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

We will go ahead and get started. There are a few people that are still joining us, but I just wanted to say hello and thank you to everyone for joining Expeditors Webinar today on utilizing generative AI in the supply chain. My name is Rhonda Macy, and I am the District Sales Manager, both the Charleston and Savannah districts and I actually sit in Charleston, I'll be your host today. But before we get started, there are just a few housekeeping things that I would like to run through very quickly. So you'll know today's webinar is going to be recorded and all attendees, you are anonymous and you're on mute. If you have any questions at any time, please feel free to put those questions in our Q&A box at the bottom of the screen, and they will be answered at the end of the presentation. Again, just -- it will be recorded, so I just want to get your heads up for that. If there are questions that come through during the webinar, we do hold those to the end or we would have to stop the recording, so we just want to keep that flowing. At the end, you will receive an e-mail in the next couple of days at the most containing a link that will give you a registration for the materials printed in it today. So when you fill out the survey that's given there, it will give you the overall presentation that has been presented today. So your feedback is very valuable. Also, your questions on this topic are valuable. We believe it's a timely topic for everyone, and we hope that you will enjoy it. So as we get going, let's talk about the speakers that we have today. We have 2 of our Customer Solutions Managers, both representing the Northeast and the Southeast region. We have Manuela Johnson, and we have Longjie Dai. And both have been with the company for 9 years. I wanted to talk about their role a little bit to introduce it to everyone on the webinar today. Our Customer Solutions Managers, they are technology solutions experts of the region. They work with our account managers and our sales teams to map out the customer supply chain process, they listen to their issues and identify opportunities to streamline and automate processes through the implementation of technology solutions. The solutions, for example, just for some of our customer-facing solutions, our online platform, whether it be our customer reporting and any reporting that's needed, systems integration, trade management, software and so on. So their heavy focus is on process improvement. So thus, leading into our topic of AI. Let's look at the agenda as we go through. Few things that they're going to cover are digital transformation, talking about artificial intelligence, generative AI, unlocking supply chain data, time lines and challenges and then the mainstream use of generative AI. So I'm going to turn it over to Manuela Johnson as we get started, and I thank you so much for joining.

Perfect. Thank you, Rhonda so much for kicking us off today. So we're going to dive right in. As Rhonda had mentioned, we do think this is a very timely topic. Actually, most companies that we talk to say this is a topic of discussion within their organization today and they're already moderately investing in that area. And that sentiment also aligns with some independent research that was recently published in Forbes. They interviewed over 1,000 supply chain professionals across different industries and across the board, the respondents were already moderately investing in digital transformation initiatives in things like inventory management, transportation, warehouse equipment and also facility management. And we do really expect to see much more interest, more investments and just more movement in the digital transformation space in logistics. So when we're talking about digital transformation, I really don't think of this as an end result or kind of like a goal for your company to get to, but rather I think it's going to remain this ongoing evolving field. So I think it's very important to look at its foundation to then really understand how to effectively use the tools within that space today. So there's 2 concepts that I want to talk about that are related to digital transformation. You have digitization and digitalization. Digitization really is the foundation of it all. And it's the process of translating analog information and data into a digital form. So a simple example of that is just scanning in a document and storing it on your computer. The foundation for that was laid in the 1940s from microchips and semiconductor shifted analog computing to digital computing. Really any digital transformation process will always have to start with this foundation, right? Because any innovative technology in that space will, at some point, rely on quality data as an input to then be able to produce a quality output. And you're going to continue to see that quality data input resulting into an effective output theme repeated throughout today's webinar. Now the second concept on here is digitalization, and it's the use of digital technologies to change a business process, usually within a limited scope, like on a project level. So an example of that would be training your employees to use a new software platform that's designed to automate a process. And some examples of those type of technologies are the utilization of computer-aided design and manufacturing that was introduced in the late 1970s. Then you can see in the 1980s, ERP systems came about. That stands for enterprise resource planning. And then in the 1990s, we saw customer relationship management systems being introduced. And of course, very important, the World Wide Web became publicly accessible and certainly an integral part of everyone's business towards the end of that decade. Then in the 2000s, social media was introduced, right? And that really brought about a revolutionary change in how people communicate and exchange information in general. By 2005, so just a couple of years later, Internet users reached 1 billion worldwide. And then once the iPhone was introduced, that really launched that mobile revolution in 2007. In general, we can say that the Internet certainly shifted us from a siloed world into a much more global one, right? Now businesses are digitally connected to their customers, not just via website, but also other digital channels, again, thanks to the birth of social media. And then in 2011 is when the term digital transformation was coined