Ep 27: Using AI To Transform Manufacturing Processes with The Heico Companies EVP & CIO Tom Gerdes

Evan Reiser: Hi there, and welcome to Enterprise Software Innovators, a show where top technology executives share how they innovate at scale. In each episode, enterprise leaders share how they’re driving digital transformation and what they’ve learned along the way. I’m Evan Reiser, the CEO and Founder of Abnormal Security.

Saam Motamedi: And I’m Saam Motamedi, a General Partner at Greylock Partners.

Evan Reiser: Today on the show, we’re bringing you a conversation with Tom Gerdes, CIO at The Heico Companies. Employing over 9,000 people globally, and with over $2.5 billion in revenue, The HEICO Companies is the parent organization for over 70 high-performing manufacturing, construction, and industrial businesses. In this conversation, Tom discusses the advantages of Generative AI at enterprise scale, the realistic impacts of AI on modern manufacturing processes, and ethical considerations surrounding AI adoption. Tom, do you mind sharing a little bit about your background and maybe what your role is today?

Tom Gerdes: Sure, currently I'm Executive Vice President and CIO at The Heico Companies. I've been with The Heico Companies for 16 months now, the first CIO they've had to build out a center led IT organization. A little bit of background on Heico, 40 years of family owned business, we have over 9,000 employees, over 70 operating companies in over 19 countries and close to 140 locations. So it's a bit of an interesting environment to come in that hasn't had the concept of IT at scale for an organization of about 3.5 billion of revenue. So we needed to help transform the organization and build out different capabilities to really push and disrupt a bit of what we've been doing in the past. Prior to that, I held a number of roles at Johnson Controls in the IT organization, and then before that, I did management consulting for about 15 years at E&Y, Capgemini, and Deloitte Consulting.

Evan Reiser: And Tom, do you mind sharing what inspired you to join? What was the kind of opportunity? What got you excited?

Tom Gerdes: Yeah, I joined to really go through a large transformation of an organization that, in many ways, didn't even have the foundations established that you'd expect for an organization with that size and scale. And so the opportunity to come in and, from the ground up, build the IT operating model, the capabilities, not just to do the “keep the lights on” activities, but how do we start to deliver value and then partner with the business to really generate opportunities for increased business value.

Saam Motamedi: Tom, I think one of the things that's so interesting, both about Johnson Controls and now your role at Heico, is you're working now at Heico across a portfolio of companies that, to like our average listener, may not appear as like technology first or technology needed. So maybe I'll just start with an open ended question, which is, could you share a couple of examples around how Heico's harnessing technology across the portfolio?

Tom Gerdes: Yeah, a number of areas we're looking at and utilizing technology. We'll start at the core on manufacturing so getting more and more connected to the shop floor. So a lot of this is focused initially on how we look at and understand performance and getting into pretty basic stuff around predictive models for maintenance of equipment, ensuring that we have maximum uptime and utilization. Going forward, looking at how we get into continuous monitoring over manufacturing processes focused on product quality and efficiency. And you think about a model where you build in quality assurance to every part that you make, every process in the manufacturing, not requiring people to be there, but utilizing automation tools. And then layering in on top of that, how do we utilize that data set, that information, to make sure we can always find ways to optimize our process and our manufacturing capabilities. And there are a couple of specific examples we're looking at. Some of them are based on utilizing digital cameras, so real-time monitoring of our manufacturing processes, looking for quality defects, and then turning those into pre-indicators. And then that will ultimately move into autonomous systems. So we're automatically making adjustments to the manufacturing process on a real-time basis. So there's a lot that goes into that, the data acquisition, the data schema, then getting into what we're doing to train against that data set and that model, and then getting back into having edge devices that can make those decisions real-time without having to communicate back to the master control system.

Evan Reiser: Maybe just a quick follow-up there, Tom, would you mind sharing other anecdotes of ways you're deploying technology in innovative ways that maybe the average listener or even the average customer of yours may not fully appreciate?

