The AI Advantage in Automotive Service

Discover how AI is transforming the automotive world with our host, Mike Chung and his guest, Eric West, co-founder and chief growth officer of Convertible AI, as our special guest. Learn how Convertible AI, in collaboration with the National Institute for Automotive Service Excellence, is setting new AI benchmarks to empower technicians. We venture into the broader implications of AI, examining its influence on competition, employment, and the right to repair movement, which could redefine vehicle repair and maintenance practices.

We navigate the evolving landscape of automotive data, spotlighting the challenges and opportunities brought about by electric vehicles and diverse vehicle architectures. With supply chain hurdles extending vehicle lifespans, the conversation turns to the lucrative potential in supporting legacy vehicles. Emphasizing the importance of high-quality data and strategic weighting, we discuss how AI can revolutionize data use, enhancing safety and efficiency in maintenance by considering factors like technician handedness.

Join us as we explore AI's role in bridging the knowledge gaps between data sources, improving diagnostics, and streamlining the repair process for technicians. We address concerns about AI replacing jobs, asserting its role as an augmentative tool that complements human expertise. Delve into the exciting potential of synthetic data to simulate real-world scenarios, filling gaps in datasets, and learn how AI is poised to attract the Gen Z workforce, revolutionize product development, and advance proactive vehicle diagnosis.

To learn more about the Auto Care Association visit autocare.org.

To learn more about our show and suggest future topics and guests, visit autocare.org/podcast

Chapters

• 0:08 Automotive Industry Trends and Innovations
• 10:00 Challenges and Opportunities in Automotive Data
• 15:36 Improving Automotive Data Quality With AI
• 28:10 Future of AI in Automotive Industry
• 37:22 Creating Synthetic Data for AI

Transcript

Speaker 1: 0:08

Welcome to AutoCare OnAir, a candid podcast for a curious industry. I'm Mike Chung, senior Director of Market Intelligence at the AutoCare Association, and this is Indicators, where we identify and explore data that will help you monitor and forecast industry performance. This includes global economic data, industry indicators and new data that will help you monitor and forecast industry performance. This includes global economic data, industry indicators and new data sources. Hello everyone, and welcome to another edition of AutoCare OnAir. I'm Mike Chung and I'm really happy to introduce to you Eric West. Eric, welcome to the show.

Speaker 2: 0:46

Thanks for having me, Mike.

Speaker 1: 0:47

Eric, tell us a little bit about your role and what Convertible AI does.

Speaker 2: 0:53

My pleasure. My name is Eric West, co-founder and chief growth officer of Convertible AI. We help automotive clients understand the possibility of what AI can do for them, and then we develop systems that help them basically do strategic initiatives that, let's say, extend their capabilities into the future and accelerate the way in which their company grows. So, with that in the context, you know, the aftermarket is something we've been focused on since my co-founders and I came from Ford Motor Company and it's through that kind of work, based on the convergence of what technology and strategy means to the future of, of course, all of our friends across OEMs and on the aftermarket. So, yeah, I'm excited to share more.

Speaker 1: 1:37

Oh, that's fascinating, and I can imagine that there are a lot of different and many exciting things going on that you're working on and have introduced to the market. Are you able to tell us about some of those things?

Speaker 2: 1:50

Absolutely. You know, in some cases clients come to us with you know, an obvious opportunity. Some they don't even know where to start, right, and so that's kind of where our special sauce sits of, like what we call discovery. It's how we start every engagement Through that process. Our proprietary methodology helps them understand, like, well, you know, what helps you most is the better question, and they alone don't necessarily or not necessarily equipped to figure that out alone, right, and it's through that that they see we as alumni again of Ford Motor Company, and let's say several others before that. They see we as alumni again a Ford Motor Company, and let's say several others before that. We help them see that. Let's just say things like whether it's supply and demand forecasting, right, there's no longer an or but an and discussion happening Through that. There's things that help them accelerate the way in which they bring new sources of data together and that, let's say, opened others' minds into well, wait a minute, if you get up our supply chain, what does that do for our parts organization, right, and laundry list of opportunities that follow. But that is a full circle point. Come back to we recently announced a partnership with the National Institute for Automotive Service Excellence and our friends at the ASE, who is now led by the chairman and president, dave Johnson, who is a familiar alum of ours from Ford Motor Company. Together we are helping develop what starts, as they've done for decades now an industry collaboration through developing what's called AI benchmarks for aftermarket maintenance, and really what that comes down to is the way in which companies use tools of any sorts right of figuring out. You know what does it mean to help technicians do their great work. And in the case of AI I'm sure you guys are also seeing this that there's people coming out of the woodworks from Tesla and the likes that says, you know, we built this AI system helps technicians. Coming out of the woodworks from Tesla and the likes that says, you know, we built this AI system helps technicians.