by a consulting firm Capgemini, in partnership with MIT. And while that coin digital transformation can include digitalization efforts, it certainly goes beyond that limited project level scope and affects the entire organization. So now we're really transforming the status quo and we're ideally developing new capabilities in order to increase organization's competitive advantage overall. And these are some of the popular tools in the digital transformation toolbox based on what we hear from customers and then also what we see in the industry in general. So AI, artificial intelligence and ML, machine learning, those are topics that we'll dive in today, and Longjie will go over those in more detail in just a couple of minutes. And we also have IoT, Internet of Things, which fully describes a network of connected devices that exchange data over the Internet. An example of that in the logistics space are GPS sensors that you can attach to your freight. A lot of our customers are now also asking us about API that stands for application program interface. And that's an integration method that's much easier. It's less costly, less time consuming compared to its traditional counterpart EDI, electronic data interchange. And then Digital Twin, which allows an organization to create a model of its supply chain in a virtual environment. And the goal of that really is to simulate different scenarios, so different changes to that supply chain and subsequently analyzing the effects of those changes all within that virtual environment. Now the idea behind digital transformation as it relates to supply chain is that it enables an enhanced version of your current supply chain. Traditionally, you're probably used to seeing and then your supply chain, right? And information is flowing from left to right, from upstream to downstream, you have that one directional flow of information, which, in turn, is limiting the overall visibility throughout your supply chain. There's probably a lack of real-time updates. And in general, a linear supply chain has a harder time adapting and it's just less responsive to changes in the marketplace. Now as supply chain enabled by digital transformation is much more circular, right? It's interconnected. Communication is now flowing back and forth between different supply chain partners. And we're utilizing software applications like ERPs, but also maybe transportation management systems, TMSs that are all becoming more and more intelligent to the point where they can alert users on exceptions and maybe even handle some of the simple exceptions automatically. So this enhanced version of your current supply chain really uses intelligent software that can help monitor, it can help manage, highlight exceptions across different functions within your organization, but also across companies to optimize your entire network. And then as a result, of course, your supply chain is much more efficient, it's more responsive and also cost-effective. And while we would, of course, love to talk about all the different tools within the digital transformation space today, as promised, we are going to focus primarily on artificial intelligence and generative AI. So with that, I will pass it on to Longjie to talk about that.

Thanks, Manuela. And I think it's important for us to understand where AI fits into the grand scheme of our digital transformation across each of our individual companies as well as the industry as a whole. Artificial intelligence, if we broadly define it, it's really an algorithmic base fashion to essentially simulate human intelligence. Another way to think about it is it's that whole interdisciplinary field of applying computer science, social sciences and really tactical and operational level uses within business of automation and a lot of these tools that we've built up on top of. The first time most of us have dealt with AIs have been what we would now refer to as more traditional AIs. They are going to be much more narrow and focus in what they're doing, and they really revolve around interpreting data that's already out there for use in specific [ task ]. So traditional AIs, we're really talking about things like voice assistance, OCR engines that are turning, maybe handwriting to text, things like price comparisons, either for home use or business use, where you're going out, you're scouring the information out there and trying to find a result that is very narrowly focused. Now these have gotten better over the years, but they haven't really been transformative in the same ways that we've seen and specifically in the last year with generative AIs. Well, what's so different about generative AIs? What differentiates them from more traditional AIs? Well, generative AIs have been the extension of machine learning that allows more expanded [ end results ] compared to traditional AIs. So instead of being able to just find a cheap flight or to be able to turn on your lights by your voice, you kind of have an open-ended endpoint where you can create text images, and this is all accomplished through the advances that we've seen over the last several decades. Now when a lot of these generative AIs were taking off a year ago, we were kind of limited not only by the fact that they were tech space that they were mainly in English as well. That has certainly changed in the last few months. There are great models out there that can work in multiple languages that can work with images, both input and output. We're going to go through some of those examples. Let's start with probably a basic example that you guys may have, if you played around with generative AIs, just see. This is something we generally hear very often within the supply chain world, right? If someone is, let's say, newly starting or just wants to be updated on how the current port situations are, this is a very common refrain. Let's say you're an importer and you're trying to bring things over to the East Coast. A common choice we have in that is if you're bringing something out of Asia into the East Coast U.S., are you going to land that in a East Coast port or a West Coast port. Now the difference for that is going to depend on several things. We know as partitioners, there's considerations of transit time of cost, of specific things that are going on in the moment, such as port congestion. Now we're dealing with issues of water shortages in the Panama Canal. So all of those things come through and are synthesized for either a choice