Expeditors International of Washington, Inc. (EXPD) Earnings Call Transcript & Summary

Hello, everyone. Thank you again for joining our Emerging Technologies webinar. My name is Gina Suriano. And before we get started, we are just going to run through some quick ground rules. [Operator Instructions]. Today's webinar will be recorded, and you will receive an e-mail [indiscernible] link to the materials and the survey. We definitely value your feedback, so we look forward to the completion of the survey. Next, we're just going to introduce our speakers. So as I mentioned, my name is Gina Suriano. I am the retention and development manager, and I'm located here in Pittsburgh. Speaking today will be Longjie Dai. He is the Customer Solutions Manager for the Northeast region, and he is located in Boston; and also Shane Wood, who is the Senior Manager of Global Business Development, Supply Chain Solutions and he is located in Denver. So with that being said, I'm going to turn it over to Longjie and Shane.

Thank you so much, Gina, and thank you, everybody, for joining us. Hopefully, you had a chance to take our poll here and some pretty cool results. It seems like some of you guys are already utilizing AI, either at work or at home. About 23% of you guys have actually already used some of the new genre of the AI platforms. And in terms of emerging technologies that you guys have implemented within your organization, it seems like they're users of AI and IoT devices. And we're going to be talking a little bit about all of these emerging technologies, but really focusing in on generative AIs and Digital Twins today. So what's new in AI? Well, before we can talk about generative AI, we really want to take a moment and talk about traditional AIs and what differentiates the newer generation of generative AIs that you may have heard about in the news or implemented or tried out recently, and try to define some of those differences. So traditionally, guys, you might have been familiar either at-home use or in the workplace over the last decade or so. They've really kind of come up the backs of both pattern recognition as well as information retrieval. Most of these more traditional AIs have very defined and specific tasks, and they usually revolve around either interpretation of data or simple calculations of data. Some examples might be facial recognition technology, that's becoming more and more prevalent, things that your bank does to make sure that when you have transactions that don't look normal to your normal spending patterns that that's automatically detected, things like voice assistance. If you have anything like a smart speaker at home or even recommendation engines. The key differentiation between these traditional AIs is that it's, generally speaking, not very creative in types of contents that create, that traditional AIs have been able to create. So a couple of examples that you guys may be familiar with that you may have physically used may involve things like doing flight tracking and price analysis; let's say, you were going on a vacation and you wanted to know what the cost is or when you should book a particular flight. Traditional AIs, when you use things like our Google Flights or any web-based flight search engine, it is really good at retrieving that information and giving you both historical and current patterns for how that would look like. We have a couple of digital examples of what that would look like if you're unfamiliar with using that, so here's a few. Say you're doing that flight search. Not only are you going to get a historical graph with more traditional AIs, will even give you an idea of when you should actually pull the trigger on booking that. And that's all well and fine, but what can the latest generation or generative AIs do perhaps differently or better compared to our traditional AIs? Well, generative AIs work a little bit differently. They really build on top of the body of work that traditional AIs have done, but one of the secret sauces behind it is that it can iteratively really teach itself based off of the data set that it's been provided to learn about very broad topics, so it's no longer limited to a very specific niche data set and the implementations that we have seen. And another key differentiation that we have seen in some of the newer generative AIs is that they can both take their input, and in some cases, their output in natural language processing. So you don't have to format this data in any special way. You can type it out exactly as you were texting a family member or a colleague in plain English, and it can output, generative AI can output things in images, it can help with things potentially in video, or back in text or other visual mediums that are very easy to interpret. And if we peel back the onion a little bit further specific to the text-based generative AI tools, it's essentially a whole recursive function that is always trying to find the next best word to put in there with some variation. So for example, if I prompted generative AI, this is -- the output here is based off of the previous version of OpenAI's GPT model, GPT-3.0. You can see kind of the logic behind it where given the proceeding words and are front about the best thing about AI is its ability to it looks into its training model set and tries to figure out what would fit best next into this string with some amount of variation. It doesn't always pick the highest ranked choice that's out there. It gives itself a little bit of flexibility to sometimes choose the second best ranked or third best ranked. And then it reiterates. Once it's chosen a word, it will then choose the next word that makes sense based off of all the previous words. And continues to iterate until it's for not only sentences, but paragraphs, and we will show you later, very detailed analysis for it. Okay. So let's take another step forward and see what generative AI can do. So I pump the generative AI, just like that Google Flights example into a much more open prompt. I ask that, hey, which date should I plan a vacation to Seville? It's going to -- instead of just giving me a listing of things that perhaps other people have written out, it's going to give me some very specific answers to those prompts. Not only