Tom Gerdes: One example where we continue to pursue the utilization of technology is actually going the opposite direction from manufacturing with connected products we have out in the marketplace. So some of our products have telematics built into them and so taking that information and being far more aware of the status of the equipment, how it's being operated, and ensuring that you're getting the maximum value for your end consumer out of that product. So while you think about things like cranes and forklifts being non-modernized technology, even when they're human operated, the real ability to get better data and better insights and then use those models to start getting, at the end day what you've seen in a lot of airline industries, which is uptime and availability of equipment and really putting that across our portfolio of products. And with a lot of the markets we serve, they haven't really been touched by technology to that deep of a level and so it becomes a bit of a transformation to take both our business through but, as well, our customer base around the additional value that you can get out of those services and solutions.

Evan Reiser: So it's hard to have any conversation about IT or technology or Cyber Security without talking about AI so I'm just going to jump right into it. Do you mind sharing any examples where you feel like you've been really successful deploying AI or Machine Learning technologies?

Tom Gerdes: Yeah, I think we're heavily looking at and continuing to make improvements in how we're utilizing those tools. I think what's going to be really interesting, not realized yet, but as semantic search starts coming to play, or semantic indexing, pardon me, that's going to be really impressive. So as we see that service built on Microsoft, and we're a Microsoft shop for a lot of our services, I think that's going to be an amazing release of value for us to look at all that data out of Graph API: how we can utilize that to understand behavior and the activity of our users, and how we can utilize that data for better security, how we can use it as better insights for data loss prevention and insider threat. But also, I think it's going to be enlightening for how we think about data privacy, data security, and broader information governance topics. I think that's going to be a real area for us. So we continue to look at just the ability for classifiable trainers that exist today. But I think as we get the ability to layer in those indexes, I think that's going to unlock a whole different set of values that we'll see. And then ultimately driving that to get the real value, which is automation, so I don't need a SOC operations team to be reviewing every incident and alert, but we can actually have the AI model make those decisions for us. I think an area that's going to be really interesting for forward development is how do you really look at trying to get endpoints that become far more self-aware? So right now we've got a model that I've got to rely on logs coming out of my endpoints, analyze those through a scene, make an alert indication, perhaps create an incident, have someone adjudicate that, and then maybe I quarantine a device. And as I can get that decision-making closer and closer to the endpoint and make the endpoint the initial determination of what that risk profile looks like, I think those are gonna be some interesting innovations that I anticipate we'll see here quite quickly.

Evan Reiser: That sounds like you're reading off my investor pitch deck, I was pretty aligned to that. But maybe going back to your business and time, I know there's cool stuff, I'm trying to pull it out. In the past, I think when we’ve chatted, you've talked about using computer vision and manufacturing. These are the concrete examples where you've seen some enhancement or transformation of a more conventional business process that you feel like you've yielded great results through some of these new technologies.

Tom Gerdes: Yeah, and we're driving that in one of our manufacturing businesses, particularly around drawing wire and so you take a rod, that rod gets drawn out to wire, then you galvanize the wire, and then it gets fooled into different configurations based on customer, consumer need. So as we go through that process of galvanizing the wire, that's where that digital vision starts to give the opportunity to look at and indicate if you've got a process failure on a real time basis. And so that's a real model where something that is very old school manufacturing, applying that technology on a real time basis gives us the ability to alert to quality issues, remediate those quality issues, and at the end of the day, it drives better yield, better product quality, and all of that delivers better value to our customers. That's just one of many examples when you start looking at how we're gathering data that exists on the shop floor and utilizing that to make sure we're driving improvements to our product quality. I think another area that we look at, and I think is going to be an innovation, we talk a lot about ChatGPT around the tech side of things, but as we start looking at robotics or broadly movement, the opportunity to look at how we can utilize models to look at where there's environmental health and safety issues, particularly around the safety of our employees and people that are in our manufacturing facilities or on the construction sites that we support. So looking at how we can predict where there are going to be potential errors or movement problems for our workforce. The same way of being able to scan images and say, there's something here that a machine might be able to identify as a safety issue with more predictability than we can see with humans, because we often stop processing, that autonomous response is to not look at anything. There's a classic example of read two sentences and tell me how many “of”s are in it, and six is generally the answer, but a lot of people read three or four, sometimes five, because we're automatically processing a lot of information. These AI models don't, hopefully, get to that level and so they're actually analyzing everything in detail and making better predictions than we can as humans. I think that's where there are some safeguards in looking forward, particularly in evolution around environmental health and safety.