Speaker 2: 3:47

But it dawned on me at least a few months ago hey, wait, if they keep doing that without the ASC and the industry, in that case it will end up in the same place the ASC or the industry started back in the 70s, when the ASC was formed. Right, it was all about this idea that the industry didn't know how to work together. Develop what becomes these industry standards for the way in which, let's just say competition on the let's say, the employment sector that these technicians know. Let's say what it means to have transferability between opportunities. Right, it develops what becomes a more competitive market in that world, and so it's through that that we say well, wait a minute. Why is this any different? Actually, because if these AI systems help companies in different ways, it's going to further broaden and exacerbate that gap between what it means to work for, whether it's Bridgestone or some mom-and-pop independent shop.

Speaker 1: 4:38

So pause there. Sure, that's fascinating. I'm thinking about all the data that could be collected. Sure, that's fascinating. I'm thinking about all the data that could be collected, say, based on technicians' experience diagnosing a vehicle for any issue that it was having. Things that were done. Is this being is one of the use cases? Perhaps feeding a database so that you have a? Oh, I go to the system with my car is having issues with this? What are some of the possible causes and solutions for it? Is that one of the potential use cases, eric?

Speaker 2: 5:25

Absolutely. You know I can't help, but of course, bring up in this context what I think you guys know more than anyone about. Is this right to repair legislation in the background, right? And in that case, that's what every OEM, especially, is aware and anxious about. Is what does that mean to our future, right? And in that case, what do we develop as the strategic moat that will have fallen in the process of the proprietary repair manuals and such and so through that, as a broad stroke behind it, yes, this idea of like well, what are we starting with? What's the prompt?

Speaker 2: 5:58

And even down to the consumer side, it's like well, consumers are more articulate than ever about. Like well, I know my car. I can plug into vice right, the EPID and whatever this OBD2 port generates. Well, does anyone even know what a P406 is? And should they even? But bottom line is, it doesn't necessarily matter anymore going forward, because there's much more to that that the OEMs have kept behind this Walt Garten until now. And so it's through that partnerships with OEMs that we are able to unlock intelligence, right, that intelligence comes down to democratizing into the hands of those that are capable of doing something, let alone being more well armed to do something about right, and it's everything from the ultimate conversation they have with the repair shop and technician down to what they know as a DIY themselves right. So it does come down to ultimately, how someone better articulates what their level of confidence is going into that conversation.

Speaker 1: 6:54

So when you say somebody who can benefit from that knowledge I'm probably not quoting you correctly, but does that include the technician who's part of ASC, who can kind of unlock the door to that body of knowledge then, so that he or she can kind of gain access to that library of data?

Speaker 2: 7:13

Absolutely so. Yes, and where we look at it and these are just broad slices of this whole opportunity it's as if in one very realistic environment, there are some independent repair shops who only focus on Japanese makes, right, and in that case this is an opportunity for that same independent repair shop to open their doors to German cars hypothetically right, not that they don't already. But this also starts opening the door for a much more mature approach to showing all their audience and building certainty in and they could be trusted to do more because of the underlying reasoning and trust built at the local level, of what they've shown, they've been able to do with those and how to be extended in a new direction. So it's just about opening the doors in different ways depending on where they come from.

Speaker 1: 8:04

So it's just about opening the doors in different ways depending on where they come from, and I'd imagine there's sidebar conversations you could have because a Japanese make specialist perhaps they can diagnose a European make, but then it becomes a question of parts and inventory, things like that. But it allows them the freedom to explore that if they want. It sounds like.

Speaker 2: 8:25

That's right and I can tell you from my time at Ford Motor Company. They have an entire division about parts and several subdivisions under that, one of which is called Omnicraft, and that opened my eyes, having worked with that team leadership for is understanding there are entire parts companies and supply chains they're under that are focused on developing well, parts for every single make and model, right, and so there are entire holds.

Speaker 2: 8:53

And so there's an entire subculture looking for, I believe, support for getting well, let's say, more funnel filling activities, of the ways in which they become a partner, more types of, let's say, distributors, let alone the actual repair shops that are looking for a shared interest, which is finding new clients in new ways.

Speaker 1: 9:10

So you touch on something interesting, eric, because I'm thinking about how this could feed back to the OEs in terms of benefits. What's in it for me, right? And the first thing I thought of was well, if we as an OE understand the problems that vehicles are having with particular parts and you can slice and dice that database to make model and part, perhaps that can be a positive feedback loop to engineering the next generation of vehicles and parts so that fewer problems are avoided. One, is that a valid use case? And then two, what you were just highlighting sounds very intriguing, right? So can you expand on that a little bit more?

Speaker 2: 9:58

Happily.

Speaker 2: 10:00

So, yes, and again, here we go back to what the elephant in the room there is electric vehicles, right, so that's a whole different ballgame when it comes to the diagnosis underlying architectures that support forecasting, the ability to do predictive maintenance, and, at hand gets the more complicated the diagnosing process gets too right and because of the well, depending how far you go back in whether it's five, 10 years out of vehicle age you start to see there's a proliferation of different vehicle architectures right In embedded systems.