that, I can have it create itineraries for me as well. And just as quickly as I can generate an itinerary, I can tell it that I would like an addition or maybe a change in something that it suggested. I can have it specifically either avoid or include a particular city. A lot of things that you may have to go out and find either a local expert or do a lot of time researching through traditional methods, generative AI is going to be able to do a lot of that heavy work for you, particularly at the beginning of any type of idea or a project, or in this case, vacation planning. Now this is a great example for personal use, but what about applications outside of things like planning a vacation? Well, we've run through a quick demo of something particular supply chain application or ChatGPT. Now this is run, utilizing the latest version that's publicly available. And I just prompted ChatGPT to do what we at Expeditors would consider a greenfield analysis towards center of gravity. We give ChatGPT a couple of prompts about demand for product and then tell it to ideally place the distribution center. So ChatGPT will run through and even explain how it's going to accomplish this task. It's going to say, hey, we need to take a look at the volumes, the geo coordinates and then plot that out, we're going to run that through and balance it based on the demand and we're going to tell you where a distribution center might ideally be located. Now as I said before, it's very easy to make permutations on this. I can give them another prompt and say, hey, what if I had 10,000 more units in Miami, Florida in addition to the demand we already have? And it very quickly takes that new information, incorporates it into the model and spits out an answer of where it thinks that distribution center should be both initially and where it could change based on additional information that comes to it. Now I will note, and you guys will be able to view this yourselves when you slow this down and view this recording. ChatGPT actually makes a critical mistake in its calculation. And actually, based of on the demand data, moved my center of gravity further north. I pointed out this mistake to ChatGPT after reviewing the results, and it very quickly recognizes that it made a mistake and makes the corrections for it. And then it very quickly then permutates again when I add in more demand information, shifting that center of gravity and ideal distribution location further south and then further west when I add more demands to it. So again, please, I highly encourage those of you who have not given these tools a try. Try them yourself, try either this example that I provided with centers of gravity and see what these new platforms are capable of. I'm sure you will be surprised, especially if you thought that this was not only a tool for home use. It's already a very capable tool when paired with some amount of expertise. So you can kind of catch it when it makes these types of simple mistakes. Now this is a basic example, but what about even more difficult examples? There was a great study done by a UPenn Business School professor, and he ran his final exam through ChatGPT to see how well it would do. Now this exam had various different topics that would be useful in business analysis. They took a look at process analysis, inventory management, queuing theory, and even things on production system knowledge. And ChatGPT, albeit with a few hiccups, generally did very well, although there were some caveats as a result. So the professor was able to run his final exam without too much effort right through ChatGPT, and he determined that it would have most likely passed if it was the student in his class. We have some very key caveats in what it was still lagging. Now keep in mind, this is a very nascent system. In fact, as we were preparing for this webinar, OpenAI actually announced and released for some journalists data access to its fourth iteration of GPT. So some of these caveats are going to already have been fixed within the iteration that's being tested right now. But the professor noticed it was able to answer the basic questions fairly well. You would have given it a B- grade. Although it make, just like the example I showed you earlier, some pretty surprising mistakes, it understood these very advanced concepts. In my case, it made a very simple mistake by taking a negative and then taking the negative again of it of a GPS coordinate which skewed the whole analysis. And in the MBA exam, it did similar things in basic math mistakes. And we see this again and again when different people test it. Now the very surprising thing is, instead of utilizing the platform to just answer questions, they found that one of the most useful things for the technology would actually be to help it create new test questions given the understanding of the model and what it will spend in as a training algorithm. It was actually very confident in creating new problems as well. So again, this is a very quickly changing subject matter, and we will likely see the vast improvements come through. Okay. But what does this mean in the business sense, particularly within supply chain? Well, the additional challenges of any AI, generative AI is that it's only as good as the data that it's trained on, and the data validation itself can be incredibly labor-intensive. And actually, surprisingly, at least in the current state, require manual training. We're not quite at that point, at least in the beginning of the data sets that use large platforms can completely train them on tasks to start with. Once you get started, it can create an adversarial model and really task fits up. But at the beginning, especially in defining what data sets it's trained on, still requires a good amount of manual scrubbing and data to get through. In the business sense that means you'll have to compare that cost and time line to utilize the AI creative solution, for repetitive task might be a no-brainer, take that time to train that AI and that system to solve that problem, especially if it's something you're seeing again and again. But maybe for niche problems that only come up once might not be worthwhile at least in its current state to that training and data cleanup. And speaking of data clean