Evan Reiser: Tom, those are really great examples because it's like a win-win-win. Like you mentioned, there's an efficacy improvement where there's some superhuman capabilities you can get. Obviously it's efficient as well, right? Because sometimes you can't hire a hundred people to go inspect all these things where technology can do that in much more efficient ways. But also I have to imagine there's some, just like, safety benefits, right? You can't have a thousand people supervising every piece in a manufacturing facility, it's just cool to see some of those examples that are a win-win-win.

Tom Gerdes: Yeah, and it's no different than what you will get in information security of putting defense in depth. Well, it's a similar concept of safety in depth. You need to have a safety culture. You need to train your employees on proper technique. But this gives us another means by which we can put in some automated controls, and if it's an integration into an AV system that makes an announcement or you have a flashing light that goes off, other indicators that, again, don't rely upon that human factor in order to make that safety notification. And the other piece is that, we see this a lot, when you do have safety incidents, sometimes it's because people don't take the action at the right time. And looking at how these models and tools can actually take that action without the, well, if I shut this down, we don't make our production quotas. These are all kinds of pressures that are on people that stop the human element from making the right choices when it comes to safety events. And these are means by which you can safeguard against that and maybe even give the employees comfort that they're going to be reinforced by making the right decisions.

Saam Motamedi: Tom, those were terrific examples. I want to zoom up and talk a little bit more about just AI strategy and this intelligence transformation that's happening in the business, and I have a few different angles on it that I'd love to get your perspective on. The first is you gave a bunch of great examples, which I would put in the bucket of predictive uses of Machine Learning. Obviously, the whole world's been taken by Generative AI over the last 12 months as well, thinking back to the launch of ChatGPT and the fact that these models cannot just be used to make predictions, but to actually generate that new data across modalities. And I'm curious, what, if anything, you all are doing on the generative side, either today or things that you might be exploring over the coming quarters and years?

Tom Gerdes: Yeah, and that's where I see a big, big transformation for us as an organization. So we're looking at a number of areas. So content creation is a big one, the ability to start looking at how we build out product catalogs and how we think about creating descriptors of our products. Utilizing Generative AI in that space, I think is a key area that we see value in. We also see that as a great opportunity to start doing language translation. We operate in 19 different countries so being able to go through a process and do forward and reverse translation into foreign languages and utilizing Generative AI to accelerate that process of work, it’s hugely impactful. We also think about real opportunities to solve some of the problems around, I'll say a hodgepodge of non-conforming data schemas. So with 70 Operating Entities, enumerable ERP systems, different CRM platforms, while we're going to move towards unifying on a core set of tools, the opportunity to look at how I can use these tools trained on a vast set of data to gain insights that would otherwise be very, very challenging for us as an organization to go through. And the opportunity to look at how you leap over the work effort to create an enterprise data schema and use these tools to gain insights that might be a bit more challenging for us as an organization. I think also looking at a fundamentally different scale to automation, and I think you talked about this in previous podcasts around how much further you'll be able to take core automation and, hopefully, automation that doesn't require me to go to the same level of process standardization across the board that you can automate for the variations that exist today, as opposed to the old model of standardize everything and then I can look at automation as a means to get something that's predictive and repeatable, which is maybe a bit more of the RPA model. Hopefully this moves beyond this and gives us the opportunity to start seeing some of these automation gains without having to do everything as a bespoke one-off automation routine.