Speaker 2: 10:37

So, as we start going down that line of thinking, it's a proliferating challenge of maintaining vehicles as they get older, across a broader number of potential configurations of those TCUs and ECUs. And so that's where I come back and point out that's not going away. In fact, through COVID, especially in that supply chain challenge, from that there's a well, there's already data out there to prove that people are keeping cars longer than ever. And so this is actually what we believe to be a partnership for OEs, right, they're looking to say, while they invest heavily in well, to say, current new vehicle maintenance solutions, we're talking at first about legacy vehicles and in that case it's all about supporting them at arm's length, through which it still fills the same funnel In center core profit center of parts in the aftermarket. And so that's where I think this the most exciting profit generating opportunity is sits is for those parts businesses that just want to show where the margin well grows right, that literally is higher volumes at higher margins.

Speaker 1: 11:43

That's fascinating. And how much data is being collected? Because I'm thinking about what you just highlighted being able to make profitable business decisions on a growing body of data, and I'm thinking, as a statistician might, in terms of what's the appropriate sample size and how much confidence is there. How much confidence is there? Can you talk to us a little bit about how large this database is? What percentage perhaps of the vehicles on the road today might be represented?

Speaker 2: 12:13

Can you just kind of dive into that a little bit for us. Yes, so a lot, right, you know this too, mike. So there's a lot of data. In fact, the volume of data has never been the problem in a weird way, but rather it's the particular data about you and your vehicle, now that is, and so that little of being served and serviced in real time. That's the opportunity also, and so I say to point out, part of our problem statement includes how this data in fact is so again, no news to you guys siloed, and your listeners understand this better than anyone.

Speaker 2: 12:51

That siloing is part of both the opportunity and the problem statement, because it really it's come back to this idea of like data quality right, and data quality is the most both talked about and underappreciated opportunity for AI, where it comes back to how do you really know well, you're consuming the right data and when for what right, and that's come back to in the world of artificial intelligence. This idea of like waiting and waiting is a strategic discussion all of a sudden comes back to why something like an industry standards body can needs to be involved in to make sure the industry co defines it for themselves, but it definitely means like certain source of data are more valuable than others and that that's not a static thing. It really depends. And back to, I think, what you brought up before even down to the individual technician it matters if they're even a lefty or righty right, you should be recommending a different, sometimes technician literally, for certain things based on that. But also like back to like well, core shop interests of like survivability of their employment base, like if you want to make sure that there's like zero tolerance for employee issues, whether it's them being hurt in the job or, let's say, challenges there, and you end up really having to think deep about what causes those that puts them at risk of of these kind of challenges in their health and things like that, as that's over literal, but like if they're left ear, righty, they'll reach differently than they would and putting them potentially in harm's way, and so that's the kind of things that these source of data back to, lefty or righty is oversimplifying it, but that's the kind of things we can consume now at the individual level and start developing systems that put people in compliance, no longer at this like annual or whatever it is six month check in that the ASC certification type technicians work requires, but rather on the daily level, right and saying how could this, this job you have in front of you, be part of what becomes the up, upskilling, reskilling opportunity, as the generative interface begins showing them things that they hadn't come in contact with before? Right, and that's where its purposebuilt for consuming as a full, let's say full-circle, point of what this opportunity and challenge is is the legacy vehicles, not the connected vehicles that we have today, right, showing them that there is no great, let's say, central, database of all the, whether it's in the world of AI, it's multimodal, they call it right.

Speaker 2: 15:26

So these modalities visual pictures are one modality, text is another, voice is another technically, and so, as you go through that, even vibrations right, that's technically a modality. And is the car vibrating? That is where this whole basic user experience starts. It's like you go to your local technician, say, hey, my car vibrates when I go on the highway. Well, okay, but which way does it vibrate? How bad does it vibrate? At what speed does it vibrate? You have all these questions. That's where you need a really experienced technician to start saying well, hey, have you tried this? It's no longer a look in the manual but, as we know, that's the spectrum we're working on is those that just get certified. They're looking at the manual for everything right. It reminds me of that.

Speaker 1: 16:10

Amco I'm sorry, it reminds me of that Amco commercial, I think it's Amco where the person is in there and he or she's like our car goes, like right, and it's that type of a thing I feel like. But more seriously, or I guess one of the things I was thinking about is just to clarify the source of the data, so you could be getting data from the vehicle itself, but the examples you were talking about are from the technician.

Speaker 2: 16:52

So can you just confirm that? So is this if I'm a technician, I'm putting in my observations into the store computer, and is that one of the data sources? Then, absolutely yes. So on the commercial side of things, when you talk about a Firestone or pick any aftermarket maintenance business, you end up seeing that they have their own system, the shop management systems, and that's a whole other, frankly, industry, right, but in that relation to they're also faced with the same siloed data challenge where, let's say, the vehicle information hypothetically repair information sits on one system, but then the actual technicians interfacing with the shop system which, let's say, communicate, let's just say, underwhelmingly.

Speaker 2: 17:30

So it's through that kind of thing that we're saying not only can we consume information from the technician real time, how they're thinking about, in fact, through almost this generative interface, could you start asking them what's the next best question to ask is like lit, a literal thing.