up, most important thing to utilize this specific to your business needs is understanding that the data you're feeding it in your organizations and master data needs to be clean. If there are dramatic data irregularities, maybe either quantities of, by order of magnitude or dates being up, I can entirely throw off the analysis that the AI is able to provide you. Now while this is my analysis, don't just only take it from me. Bloom + Gartner took a look at this -- they too had their reservations. They have already had generative artificial intelligence within their Hype Cycle. They really feel that this is going to be accelerating fairly quickly. And at least within non-supply chain applications, they're going to see mainstream adaptation in 2 to 5 years. It's actually incredibly quick. That said, within supply chain applications, due to the need to have clean master data and clean understanding of the models that actually are behind us to properly utilize this probably about 10 years out into Gartner. Now that doesn't mean we should sit back and wait 10 years. This is exactly the time of action to start now and start understanding your data, start building the models around your supply chain organization so that you're ready to take advantage of these exciting new technologies when they're a little bit more capable. And to talk a little bit more about how you can prepare your landscape, I'll turn it over to Shane.

Thanks, Longjie. Interesting stuff. I love the video and kind of seeing it in action. I think you told me that the speed, if anybody's played with it, it's actually very fast. I think for the sake of time, Longjie did it at 2x speed, but it's amazing that the ability for it to query and respond back with just the intense amount of detail. So I think as Longjie pointed out, a lot of this requires good data and the foundation to be able to use it, you need to have those pieces in place, whether we're talking about doing some of the work that I'll get into in terms of supply chain optimization or to be able to leverage this type of capabilities, obviously, you need to have some of the data and various portions of the data. So I highlight this particular e-book, a colleague and myself had sort of in the middle of the pandemic when we were all in lockdown and the logistics crisis were starting to begin, we wrote an e-book that we really aim to try and to help customers. It's not a sales-oriented material, but we really looked at what are the foundational elements of building out your supply chain data. So I'd encourage anybody that is interested to download -- you snap the QR code and download it. We have some worksheets in there and we sort of talk about building out the foundations to do more advanced things. Two years ago, I probably wouldn't have envisioned the capabilities in terms of AI and where we are and just how quickly it's getting built out. But there is a -- I think we talk about lots of data -- and my overall feeling is that the foundations are bringing in more data. So finding your data deserts, where you don't have data and focusing on fewer items and being able to get really good at using those particular items. And as you use those items and you can't answer specific questions or understand certain things that you're going to find where there's additional gaps and you need to expand your focus to more areas and using different technologies is certainly going to help you go about building those out. So one of the areas that I'm really keenly interested on and certainly the last couple of years in the supply chain space, has prompted of the ability to build what-if capabilities. And I think AI actually will have a large play in this of being able to sort of surface and talk about some things that you should be aware of in the supply chain area to incorporate, to understanding the what-if capabilities. And so I have here just a simple image from our friends, the Wright Brothers who were the first in flight. It's sort of in quotations, the "first in flight." But if you actually look at the history, they were not the first in flight. I think they were the first in flight that we're able to maintain 5 different characteristics, right, that they were -- they could pilot it that they could -- it was lighter or heavier than air. So it wasn't a hot air balloon, that it had an engine that it was able to do certain things that previous attempts hadn't. So they were first to be able to master these 5 criteria. And if you look at the history and what they were able to accomplish, really, they were brilliant engineer minds. And so they really mastered modern engineering to be able to test ideas and do what-ifs. And so a lot of the inventions that they came up with were actually -- they're still used today in aviation. And so they were -- if you know the history, they were bicycle manufacturers. They started investing all their profits from building these bicycles and selling them to them being able to get into flight and build their aircraft and test different ideas. And so over the course of a couple of years, they modernized or were able to identify the needs to be able to have a bi-wing plane, and they identified the wing working to be able to capture the lift and be able to balance across the 2 different sides. They had propellers that spun opposite directions, because if they spun in the same direction, then of course, it would turn the aircraft. They identified an elevator and a rudder to help them be able to steer the airplane and be able to capture more lift as they needed it. And so I bring this up because I think that it really ties into the progress they made was really developed because of their ability to test what-ifs. They built an actual wind tunnel where they were able to test these different types of propeller and wing configurations with great detail for that time. And so all these different things that they developed was really a scenario of what-if tests that they then applied and used and were able to live about it and then go down in the history books that we all know today. Just last week, I read about this great article from McKinsey, and they talked about scenario testing, which is what-ifs, scenario testing mostly from a geopolitical standpoint, but I thought that the article