or a combination of choices. How can generative AI help us in that? You might take this search result and put it into a traditional search engine. And you're probably not going to get a full-fledged view of this picture. What generative AIs really good at doing is synthesizing a lot of these results and putting it together. So this query I made asking about the differences for East and West Coast, asking it for cost and lead time, I know the text is a little small, but the generative AI, this one that I've shown example of came from Bing Chat, we created this just last week, did I think, a pretty good job of laying out the benefits and [ downsize ]. Now I will say, there are always caveats. So we always try to push this to the end and say, we tried to query, well, tell me exactly what these costs -- you would think these costs might estimated to be. And that's where it started to get a little bit tripped up. Inherently, the generative AI knew that it was more cost-effective to go to the West Coast. But then when it tried to calculate out the inland portion and combine that with lead times were starting to trip up a little bit in the full analysis. So again, great starting tool for this kind of query gives us way more information than we get out of a general search query that just doesn't fully flesh out. Now we would have ended on this, except for last week, we actually got some interesting news as it relates to competition within the generative AI market. Google finally a little bit late to the game came out with its engine, I rather see an improved engine behind its front end. So Google Bard was recently updated last week. And we tried querying the same question to it. And interestingly enough, Google's engine, actually, I think, did a much better job at providing the details as it related to the same exact question that we have queried just to week prior. It was able to really summarize for us the differences in transit time, some of the other considerations to make. And I will say these additional factors I think are very interesting. So instead of just listing out what was requested of it in terms of cost and transit time, you can see that in the tables, it also prompted as part of the result, additional factors that we haven't even created. So I mentioned you might want to consider things like inventory management. What does this mean for actually how long you're holding on to inventory for as well as sustainability. If you're bringing things over to the West Coast and trucking across the country, what does that mean in greenhouse gas emissions, right, versus bringing all the way over to the East Coast. So really interesting things and evolutions that we're seeing as part of these generative AI responses, and again, they're getting better kind of as we're utilizing these tools, a lot of this presentation that we did today, we had to do a second rap last week just because of how much a change in the industry. Okay. Another example, those of you who may have attended our previous webinar, might remember this one. I query ChatGPT, which is based off of what people refer to as kind of Version 3.5 of ChatGPT's engine to do a center of gravity. Those of you who have worked with industrial engineers or operations research specialists may have done this before. It's a great tool to be able to at least start a search for where you might place a warehouse or a distribution center. All that typically takes is some kind of demand data as well as geographical distances. And it tries to find a center of it. Now what I think has improved about this is all of these generative AIs are getting better at showing kind of some of the inner workings. In fact, one of the advantages of Microsoft's engine is that they actually cite almost all of their responses. So they'll actually have citations where they're pulling that data from. ChatGPT does it a little bit differently. It shows you the math behind it and tries to do a calculation. So all I did in this calculation was I actually had it in a previous query, pulled the population for children under the age of 5 based on the census data, out of each of the states within the Northeast and the Southeast, the 2 regions that we have kind of represented today here on the presenter side. And I tried to have it do a center of gravity. So think Northeast states and Southeast states where the center of that would be. So ChatGPT very confidently calculates out the latitudes, longitudes, [ weights ] them and calculate a center of gravity approximately around Trenton, New Jersey. So if you know anything about where distribution centers are placed in kind of the general RFPs, that might make sense if we were talking strictly about the Northeast. But unfortunately, if you include locations such as Florida, Georgia, Puerto Rico, those numbers don't quite make sense. I tried to add in a little bit, give another option and say, hey, what if you broken out and did 2 centers of gravity? It just responded, the 2 centers of gravity would actually be the same as one center of gravity. So we can see that ChatGPT still has a little bit of way to go with this question probably requires a little bit more prompt engineering to get the right answer. And again, we thought this was it, but, of course, last week, we had the new release. So we punched the same exact query into Google Bard. And surprisingly, it was able to get the right answer. So it calculated based off with the original data set at the center of gravity between these 2 regions and the demand points that we put was in Newport News, Virginia. We ran this on this side. We validated the information through our own more traditional tools and did validate that, that is the correct answer. Not only that, when I gave it the additional prompt on, hey, what does this look like if we ran this with 2 centers of gravity, it was also able to get the [indiscernible]. So in a 2 kind of DC model, incorporating demand from the Northeast and the Southeast, you kind of have a 2 DC model, somewhere between kind of Pennsylvania and New York-ish area and then you'd have a separate distribution center kind of mid-Florida, somewhere in that Orlando area. So we're very cool again to see generative AIs in supply chain applications, really do more than we were expecting and what we had seen even in the previous months. I'm going to turn this back over to Manuela to talk about how now that we know some of the capabilities, we can actually apply them in our day to day.