Evan Reiser: Tom, the one example about the kind of interoperability of data, that actually makes a lot of sense to me. I haven't really heard anyone else mention that concept, but it makes sense when you have got all these different entities, you have different data systems, whether it's accounting or customer records, and traditionally, if you want to unify all that it requires someone to go map out the data schema, centralize it, build a new thing for all the piping and flows throughout the process on the inputs and the outputs to make sure it comes in the right ways and deals with all the gaps. If you think about AI as like a really smart translator of stuff that can dynamically figure stuff out, you're right, I think more of that can be done in the future with AI in ways that would require humans to do the planning and execution today.

Tom Gerdes: Yeah, exactly right, and I'm working through this right now as we look at, say, it's 40 different ERP data sets we deal with across all of our locations, trying to get a viewpoint of how we understand our core customer base. Who are we serving? Who are those customers? And when you're working across 70 operating entities, understanding that becomes a challenge and it becomes a massive mapping exercise, you're looking at trying to correlate data together. I’m hopeful that this is a way in which we can create a different way to solve those kinds of issues. It's going to be really interesting to see how that evolves here over the next few months.

Saam Motamedi: Yeah, absolutely. I think one of the related things I want to talk about is, in my mind, I think of it as augmentation versus automation, but just more generally speaking, how do you think about the adoption path of AI in the organization and how your end employees interact with it? What does that look like in the short term? What does that look like in the long term?

Tom Gerdes: Yeah, it's a really interesting problem to work through in our organization. So part of it, I think, is talking about what I at least see as the journey of where we utilize search and tools and now moving towards AI that used to be around doing education and research. Then it moved into how to troubleshoot something or find an example for a solve. Now it's moving into the utilization of these tools to actually take action. Same way of it used to be reporting, and then it's analytics. Well, now we can get insights, and those insights drive action. So instead of the financial statement outcome, I actually have a recommendation for what I should do in order to affect something to change training, working capital, or to impact days payables outstanding. So I think looking at ways in which you can translate the user's context, what you're trying to achieve as a business outcome, and how these tools will get them there is a way to start to bridge that gap. I think one of the other challenges is also building up trust in whatever AI you're deploying within an organization. And so how do you gain the confidence within your employees that something that they may not fully understand is an outcome that they can still have trust and confidence in? And I think it's not new to this ecosystem that we're dealing with around Generative AI in and of itself or AI for that matter. We've experienced this throughout the sort of arc of technological change throughout the world, and throughout history, the pace of change is increasing and we can't really understand all of it, but we have to create a system to build trust and to be candid. I don't have a solution for that within our organization yet, but I think that's going to be a piece that I think there's going to be a lot of conversation around in the ethical use of Generative AI. Now we're seeing voluntary marking of AI generated content through a lot of providers, I think that'll start to maybe build a context where people far smarter than I are gonna figure out the means of establishing trust in these services.

Evan Reiser: What just clicked for me is the importance of that, especially in your operations, your environment, when physical safety is an issue. You have that flashing green light or the flashing red light, people need to have trust that's like a trustworthy signal. Otherwise, it can work against some of the potential gains of some of these technologies.

Tom Gerdes: Yeah, and we've assumed that trust in a lot of areas because we built on maybe a slower pace of technological change, right? We trust automated machining centers because we took time to build them out. We've measured the parts. We understand that they're conforming. So as you move towards technologies that are using digital image recognition and other more advanced capabilities to identify good part, bad part, and you're doing that in a less than a second to make that determination, building the confidence with our employee base and with our customers that the parts they receive are parts that they can trust. And I think that becomes an interesting component that, as these technologies potentially allow us to go faster and faster and make more autonomous decisions, how you establish that chain of trust throughout that decision making process becomes a really interesting problem to be solved.