was really interesting. And they talked about 3 different characteristics. So Black Swans, which we all know, which are these very impactful events that we can't predict; and then you have Grey Rhinos, which are very impactful events that we probably can predict. And then they talked about silver linings. And so these are the opportunities in the middle of crisis or opportunities that you can expect when one of these events happen. So a Grey Rhino. And they really highlighted that in an organization, it's good to be able to understand these events and think through what would that mean to our business? What would be some of the deficiencies, where would be some of the opportunities? And I think certainly, the logistics challenges and the challenges with COVID highlighted many of the companies that were better prepared, were able to capture market share, take advantage of that volatility and use it for their benefit. So maybe they have advanced awareness of some of the lockdowns and some of the logistics challenges based off of information they were getting. And so they were able to ramp up suppliers and get key components sooner than other -- than their other competitors they're maybe using those particular products. That's 1 example. I think one of the examples that they talk about it in this white paper from McKinsey or the article was renewable energy companies. With the Ukraine crisis, the Ukraine war and the energy challenges is that it's really been a windfall for a lot of the renewable energy companies. And so if you've been able to predict that and been able to sort of understand what was going on in the ground to know what the connection to that would be sort of the next layer of effects, well, then there's an opportunity for you to capture market share and take advantage of it. So I'm going to transition to talking about the what-ifs to Digital Twins, which is a huge component of the capabilities. And just maybe before I get into it, I'll highlight about our team. So we are a supply chain design and consulting group. We offer independent consulting services so we can support customers not only about the business that we're supporting within the Expeditors realm, but also the other pieces of business. And we have a Digital Twin platform that we are working with customers today that I'm going to go through. And really, we're doing probably 90% of the activities in the platform is non-Expeditors business. So we can have an NDA, reinfense engagement to support customers looking at their overall supply chain which may incorporate part of Expeditors but portions outside of it. And we're really in the supply chain design space and our Digital Twin is oriented to mostly supply chain and logistics areas. So the first Digital Twin, if anybody knows Apollo 13. So Houston, we have a problem, right? But not that one. Houston, we have a problem. So that was the really the first original Digital Twin as most sort of researchers in the space to identify it. And essentially, a Digital Twin is a virtual representation of actual processes or things that are happening in the real world. And so the communications technology, the information and computing power was produced well enough and the technology was good enough to be able to reproduce the oxygen tank issues for Apollo 13, they were only take the sensors, take that information, extract it and then reproduce it in one of their simulators. And so they have the benefit of being able to see what happened if we shut off up both engines and we just try to use gravitational pull to get down, oh, no, we're going to be going too fast. -- we need less electricity. Let's see if we shut down the electricity. They could do -- they did all these different configurations with the engineers and pilots to be able to test ideas to see what works versus what doesn't. And ultimately, that is the power of the Digital Twin, not only give sort of the monitoring capabilities to understand what's going on in terms of your supply chain, but also be able to understand and test those what-if ideas across different dimensions. And this is, of course, this was over 40 years ago, -- and since it's been widely utilized. If you look at the Gartner Hype matrix that it's been on that matrix for a number of years. It's going around the Hype matrix -- you can certainly see it in other industries and there are specific programs and solutions that are developed specifically to launch a Digital Twin. And so I've highlighted just a couple of other examples. So you can see the Hong Kong airport. If you ever fly into that airport, you are part of their Digital Twins. So they're keeping track of all the different dimensions of the airport operations. So not only the aircraft coming in, but the security lines, the cuts on blind, where folks are in different portions of the aircraft so that they can make a holistic decision to see where should we bring the next aircraft in? Should we bring it on this side of the terminal or that side of the terminal, how may that impact all of these different things, where do we need resources, how are we using the electronics, the electricity and the systems and ultimately trying to continue to manage and be able to optimize their different operations. Formula 1 uses it. You can think about all these different race tracks, they have to engineer it specific to the race track and the conditions. And so they can understand with the sensors and the activities, what things may work best for their vehicle on those conditions. And then engines and predictive maintenance. So engine manufacturers, most of them today have a revenue model that is based off of uptime. So they get paid as the engines are flying aircrafts. If the engines are down, well, they don't get paid. So it's to their benefit to get out in front of any potential challenges to be able to do predictive maintenance to say this part is going to fail in another 5,000 miles, let's go ahead and replace it so that we don't have any downtime. So it's certainly well utilized across multiple industries. In the supply chain space, and specifically in our offering, so we take all of the transactions and data for the different portions of the supply chain, whether that's a brown truck, a purple truck, a yellow truck, an orange