Speaker 2: 17:45

That is an opportunity for AI, which is a really challenging but huge opportunity for understanding how to what is broadly called embed the reasoning that a technician applies to diagnosing issues, and we're talking about the root cause diagnosis. Right, that's always the goal, and so that's what brought strokes here again, but that's the idea of. One source of data is the actual technician. The other is both literally repair manuals, but also the manuals of what a certain part number comes along with in the aftermarket. That if you're getting the parts from Castrol or from Ford, they likely have different ways of talking about the installation, have different ways of talking about the installation, right? So we're showing like you could bring all that together and not say that either one is right or wrong, but start to bring it in the context of what is the master technician who's done this for 35 years? Focus entirely on Japanese makes as an example again. What do they do to make this work for them? Right?

Speaker 1: 18:46

And that gets to the weeding issue.

Speaker 2: 18:48

perhaps Exactly right, and that gets to the weeding issue, perhaps Exactly Right, and and microcosm of that is what is happening already, with technicians that literally go to YouTube. Right, and they'll go to YouTube and say, oh, how do I replace this part? And they type it in, and we've heard this, we've heard this directly that they end up saying, oh, it's a 2014 Honda civic break replacement, blah, blah, blah, blah, blah, right, but then they go through the watch the video, do part per part, whatever they need, but then at the end they realize, oh, but that's mislabeled, that was a 2015,. But that means it's irreversibly done with something that was not correct in the first place. Right, and that's a microcosm.

Speaker 2: 19:37

But, like you start to see, this data quality thing is a big challenge, and that's also what shows that the moment that, well, that underlying data evolves right, whether it's new attributes introduced or however they want to start tagging these things it breaks these static, conventional systems in a lot of ways, which is why there's a big hesitancy and, let's just say, there's a lot of inertia in updating these legacy systems. For that reason Right, and so it's frankly, why the SAP, the infrastructure around that marketplace, is so robust is that it's very costly to do that.

Speaker 1: 20:09

So, if I'm hearing you correctly and please correct me, where need be your organization you're able to access data across a number of silos, whether it's one shop management system or an OE getting data from. So, the shop management system being fed by technicians in local shops, the OE perhaps getting data from its onboard computers, and that's where I could see an AI solution to be able to take the best of the data. I don't know if normalize is the right word, but to look across and find to, I guess, have algorithms to make up for those differences in how something could be categorized. The data structure might be defined in one system differently than another system. So I can see where those challenges can be right Because, again, excel spreadsheet, a column for year, a column for make, a column for part, but there could be a mislabeling.

Speaker 1: 21:17

There could be an open-ended text field in terms of I installed it this way, or is there a column for right-handed, left-handed, you know? Is it a winter? What's the environmental conditions? It's not going to be consistent from system to system. So, if I'm hearing you correctly, what you and your team are able to do is to use AI as a way to bring together those sources of data to come up with, I guess, a better system to solve those problems.

Speaker 2: 21:51

That's right. So, yes, and again, here we get back to what I think we're helping these companies overcome is both the problem and the reality of working with these static systems, right, they're not built to accept new types of data, let alone well, data that's tagged in new ways. Even right, the same data in that case and so that's kind of where the opportunity lies is saying it depends, right, it depends on their goal. And that's where I come back as a well to the alumni of Ford Strategy Group. I can sit with these executives and say, well, what do you care about? Right, in one case, an energy company acting in the mobility sector.

Speaker 2: 22:31

For, in this little case, maintenance for automotive.

Speaker 2: 22:35

They care much more about things like the, the individual daily consumer experience that informs the decision of when the right time is to get their cars maintenance completed.

Speaker 2: 22:48

And, as a literal example, and how it depends, it's not just about what the car needs and what a technician needs to know about fixing the car, but how does the decision of when the right day is to go in to get that service done get informed by climate data in their case, right, and in that literal case, we're pulling in data about if it's a catalytic converter, hypothetically, that is no longer in compliance.

Speaker 2: 23:11

There's a way to objectively measure how much impact on a daily basis you are now causing in a negative sense, the environment, in detriment to what it is. That is, the, let's say otherwise hidden cost of continue to drive that car without being maintained. And so that's the kind of thing we said, well, what would you and your customers value most? And that's really what we're all about. It's no longer this purpose-built idea of like it's just for technicians, because absolutely, don't get me wrong, but it's multivariate in so many ways. It's much more nuanced than ever about the types of customers, what they care about, and we can bring them all into what ultimately becomes requirements for development, development of a system for them. Ball right, it's all about one or the other.

Speaker 1: 24:36

So what I'm hearing is thanks. Thanks for expanding on that, eric. So perhaps one use case could be if your client is an. On that, eric, so perhaps one use case could be if your client is an OE, it gives them the information to go to their customers to say, based on your vehicle's history and climate conditions. Currently, we recommend that these things we have seen that these things could come up, that these things could come up If your client is a service provider, perhaps another use case could be. Here is a tool that your technicians can use to perhaps engage with, to get, to access a wider body of knowledge on best practices for diagnosing and I don't want to say treating, that's the wrong word but to repairing a vehicle, right, yeah, that triaging exactly.