Evan Reiser: Tom, one thing I appreciate about you is your ability to simultaneously acknowledge the reality and the practical constraints, but also dream big about what that future can be. Do you want to share a little more about, maybe vision is the wrong word, but as you kind of imagine the future, what are some of the ways you think new technologies, whether it's AI or something else, have the ability to have an impact on a business? And what do you believe will be true about the future impact of AI that maybe most of your peers would disagree with?

Tom Gerdes: I don't know if I have any terribly controversial views there. I think when we look at manufacturing at the core, it still takes people to make products. And I don't see that that gets drawn out maybe as much as some think. We can certainly look at more and more automation around how we drive and how we develop more efficient processes and put in more automation tools but I don't see maybe a transformational shift in taking a lot of the labor out of that process, I think it's about skilling that labor up and utilizing automation as a means to get more value out of what we do in a manufacturing organization. But when you're running a furnace to produce steel billets, and when you're rolling those into rod and we're drawing it into wire, there's certainly a level of automation that can work there, but they're intense processes, they're hard on the environments that we work in, and they're hard on equipment that processes that. So I don't know if that's gonna be really the big fundamental evolution for us, it's maybe better insights into how we control that. I do see though, the core components when we think about back office business processes, those are where I do think we're gonna see a tremendous ability to apply this technology in a much faster time horizon. To be clear, I'm not saying that there's not a world where there's far more automation in our production environments, I just think the spend versus value is going to yield a much better return for the business by focusing on back office processes. The context that Microsoft's put out there with Copilot and utilizing Generative AI tools to enhance the productivity of our workforce, I think that's where we're gonna see the big transformation here very, very quickly. And then I think there's a bit of a longer tail, that I see because of the cost involved, to try to drive automation and insights because much more of it's, what you think about traditional automation of utilizing third parties to come in and produce automation equipment, I need that investment to create a bunch of insights in order to drive better automation.

Evan Reiser: So at the end of our episodes, I like to do a quick, more of a punchier lightning round and we’re looking for your one tweet version. Saam, do you want to kick off the lightning round first?

Saam Motamedi: Yeah, absolutely. Maybe to start, Tom, how do you think companies should measure the success of a CIO?

Tom Gerdes: I'll give you three items. One, participation in the strategic vision for the organization, thinking as a business leader. Number two, creation of value. And then number three, management of risk. Not necessarily reduction of risk, but management of risk. I think CEOs these days need to be risk accepting, not driving risk to zero, but finding the right areas to make bets or be disruptive.

Evan Reiser: I just feel like it's one of my pet peeves when someone says, well, I have an abundance of caution within this. I'm like, I don't want an abundance of caution, I want the appropriate amount of caution.

Tom Gerdes: Exactly right, yeah. And if you're trying to drive risk to zero, you're, in my view, driving value down to negative, or certainly below zero. I think it's very, very hard, and I think the whole component of generating value is about understanding and accepting risk, making sure that that risk is also effectively communicated to the business.

Evan Reiser: Tom, what is a piece of advice maybe you wish someone gave you when you first kicked off your journey as a CIO? And maybe what's that piece of advice you might want to share with other peers?

Tom Gerdes: I think the first one would be don't rely on what you've done in the past. I think you need to be willing to break what has made you successful, because repeating the patterns of the past with the changes and shifts we see in technology are not going to be the patterns of success in the future. I think making sure you're always conscious about challenging the status quo, and particularly the status quo of how you've been successful.

Saam Motamedi: How do you think CIOs should position themselves to best collaborate with the rest of the C-suite and leadership team?