Great. Thank you so much. Yes. So as Longjie had mentioned, today, you can already use artificial intelligence to unlock the full power of your supply chain data and essentially move from data to insights much, much faster. So traditionally, if you're asked as a supply chain professional within your organization to potentially present an overview of this year's supply chain activities to your upper management team, there's a couple of things or, I guess, problems that you have to figure out. First of all, you have to get your hands on the correct data. And hopefully, your organization already has some standard reports built into your ERP system that you can simply run to retrieve that information that you need or maybe you're reaching out to your service provider that, again, hopefully has the capabilities to kind of put all that data together for you in a seamless fashion. And then in other cases, maybe you do have to reach out to an analyst on your IS team to write a completely new query to get that data that you're looking for out of a database that it's stored in. Once you do have that data in hand though, you still need some sort of visualization tools. So maybe you're using Excel, maybe you're using PowerPoint or Power BI. But ultimately, you're going to visualize that data. And again, depending on your skill set, that can also be quite time-consuming. So right now, Microsoft is actually in the process of building AI-powered copilots into many of their products, including Power BI, and that is the screenshot examples that you're seeing here. So these are screenshots of the copilot preview in Power BI and that was actually just released last month in November. The long-term idea behind copilot is that it's going to help users get more done and create more value from their data, of course, than last time. A user will be able to create an entire report, so not just one visual but an entire report page, as you can see on that left screenshot there, simply by prompting copilot with the description of the visuals and also the insights that you're looking for, all in very simple conversational language. So I thought that's pretty exciting. And then copilot can actually go beyond that and they can also analyze your data by generating new or also editing existing DAX calculations within your Power BI and for users that may be new to Power BI that DAX is just the language that Power BI uses to calculate certain things like averages, et cetera. And then on the right-hand side, that screenshot is an example of copilot summarizing all the data and insights on a report page through a very short narrative. So it also has the capability of looking at all of the visuals and kind of summarizing for you what insights a consumer may be retrieving from that report page? So all really exciting stuff, in my opinion. Now today, I'm not able to show you the copilot preview live, simply because I don't have access to it. However, instead, I am going to show you the Q&A virtualization in Power BI, which has very similar functionality. So really, I'm thinking of the Q&A virtualization of just a lower tech version of copilot. So I went ahead and loaded this Power BI with global supply chain data. Now again, in an AI-powered world, you would probably be using some sort of AI tool to extract that data for you automatically. So that tool would be accessing your organization's databases again to pull the correct data needed to generate whatever visuals, summary report you're trying to generate for your upper management. All right. So in this example, again, we have the global supply chain data loaded into the Power BI dashboard and then I'm adding a Q&A virtualization. It is quickly based on the data set that is loaded, throwing some to [ distance ] out of topics for our report page. So I could either click on one of those suggested topics or I can prompt it through natural language to create a visualization that I'm looking for. And here, I'm entering a very simple example question that we get asked very often from our customers. What are the number of shipments per lane in 2023. So you can see it's quickly generating a bar graph, but I can also see because I'm familiar with the data set that it made an incorrect assumption. It actually is using the data element labeled freight to count as the number of shipments. But in this case, the data element freight is actually referring to the freight cost for each shipment. That's okay because we identified that error. So now we actually need to go and train the tool to pull the correct information. All right. So -- here we go. So one way to train the Q&A visual so they can better understand natural language puts and then, of course, in turn, provide the right answer, higher-quality answer is to add synonyms for the different data elements in our data set. So by adding synonyms, we're explicitly telling the Q&A what field people are referring to when using specific words or phrases. Again, shipment was a suggested synonym for freight that is incorrect. So when we're scrolling to the freight data element, we can see shipment listed under the suggested synonyms. Instead, we actually want to add that shipment synonym to the HBL data element or House Bill. And once we add that data element there, we can see that the tool is now learning is already generating new synonyms for the HBL data element based on that new information that we just entered. And it's also removing the shipment synonym from suggested -- from the suggested synonym portion for the freight data element. Of course, adding synonyms for every single data entity in your model can be super time-consuming, even if those are common synonyms for those names. And that's where, again, copilot is going to come in. It does promise a more intelligent and creative source for automatically generating all that work. But little caveat, of course, it's still not going to eliminate the need for an expert or someone that's just knowledgeable and familiar with the data set to again review everything for accuracy. All right. So we have trained the tool. And you can see the visual is now updated with the correct shipment count using the House Bill data element. There's a few other things that we can actually do in the Q&A function. So for instance, if you do not like the original visual chosen, you can also prompt it to show you a different virtualization like a pie chart. In this case, I think that looks very busy. So maybe we'll try a column chart instead. And I'm also not too much of a fan of the long tail on this one. So I think the best idea is to just stick with the original suggestion. So you can always prompt the tool to go back to the bar chart instead. And then once you arrive at a visual that you are happy with that you think best tells the story that you're trying to convey to your upper management, you do also have the option to turn this Q&A virtualization into a standard visual. And then once that is done, you could just repeat the process over and over again until you have your report generated maybe with 4 or 5 different visuals that again summarize your supply chain activities for this year and support the story that you're trying to tell to your upper management.