Speaker 1: 25:37

Right, so are there other use cases? I guess it could depend on the client of your system, right?

Speaker 2: 25:44

Absolutely. And actually that's an easy segue into yes and again. We're also talking to similarly analogous OEMs in the HVAC space, because same underlying technician challenge and same underlying OEM's interest of building walled gardens in the world. That's upcoming up against. Give me the right to repair legislation that your, in fact your residential condenser. That's really complicated too, and their incentives are to sell that same make and model or to say that another model, that same make, um, but the way in which those technicians are trained are are less, um, consistent than ever. Right it's.

Speaker 2: 26:33

And it's simply because this very similar to the automotive aftermarket maintenance world, uh, mentorship models have been eroding for years. The access to data has been becoming more and more siloed. It's through this idea there's way more questions than answers about where is it all going. And so sad to point out, like the automotive sector has the ASE, which is that developing benchmarks. And for technicians, we're talking to the HVAC industry's analogous group called NATEX, and N-A-T-E is basically that literal say, aftermarket technician certification team that talks about technicians' excellence. Similar way, so through this idea that we could do this for any company that's servicing and supporting technicians for what we call simply embedded systems, and HVAC has literally that just like automotive and lastly, aviation, which is where my career started. So talk about the way in which planes and helicopters have been serviced for years. It's a lot of opportunity to address a similar, let's just say, aftermarket support and, in this case, uptime that they require. They have a very different customer base.

Speaker 1: 27:46

So complex systems where you have a technician shortage and where additional assistance could be had for maintaining vehicles or appliances and so forth. It sounds like there are a lot of parallels. That's fascinating. Appliances and so forth Sounds like there are a lot of parallels. That's fascinating. And what are you hearing from across the board? Whether it's technicians? How are technicians responding to this? How are shop managers? It could be. Whether it's an independent shop, a dealer shop. What's some of the feedback? Are there concerns?

Speaker 2: 28:22

Tons. Yes, I think so much of the anxiety about the market for AI has all about the risk, obviously, that it will replace people in their job. We always start the conversation with we're just here to augment you, not replace you. Right, and starting there, it starts to get them to see because we're only practitioners of industry, we come from the world in which they operate that we can show them quicker than most. Hey, we believe you Like, we believe you need help, and that help is not, well, just handing you more money, but rather handing more opportunity, and that's the thing that we kind of get them indexed for is, for that reason, what ultimately unlocks their trust.

Speaker 2: 29:08

The final straw has been how we show them. You know, the answer is not just focusing on recent graduates and helping them grow in new ways alone, but also helping recent retirees, who, in fact, have the most amount of knowledge and reasoning embedded in their brain, but no way to communicate that to someone in a time of need, which is the recent graduates in a lot of cases. And so this idea of like building infrastructure for the thing they've always wanted but are struggling more than ever to do, which is ultimately democratize access to those systems through better infrastructure are the ways in which they help each other right. So bringing people together through this technology has been shown that that does open people's eyes. When they then hear, as a final point, that we could do it in a way that well builds that margin story in a better way, that accelerates service operations and whatever their P&L initiatives are, that's where they say, well, hmm, interesting, let's talk more.

Speaker 1: 30:05

That's fascinating and side note here, you mentioned that knowledge retention sort of that benefiting from the gray hair, if you will and how does that knowledge get passed on to the subsequent generations? I've seen that in engineering consulting as well, and certainly there is that opportunity. That makes perfect sense for the automotive industry. And I think we've talked a bit about data. What are, say, technicians' concerns or responses in terms of some of the things we've talked about? Are they going to have to enter data a different way? Is there an implication for shop management systems? Can you talk about that, eric?

Speaker 2: 30:45

Sure, that's back to it depends. I'll start with that last point, the shop management. There are several shop management tools and platforms out there, as you know, and it depends. We're happy to integrate with any of them and most of them are built with an API forward architecture, right. So that is the relatively more straightforward part, whereas figuring out what is sitting in that, that. Back to the data side of things what warehousing have they invested against? That's where it's gonna be interesting to see what we can pull out and tease out from what they've we can glean from that, right.

Speaker 2: 31:16

And then that, combined with the fact, not every technician thinks the same way, that's very clear, but those that see the opportunity to show that the way that they do their great work could be valued more broadly, there's something to that right, and I oversimplify it for some of these conversations just to get to the point that you know, what we're ultimately building is a mosaic of the most valued technicians' work, right, and we want everyone to be represented in this because that's the fact, right, that's not new or going away, just because because, as you do your great work, you know the senior technicians, master technicians.