Tom Gerdes: You have to understand the business. And by that, I mean be intimately involved in the knowledge and operation of the business, go out and see the sites, go meet with customers. I come from a bit more of a non-traditional background, I did management consulting for 15 years and was in a number of roles at Johnson Controls, none of them would be viewed as coming through a core technical pedigree. What I do and how I approach my role is that I'm a business leader first, and then I run IT as a function to deliver value in the organization and to help realize our strategic goals. And so I think that's really a core component of how to work and collaborate with the business is to be part of it, to be intimately involved in understanding how we operate, who our customers are, and how we deliver value to the market.

Evan Reiser: I'm going to switch up a little bit, maybe go more on the personal side. Tom, do you mind sharing if there's a recent book you've read that's had a big impact on you, and if so, I'd love to hear what it is and why?

Tom Gerdes: Yeah, the recent one that I go back to is Essentialism by Greg McKeown. I think it's just a fantastic book about how we take noise out of our environment and the distraction of focusing on too many things and really getting down to the core that drives value. Fantastic book and I have to give a shout out that was recommended to me by Ben Johnson from Obsidian Security. When he knew I was taking a new role, he was like, hey, this is a great read because you're going to be inundated with more stuff than you can deal with, especially with this degree of transformation and how do you get down to focusing on the things that really matter so you can make shifts in scale that matter as opposed to spending all your time dealing with 20 to 30 things that are really distracting from making progress?

Saam Motamedi: Maybe the next question, also staying on the personal side, is what's an upcoming new technology, again doesn't need to be related to your role at Heico, that you're personally most excited about?

Tom Gerdes: I don't have the risk profile yet, but I'd like it to be to a point where I can, space travel would be pretty cool. So something more practical and pragmatic that I think would be great to see, actually it's the evolution of technology to solve broader issues in society, I think is what I'm really excited about. I've got two young boys, watching them grow and engage in or act with technology is fantastic, I just hope we can safeguard the world for them years down the road.

Evan Reiser: So maybe one final question, which I think attaches onto that. What do you believe will be true about technology's future impact on the world that maybe most people consider science fiction today?

Tom Gerdes: Maybe this is controversial, I do actually think that we could run into an issue of containment risk. So what's going to be interesting to see if it's that question and maybe the science fiction positing that if AI becomes truly self-aware, is there a problem that it can't solve? And the old question is if you put a supercomputer in a room and all it had was its processors and a power supply, would it be able to get out and break containment? So it'll be interesting to see as tools progress and where and how that issue moves forward. Yeah, I think that's an interesting one. I debate that with myself on occasion, whether that's a meaningful issue or not. But the way it's been described is through signals it could produce through electrical wire, could it somehow manage to program something? And it's an interesting problem to think through what the implications would be if you have truly self-aware, non-contained AI, and what's the basis on which it would act and behave.

Evan Reiser: There's some historical precedent for software breaking containments, right? Like Stuxnet and things like that. It's a scary world that we're going into, but I think it’s going to be filled with more good, positive things to the world than negative.

Tom Gerdes: Yeah, I definitely agree. I am very, very excited about the opportunities that technology evolution and advancement are going to have for us as individuals and I hope for society as a whole, and that we do see the positive side. I'm hopeful, but I hope it's not false hope, that it's going to provide much more good than it will potential for harm.

Evan Reiser: Well, Tom, that seems like a wonderful note to end our episode on. So I appreciate you taking the time to speak with us today and thank you for all the hard work from you and your team to help keep the world running.

Tom Gerdes: All right, thank you, Evan. Thank you, Saam.

Saam Motamedi: Thanks, Tom.

Evan Reiser: That was Tom Gerdes, CIO at The Heico Companies.

Saam Motamedi: Thanks for listening to the Enterprise Software Innovators podcast. I’m Saam Motamedi, a General Partner at Greylock Partners.

Evan Reiser: And I’m Evan Reiser, the CEO and Founder of Abnormal Security. Please be sure to subscribe so you never miss an episode. You can find more great lessons from technology leaders and other enterprise software experts at enterprisesoftware.blog.

Saam Motamedi: This show is produced by Luke Reiser and Josh Meer. See you next time.