vessel, the activities in your warehouses, that data can then be reproduced, engineered into a database and can be reproduced to be able to capture what is the workings of the supply chain, the processes, the spend, the carbon footprint and the different activities. The biggest characteristic is that it is connected. So as those activities happen, it can be connected, and we think about it more, not of real-time connections. So this isn't an execution system but a historical system. So the activities for last week or last month from these different data sets to connect them so that they continue to build up over time. And so you can take those same ideas of the simulator and be able to apply those to your different supply chain ideas. And as I always say, there's always more ideas and more questions than companies have resources to. And so this gives us the ability to be able to understand, to test those ideas much quicker and more efficiently, given some of the capabilities that we've developed with the Digital Twin technology. So kind of getting into the actual sort of workings of it. So as we think about data, right, you can have different sets of data. So you could get just your internal system. So your ERP data that gives where you're buying from, where you're selling to, maybe it's your WMS, or TMS or those internal systems. Those can be ingested into the database. And we typically think about ERP as a great source of data because as we optimize the supply chain and we test ideas, we not only want to look at just freight that maybe your company pays for, either delivering or buying from suppliers but also the freight that maybe customers pick up from your warehouses and suppliers land into your facilities. So that ultimately, you're optimizing the entire network and not just the portion that you pay for. Because ultimately, if you optimize the entire network, you can make up for that in product price, in the negotiations with your suppliers or your customers. Another sort of rising area are visibility integrators. So this is your P44s, your E2Opens, obviously, have gotten a lot of increase in usage over the last couple of years as people are trying to track down where is my stuff. And so we use that data, it's a great data source that gives different milestones, a single view into a lot of the visibility activities. Their problem set is typically around where is my stuff, and we use that to, given one portion of the data of the overall portion of the questions that you may ask. And then the logistics service providers. So these are all your different service providers that give information of what are the transactions that they've handled, the spend, the modes, the location, the suppliers, all those different items. And in terms of the integration, I want to highlight that this can be -- we try to lean towards simple. So simple in terms of using scheduled report, dfb excel format that can be scheduled on that given cadence. In terms of the systems, we can use -- FTP. So we're not trying to do these massive integrations, and we're really trying to help connect to the data at its source or develop some simplicity with the different data sets. So those are typically the transactional sets that we're looking at and working with customers on -- and then we have the ability to also bring in market information. So this is market information that maybe -- that you have in terms of where you're measuring yourself against others in a particular industry or we're starting to build the capabilities to use market intelligence mostly around rates. So what are the average index values for LTLs in North America from the Northeast to the Southwest. And so if we bring in all these transactions and then we have this market information, we can ultimately use that information and be able to do lots of cool things with it. But first, we have to make sense of the data. So everybody probably has a lot of data challenges. And so you work with lots of different spreadsheets from your different service providers, from your different system. You know the challenges of it being inconsistent, you have different naming convention service levels, you use 15 different names across the 15 different service providers. You're not able to match the different event codes. Maybe you don't have the connection to the order level that you would like. And so maybe you do some of that additive work using just some Excel formulas, doing some manual cleanup. And so we try to take that same approach, but use computing power and ETL processes, so algorithm-type processes, to be able to build logic and business rules and cleansing using code. So we take those 15 different service providers, and we can say, all right, let's get down to 3. So now you can look at everything across all of your supply chain is some -- is a single window into where your spend is and where your activities are. And so a lot of the work we do and just the foundational level of the data is giving customers a clean and consistent view of their data across these different data sets. And if we can build these processes, use these rules and we get this data on a weekly or monthly basis, for example, Well, now we can start to take care of the exception. So as we do more data refreshes, we're going to have fewer and fewer exceptions because they're going to be caught by the code and adjusted -- and really, customers get the best of both worlds where you not only don't have to deal with the manual pieces, but you also get the benefit of us being able to surface to you the exceptions that you can go and correct at the master data level. So you're not messing with it on a manual step, but you're also correcting the Longjie's earlier point, the master data area so that you're actually correcting the root cause. So we have this Digital Twin now of this data that's connected, clean, consistent, a single window into your supply chain. And then once you have that, you have the foundations where you can use -- you can do different things and use technology to support you in different portions of your supply chain. So I'll just talk through some of these. So adapt board. So interactive adapt boards, that's the visual part of our platform, being able to see a 30,000-foot view across all the