Speaker 2: 32:00

We don't want you to, we don't want to lose you, let alone the way in which you clearly earned the stripes, and say, well, we want to embed the way in which you think into the model that gives those next generation the opportunity and that, as a final point, is showing well, this even is really attractive to the Gen Z audience that we're talking about right. In fact, as we keep talking about these well paths to certification, it's clear that there's a change in higher education and trade education at these younger generations and as they start seeing the opportunity to work with technology, to do whatever their job is, that's attractive, right, and that's almost what I use them as the leading indicator for, it being the future of work. Pretty broadly, it's working with technology. So that's, I think, builds a competitive advantage of there becomes a certification at some point in the near future for what it means to work with technology best right, and in that case it case the job with as opposed to the job for.

Speaker 1: 33:03

Right, right, Thanks for highlighting that. And when you look out into the future say three years, five years, 10 years from now, what is in the realm of possible here with regard to how some of the systems you've described, how much they grow over the coming years?

Speaker 2: 33:21

Well, both that's a great question on its own, but then also a great way to bring this full circle back to the beginning of one question you asked that I skipped over. Strategically, it comes back to the OEMs and product development, right? I think cars, once they get to this, gets to a point where this model, or, let's say, set of models, can be brought together in a way that actually does the proactive diagnosis of a given vehicle architecture. That should greatly inform the way in which vehicles are developed and architectures are developed going forward. Right, and I think that's going to be one of the most underappreciated points of OEMs and that strategic return they get is how their core product development activities get informed by this right and generatively created based on all the things they've come to know about the underlying reasoning that a technician, as a product of all those vehicle architectures until this point, have learned about what works best or that matter doesn't, in the process of maintaining a given system that matter doesn't in the process of maintaining a given system, and how, without you know.

Speaker 1: 34:25

This could be a whole different conversation. But from an AI model, enhancement and recalibration or maybe enhancement's the better word how do you expect your AI solutions to continue to improve over the years? To accommodate those solutions and more?

Speaker 2: 34:46

We'll see in real time, right? I don't have a magic wand in that case, but I can tell you that there are some future-proofing activities as well. Our CTO my co-founder it's his favorite activities to become. What we joke about is, instead of a CTO, for chief technology officer, chief tinkering officer, right. It's about this idea of like, how can we be on the frontier of whatever it is that would be the next best tool objectively compared to what it is? We built to that point, right.

Speaker 2: 35:25

And that continual pressure testing against these feature developments is part of what has gotten us to see whether it's literally this rag pipeline, If you've heard that term, the idea that retrieval, augmented generation is that interface part where it says it knows almost what to query for the context of bringing up, in a certain context, conversation.

Speaker 2: 35:39

That's one module of this whole thing that leads someone to see well, that's just the current state. What ends up being able to be done when you have this really robust conversational interface and it's way beyond what we have come to know as this natural language processing, NLP, right, what that has been for years. It's not new or going away, but rather that through all the data you start to consume down to the individual level of again what that specific technician knows about, not just the right or lefty, but remembering in order to generate hey, you're righty, why don't you grab this with your left hand and then reach over there with your right right? That kind of thing is what makes everything become customized to every single technician every time they use it, and I think that's going to be a lot of the special sauce, no matter what next tool comes of showing them that this is all being invested in for them.

Speaker 1: 36:33

And one last question in terms of the technical aspect of our conversation. You talked about data quality and some of the examples just now right, left-handedness as challenges. Are there any other challenges that you and your team have either encountered or foresee and it could be. Maybe it's a structural limitation, whether it's energy or computational ability, or there's just too much data and it takes too much computation time. Are there any hurdles that you and your team have to cross or that you foresee in the coming years?

Speaker 2: 37:08

Several.

Speaker 2: 37:10

Yes, one that comes to mind, and this is a solution to a problem I'll start getting into. But the idea of synthetic data creation, synthetic data is a solution to a problem I'll start getting into. But the idea of synthetic data creation, synthetic data is a pretty widely discussed term at this point about the idea of just creating data right and it's done strategically to plug holes in data sets, and I think it's one interesting aspect that comes along with the process of developing AI. It shows pretty objectively what a given data set is missing, not just what it is or isn't right. And that idea is like what we've come to learn. It goes through the harder conversation of saying to a given executive you have some of the right data but not all of it, so could we help you then augment that and show you we can build the representative database you need in order to move forward in the most strategic way. And that's been a harder conversation than we've expected, as we think that objectively, could make most sense but doesn't necessarily resonate nearly as deep as we thought.

Speaker 1: 38:11

Can you give me an example of what a synthetic data point could be?

Speaker 2: 38:15

Absolutely For discussion's sake. If I was technically to work with my town here in Connecticut and with the challenge of doing something about the heights of every single person in my town I don't necessarily need the heights of every single person in my town to do that. Instead, I can make a representative data set, generate, in fact, that data set, simply knowing the town's population. It's measured in feet and inches and it generally follows a bell curve, right? That alone could help me generate a data set that is representative of the way in which I would do any analysis on it, data set that is representative of the way in which I would do any analysis on it, right, so that I can then prove whatever it is that the goal is and then show them that if I could do that for this, imagine we do for yours, right?

