
Machine Learning in the Pit Lane
Well and and I think we’re we’re living in this incredible era of AI development. And, usually, we think of AI living in server rooms or corporate offices, but you are throwing it out onto this loud, unpredictable racetrack. And then, you know, just talk to me about what was some of the most exciting and maybe AI specific challenges we could the the team just really couldn’t wait to get their hands on when we’re talking about bringing machine learning machine learning into the pit lane.
So, I mean, I think one of the really kind of exciting things was just that, you know, it’s it’s a a lot of the times in AI and data science, you’re working with data that isn’t really tangible. Right? You know, it’s marketing data.
It’s economic data. But with this, you know, it’s it’s a all the data kinda ties to an object that’s right in front of you. Right? It’s this motorcycle that’s incredibly powerful, and fast, obviously.
And so I think that was kind of the most exciting thing is just seeing, you know, a one to one from, hey. The data was saying this, and, you know, the bike was also saying that. You know, watching the bike, you can see it act a certain way, and it’s reflected immediately in the data that you look at or the AI can analyze.
High Speed Anomaly Detection Challenges
Also, another thing was just that, I think it was kind of a challenge for the team, but something we were all kind of excited for, which was that anomaly detection is you know, it’s hard on, like, a factory machine. Right? You’re detecting anomalies on a factory machine. But now imagine that that factory machine is moving around the factory at a hundred eighty miles an hour.
So it’s it’s kind of this whole new challenge that the whole team was kind of excited to take on. And so yeah. I mean, it was Yeah.
And and there’s so many things that can go wrong too. It’s not just you get this clean data every time, and it’s all perfect. Data gets corrupted. Data gets sensors get left off.
Identifying Faulty Fuel Injectors
All kinds of things can happen to make things really challenging, and you need as much of that good data as possible to be able to make an accurate prediction. Maybe talk a little bit about, you know, in the very early stages of this and when you’re building the model, what was the first quick win where maybe the AI predicted or highlighted something or spotted a pattern where everybody said, woah. Yeah. It knows what it’s talking about.
Yeah. I think pretty early on, one of the issues they had some issues with their fueling. Right? So, even before we had all of Pitlane IQ’s, you know, app fully flushed out, we had that model.
So that model was already kinda going, and, it was able to surface some issues with, you know, their fueling behavior in one of the bikes. And it kinda pointed out that, hey. This is not normal.
And from there, they kind of investigated further and were able to realize that one of the fuel injectors was, completely faulty even though there wasn’t any visible sign on the outside.
They were able to replace it. And turns out that was kind of the leading issue for the poor performance in the bike.
Yeah. I think for them too, it’s it’s almost like a virtual another team member. Someone that has opinions that may you know, they may not have thought about or they haven’t had enough time to process everything. And it’s really exciting when it throws out something that the team isn’t thinking about and that that becomes something really relevant.
Building Human Trust in AI
I mean, at the at the end of the day, you know, I think a motorcycle requires a rider, requires a crew that rely very heavily on this physical feel. I you know, they come off the track, and they’re always making these noises about the the sound that’s making or, you know, there’s not a a lot of descriptors other than, you know, those kinds of things. A lot of muscle memory, lot of gut instinct. And that type of environment and it being so fast and intense, how do you ensure that how do how do you build the AI so that veteran mechanics that aren’t used to using AI actually trust it?
And I think that’s been really interesting to see them adopt it, and they’ve been really great at it, very welcoming of it and interested. But were you nervous about that?
Combining Machine Learning and GenAI
I mean, that’s obviously, you know, a huge thing is trustability and, you know, explainability in AI, but most specifically machine learning. Right?
You know, like machine learning, a lot of the times, it’s a black box. There is no it can tell you something, but a lot of the times, you won’t really have an explanation for why. So I think that was kind of a thing we highlighted on early on just to say, let’s focus on this. And kind of what we ultimately ended up with was, we combined machine learning and GenAI. So on one hand, you had the machine learning model kind of saying, hey. Here’s an anomaly, but then we combine it with GenAI where you get, like, a summary, that looks at the anomaly results and the data and provides a interpretation using a language model.
So that took care of the explainability portion. And then for the kinda trustworthy, part, that was more so just gained being there on the weekends this whole season, you know, using the tool with them, you know, working you know, being there every weekend, every Saturday, every Sunday, every Friday, you know, correlating what they’re finding, saying, hey. We found the same thing.
And just building that trust over time using the tool with them. I think that’s kind of just been the the the the way we did it.
AI Change Management and Implementation
That I mean, there’s nothing better with change management around the AI than hand holding and being a a face to the AI with people. Right? Being able to be that that intermediary, not only explaining what’s happening, but also being able to fix it, adjust it very quickly so that you don’t erode that trust, and there’s just no one no explanation there or no no one there to help explain that. And then, also, I think, you know, the team understands you’re there with them. You’re in it with them. You understand what’s going on. And sometimes even things that they may not think to tell you or the AI model, you can pick up on and add add independently, which is a really big deal as well with everybody moving so fast.
So well, I know we we’re doing everything from, you know, predictive maintenance one minute, back office AP automation the next minute, and then you’re out on the racetrack on the weekends. So it’s been kind of an exciting thing to see, and thanks so much for sharing your experience with Arch and the team, and good luck in the the future races. Thank you.
Great. So just thank you everyone for joining us and being a part of People Behind AI.
Leave a Reply