portions of your supply chain, what is going on in my business and then being able to go down to a business unit into a node, into a service provider, into a particular flow. Where are there opportunities? Where are things not going according to plan? Beyond that, you also now have the foundations for your carbon footprint because you have one single view for the different activities. So you can now, as a native part of the activities actually start to calculate carbon footprint with a single methodology. What we've really seen is that there's a lot of customers that are taking the different service providers, the methodologies, which may, in fact, be different and they're trying to construct it and do a little bit of a jigsaw puzzle to get to a single carbon footprint. And now we can do it with one methodology. We don't have to worry about the different calculation steps that are being done. We can calculate as one methodology to give a comprehensive, baseline carbon footprint that ultimately sets the foundation to do other things and connect sustainability with business initiatives. You can also start to model out costs. So very much like a freight cost company would do you can model out the expected cost to be able to surface those exceptions and do that much quicker. So you can do that on a weekly basis after the shipments delivered to either be able to understand what happened. And if it's a one-off approval that I need to understand with the business, or there are some exceptions that you want to get out in front of as opposed to it all going through the freight cost and payment center, which may be 60 or 90 days, and then you're addressing things way after the fact. RFQs are a hot topic these days. So now we have a single data set -- we're working with customers to be able to do that type of RFQ work and help them do simulations and optimizations for the different carriers that are coming in. So it supports their procurement strategy to look at, of course, different cost levels for the different dimensions; air freight, fuel, handling, delivery, all these different areas that go into the landed cost you can simulate but also you can apply cool rules to be able to fit your procurement strategy. You want to be dual sourced, you don't want to just go with the lowest cost. Well, now you can put constraints on it and do different types of work to help you really support your procurement strategy with actual really detailed math. And then we can do where is really the secret sauce developing a baseline with that data and doing simulation. So anybody that's done a supply chain design study will probably recognize that the hardest part of doing that study is actually collecting the data, getting the tariffs, applying those tariffs to the transactions to give a [ caustic ] baseline model. That in our experience has been about 60% to 70% of any study over the last 17 years that we've been doing it, and it resounds with my sort of experience. I've been doing this work for 8 years, and that is definitely the case. So there goes a lot of work into building assumptions, cleansing the data, organizing the data. So as we bring customers into the platform in their different data sets, we can apply their tariffs or market tariffs that are representative of a baseline that sort of is a proxy for their supply chain. And ultimately, you can jump in to test ideas much quicker. So what may have taken months now can take days or weeks. So ultimately, you're able to test a lot more ideas because you're jumping right into those ideas. And so you really have the benefit of being able to do more over time with less. To bring that point home, so we would really think about it in terms of a journey. So we would never say, hey, let's give all your logistics, all portions of your supply chain, let's try to bring it into the tool, let's do a transition over 2 years. We really think about it. Let's start with -- think about this as a journey starts small. So we're going to like capture a segment of your supply chain, but you get the benefit of a lot of the things that I had already talked about with just that data. But then as you expand, you can get more portions of the supply chain into the lens and more portions that you're starting to understand. And then once you have all of those nodes really mapped out and understood, well then you can do things like scenario testing for resiliency. So customers are able to test ideas of, well, what happens if our DC shuts down? What happens if our suppliers in Thailand have another flood? What happens if we introduce these things. And ultimately, you can test these ideas to be able to see, well, if that happens, what's the best way for us to be able to manage that? How much inventory will we have given we're not going to get any supply? How would we allocate that inventory to our different customer levels? And so you ultimately get to the spot where you typically would have a whole host of ideas that you're looking at. First, around cost. So customers are usually focused around cost. And then they get into some of the more proactive management, risk management, resiliency oriented items. So this is an example of maybe a project pipeline. Of course, you got to have at the top of the list, the CEO's pet project. And then after you have that, you can get down the business and identify lots of other things. So I'm going to echo what Longjie ended about the fine 10-year age whiskey. I'm going to put it another way and sort of think about this in terms of the future is coming, and really the foundations are necessary. We really hope that we can help customers in building those foundations. We don't see it as a challenge with data, but we see it more around being able to support customers and helping them get their data organized and expanding over time so that in 20 years, you have a beautiful tree that can give you lots of shade and you can enjoy it and read a book and talk about all the days of the past when ChatGPT was simply just doing centers at gravity. If anybody is interested and want to hear more about the things we're doing and look at the platform, certainly reach out to your account manager, your sales executive and me directly, I'd be happy to talk with you about it a little bit further.