Speaker 2: 39:09

And in that case it's about generating something that I think the term representative is the most foundational element here, because it shows like we can make something that as if it was real, right. And just because it's not real doesn't mean it's not representative. So that's what we start sowing them. Just because you only have data about the 70% of whatever we're looking for doesn't mean we can't make it 100%, and that's the opportunity we can do. Come back to it's just step one, right? It's just a starting point through which we start consuming more actual data to then augment and replace, right, what it is that we've built at that point and that's what this becomes a living, breathing experience on the backend side. For that reason, engineering background.

Speaker 1: 40:02

So what I'm thinking about is like iterative solutions. You have this really long chain of equations, a whole bunch of variables in the beginning, but you have to make an estimate for let's just say delta, and let's say delta is your height of those individuals in your town, and you start with an estimate, let's just say five feet nine inches. You start with that as a starting point. You run your calculations, you get your solution and it helps. You say okay, well, maybe delta should be five feet eight and three quarters inches. And then you run it again and you get closer to your analysis or your solution. Am I thinking about it the right way, Eric?

Speaker 2: 40:39

Mike, it's like we went to the same school and program. Yes, exactly so. Yes, exactly right, and I say that all to point out. It depends again, it depends on what the goal is and, ultimately, what we're there to work with and what's the problem you're trying to solve right.

Speaker 2: 40:56

Right, and this is consulting 101, but, like, that's literally where we start our work. That leads to the opportunity that we show them and basically hand over to them, saying that this is how it would be done. If you wind up doing it, of course we're there to help you, but, like, at the end of the day, that's ultimately what we've proven to be most valuable to them.

Speaker 1: 41:13

And I can imagine where we talked about some things that perhaps may not be captured within the shot management system. So let's say it's weather, but you have the date and you have the time of the repair, presumably, but perhaps it's okay, I'm going to add to my data set data from NOAA for weather climate and those can be part of the input variables into my generative AI solution. To say, yes, this aspect of the weather was an influential factor in this issue at hand. Is that what you and your team do in terms of combining data sets and then identifying additional variables that could?

Speaker 2: 42:02

that could influence the outcome. Mike, it's like you've been in these calls with me, so yes, so I just didn't matter until this exact point. I'll bring up a prior client experience and body of work. We have worked with a golf equipment manufacturer who, who let's just say that the team we got engaged by was their engineering and product development teams. So at that point they were literally coming to us saying, hey, we have a knowledge management issue. How do we help? Can you help us, like figure out what we can do with knowledge? And that's unstructured data, structured data, documents, reports, test data, right, all this stuff. Absolutely Short answer.

Speaker 2: 42:40

So we started working that process and didn't take as long to realize, well, who would value this most in the organization. And the answer of all places was customer support. Because, at the end of the day, customer support was getting calls every day saying, hey, my club didn't work right. And in that case the obvious answer was all right, we'll send it to this. You know, we'll send you a prepaid envelope, we'll bring it for testing. And that's very expensive, let alone that person's out a club for having it.

Speaker 1: 43:09

Yeah, it's disruptive to the customer too.

Speaker 2: 43:12

Exactly. So I said well, hold on, what if I could tell you what we just built here using your knowledge base of engineering knowledge base? Nothing less is knowing whether or not whatever they're telling you in real time is something the company knows right, and that knowledge application is what came back to the exact user story you just shared. Actually, is that same person on the phone the really likely angry customer saying hey, the club slipped out of my hands last week. Clearly the rubber is malfunctioning. They say All right, sir, just give me a little information about the exact situation you're in.

Speaker 2: 43:54

When did this happen? Last Thursday, oh, I see you're based in San Antonio, texas. Did you play in San Antonio? Oh, last, aren't our gloves? In fact, we never test and in our disclaimer say that our product not tested against any gloves other than ours. So in fact, that can't be something that we're liable for, and that's the way in which this becomes a customized response. But it's purpose built for supporting what their knowledge is applied best to in real time, and that's the idea of what this could do is be dynamic in that sense of being generative for what that next best question is.

Speaker 1: 44:36

And I think that's much more efficient. The customer feels better cared for and, wow, I talked to somebody who's really smart. I should really keep that in mind. Yeah, because it could be, one thing Well, of course you're hitting into the wind. Play a club higher, but well you know you're hitting into the wind. Play a club higher, but well you know you're playing in the barometric pressures lower. You've got to play half a club higher. There could be some knowledge like that that is to be unearthed.

Speaker 2: 45:01

That's fascinating and an opportunity, right, Because at the end of the day, you could say no, it's your fault, not mine as a company, but the company doesn't want to ever say that. But rather, hey, could I give you a 50% discount on our gloves and we can have that shipped out tonight.

Speaker 2: 45:15

right, that's the opportunity to build the relationship, and we do that in a way that comes back to what is the core offers we have to play with, let alone what we're willing to provide. And again, this all could be generated in a simple user experience that gives that and arms that same customer service agent with the augmenting power that they're actually looking for.

Speaker 1: 45:35

That's fascinating, and thanks so much, eric, for not only joining me on this show, but just drawing analogies to other industries that are that. That just really makes it fascinating and tells me at least just the applicability of data and creative solutions to meet customers where they need to be met.

Speaker 1: 45:54

So yeah, thank you, eric, my pleasure you know a couple of questions as we round out here. You know we talked about a little bit throughout the program how we have engineering backgrounds, and I saw that you attended Carnegie Mellon and you were into robotics. Saw that you attended Carnegie Mellon and you were into robotics, and you mentioned earlier aviation time at Ford. Just going back to your time in college, tell me a little bit about robotics. Have you always been engineering, science, math science type of a person, and how did it ultimately lead to automotive and data?

Speaker 2: 46:31

Oh, this is great. Yes, so I had an incredible four years in Pittsburgh at Carnegie Mellon. It opened my eyes to way more than what I thought engineering was, but what it could be. And that's where I started to see. This is in 2004,.

Speaker 2: 46:47

Well, the early innings of self-driving cars. Right, it was happening as a form of research with DARPA, and I got onto this research team as an undergraduate looking at how we can use their well, literally, h1, hummer platform to operate autonomously in the desert, and that's where I started to see this is no longer just a mechanical challenge. The real world is much more complicated than that. Right, and as embedded systems were starting to become more and more invested and understood, that sort of opened my eyes to what were all these called robotics. Robotics was building and embedding intelligence into mechanical systems to start, and it was through that that I. My first application of that was in aerospace manufacturing at Sikorsky Aircraft, and I got, let's say, assigned first to the actually another DARPA R&D program. That was about embedding electromechanical technologies into the helicopter and Blackhawk's blades so that it can morph in real time as it turned around its azimuth, and that opened up a whole host of other pilot interfaces which it was the dawn of the ios app store. Actually, that got, let's say, some of our developers be like maybe why don't we do that for a pilot? They could choose. And what if you could? Because you couldn't before this, I want to fly quietly now. Well, that means it's, you know, the payload has to decrease, so in fact, you can't fly with nearly as much payload. So that that's what the cost of doing so is, and it's, it's, I don't know, you can't fly nearly as far, hypothetically in range. So, but you got to figure out what is the trade study in real time and embed that intelligence into the helicopter flight. That's what we started to unlock and explore is like how these electromechanical technologies start building new ways to use these machines right, and that through an MBA.

Speaker 2: 48:37

Even further up my eyes back to automotive, saying, hey, wait a minute, those are 30-year product cycles. I'd think a little bit more quickly than 30 years and want to. And so that's where I found wait, I could do it 67% quicker. So all of a sudden down to a 10-year cycle, and that was what, let's say, got me excited about. Moving to automotive, where I saw that the dawn of autonomy was starting to hit the roads and that's where I started starting it from there after my MBA into Tier 1 called Harman. Harman got acquired by Samsung, brought me from New York to the Bay Area and started my career. That led me into Ford's Palo Alto team. That's kind of where my journey with them started.

Speaker 1: 49:21

That's fascinating, and you know, eric, this has been such a fun conversation and I'll just have one last one here, kind of a random, personal kind of question. Let's say you're having a party, you're having some friends over, it's going to be a pizza party, and tell me the type of pizza toppings you might put on, either that are your favorites or that you know some of your friends would like. Um, would love to hear some of those.

Speaker 2: 49:43

Assuming you mean pizza, Definitely. Um, it's probably very telling, but uh, no, we are a pepperoni pizza family, and specifically with the, let's say, some blend of onion and garlic powder on it. Right, so we like the spices, but with the pepperoni slices.

Speaker 1: 50:06

Okay, great. Yeah, I went to college in the Boston area and I grew up in Ohio, so I was used to kind of Domino's Pizza Hut pepperoni, mushroom, green pepper, onion sausage like the Supreme right. And then I go to college and I remember going to this place called Bertucci's or maybe it was the newspaper staff where they would bring pizza on Sundays and there were sliced tomatoes, there was eggplant, there was broccoli and I was just this is weird. It was just like a kind of a culture shock to me. But I've really come to appreciate all the different varieties of pizza and eggplant pizza, for instance just fantastic.

Speaker 2: 50:45

So well, I'm going to say one more quick thing is, Michael, whenever you're passing through the Connecticut area, I'd love to host you for a white clam pizza, which is something as you go back to unique things about geographies. During Connecticut, some people call the birthplace of the modern pizza. There's some incredible, unique dialects of pizza consuming experiences that I would love to share.

Speaker 1: 51:10

I'll definitely take you up on that. So, eric, once again, thanks so much for joining us. I hope you had as much fun as I did and, to all of our listeners, we hope you found this one engaging, enjoyable and informative. Thanks for tuning in to another episode of Auto Care On Air. Make sure to subscribe to our podcast so that you never miss an episode. Don't forget to leave us a rating and review. It helps others discover our show. Auto Care On Air is proud to be a production of the Auto Care Association, dedicated to advancing the auto care industry and supporting professionals like you. To learn more about the association and its initiatives, visit AutoCareorg.