Thanks, Shane. If you can just go to the next slide, I'm going to show everyone informative resources and upcoming events. So before we get to the questions, we just wanted to share some of these resources for you. You can register for these communications through the QR codes or the subscriber here links when you receive the actual presentation. We definitely hope that this session was informative, and we look forward to your feedback once you do receive the survey. And just from a question perspective, we do have one question. I am [indiscernible] but I fear that my data may not be ready to utilize this. Is this a technology where I should focus on my data foundations first? And how long does it take to develop and implement?

Okay. Yes, well, so you definitely need data, but I actually wouldn't worry about the data not being perfect. We sort of think that with our solution, right, you're not buying just a software. So you could buy Digital Twin software. I think Microsoft has got a great version plus probably into the ERP, looks at different things, -- but that's much different because you do need somebody to run it, you need resources, your data needs to be very good. Our thoughts are a little bit different in that we actually think that our product fit is geared towards a lot of customers that have data challenges, and we can help them on that journey. So I flip it the other way of like if you're not making progress on your data today, what does your road map look like? And this is a little bit of the easy button where you can start small with certain portions of your supply chain. We can help you organize it, identify those gaps, cleanse it and then we think about expanding into more portions of the supply chain. And I say it's the easy button, which I think this ties into the second part of your question, Gina, of we're right now onboarding customers in about 90 days. And we certainly wouldn't advocate to take every portion of their supply chain. We're going to start with the big hitters, the first priorities and go through that. configure the dashboard, work with them on sort of setting up the business and the activities and showing value and then expanding into more pieces of business as part of that implementation process. So I hope that helps. Certainly that presents a question, they can reach out and we can talk for more specifics.

Shane, I just wanted to add, since we have that survey response early on, there are ever-increasing sources of data nowadays, right? So they saw the survey we had at the beginning, it seems like a lot of people within their organizations have worked with or exploring IoT. So those are all data points that we may not have had years ago to incorporate that, and to your point about not letting -- if I can summarize it, it's not letting perfectly the enemy of good. You take the data that you can and put it together because that's going to be instantly better than not making that attempt with the data that you have. And as we get more and more devices, data points, to really just going to make this solution even better, but we should certainly start as soon as possible.

Yes, yes, and I've seen where as you start investing time into sort of building this, you're going to realize where you really true data challenges are and what gaps you have if you're not using it in your business as much as you'd like, if you're only looking at certain points and not looking at the holistic view, you have a pretty -- you have a more narrow lens -- and ultimately, you're not finding those gaps. So you're trying to get to the questions you want to answer because you don't have the data points to be able to answer those. So this is all a journey, right? I mean, I think ICAI is going to help with a lot of these data challenges, right, and sort of the capabilities to bring another data set, maybe do a lot of the organization and cleansing kind of in the same way that we're doing today for a lot of organizations, and this is going to be embedded into a lot more platforms and capabilities. But also the challenges are growing even more because of all those new data points, right? And sort of the frontier of competitive edge is really going to be leveraging that data to drive your business, to understand patterns and behaviors. And use it, whether that's from a logistics standpoint or from a forecasting standpoint and from a risk management standpoint to be out ahead of it and use it as part of your business.

Well, thank you, everyone. Thank you, Shane and Longjie, for speaking. Thank you, everyone, for your time today. Keep an eye on for the e-mail with the survey and then the presentation. And if you have any questions, please reach out to your local representative. Have a great day.

Yes, thanks for giving us an hour of your time today. Have a great day.