That was a great example in well showing us. And I think both of our examples, really highlighted some of the caveats of using some of these tools. Generative AI can end up showing results that are quite convincing upon a first glance, but it sometimes doesn't quite exactly yet what you would expect in terms of either the metrics that we saw with your example, it was actually pulling in cost as opposed to run a number of shipments. We saw the center of gravity results and we saw that it was just a little bit off. So it's still, at the moment, requires a fairly skilled practitioner to at least to a [indiscernible]. Now that doesn't mean that in the future, that will always be the case. But at least in the short to medium term, we really see these tools as being a force multiplier, really a tool to be more efficient. We can liken it to having a pocket calculator instead of doing everything by paper or on a slide rule. These are the things that we're seeing in terms of tasks that AIs are good at, but not quite enough there to replace people. Now one of the functionalities that I had mentioned earlier was the ability now for generative AIs to be able to not only work via text, but also via data as well, not only that but it has the ability to truly synthesize data. We wanted to incorporate this from one of our partners out in the Netherlands, and they had come up with what I thought was maybe a little bit of a busy slide in terms of talking about the paradigm shift in the logistics industry. And as we were trying to figure out some way to either paraphrase or synthesize this, we thought, hey, is it actually be a great way for us to utilize generative AI, some of the functionality of it to take something like a slide, a visual slide and synthesize something out of it. So we punch this in. Again, we utilized based off with our previous text, the Google Bard, generative AI. And when we first query it, that kind of gives us the not so deep description of it. We showed at the slide -- the slide shows a new paradigm shift in the logistics industry. It compares traditional activities with our growing digital footprint. Well, if we take a look at the [indiscernible] well, that's true, but all it did was take it the title in the largest elements and just summarizing incentives. But we've done a little deeper. We wanted to give it a little bit more of a chance to say, can you elaborate on that, right? And what it gave us was, I was pleasantly surprised. Now it was not only able to take these bullet points that basically gives us these things out in incentives form. It wasn't like it gave us a bunch of disconnected thoughts and just link them together. I think it really did a good job at providing synthesis. Not only did it provide synthesis, I think this analysis could stand on its own. So the summary I gave us with the logistics industry is moving towards more digital and connected future. The shift is being driven by emerging technologies such as artificial intelligence, machine learning, the Internet of Things, big data analytics. These technologies are helping companies to improve efficiency, reduce costs and provide better customer service. So first, so good, right? For example, mobile devices and smart scheduling are helping logistics companies to improve the efficiency of their operations, hyperconnected instant transport and logistics is providing real-time visibility into the movement of goods, which can help reduce delays and improve customer service, vision that can advance consumer analytics, are helping logistics companies to better understand and anticipate customer needs. Finally, the logistics industry is still in the early stages of its digital transformation but is rapidly changing industry. Companies that are able to embrace digital technologies will be well positioned to succeed in the future. All of that based off of a slide, which I think did an okay job of showing the information that was on here, but maybe not the kind of synthesis that we're really looking for. Now I could argue that we had some very small mistakes on where it plays to certain things. But to go from this into a synthesized description, I think is really powerful, and it shows the direction that the generative side of generative AIs be able to do. And all I did was give this a picture, right? So you're seeing a lot of the built-up technologies, the ability to reduce simple OCR, the ability to organize data and then to organize in such a way that we're taking bullet points into a very sensible summary that you could take -- that you could tell someone that makes sense. I think this is -- leads better than what we've seen in months past. And I can make an argument that it did a better job at summarizing it than maybe I could have done with a similar amount of time at least. So with that in mind, what are -- what do we see being kind of the bread and butter supply chain applications of generative AI that you could use today right now? We have different opinions on this, but in what we see in terms of capability, generative AI already seems to be very capable in areas such as process analysis, right, especially now that we have the ability to feed at things like flow carts and visuals, it can help us determine if we have the right kind of data and the right kind of a framework in [indiscernible] where things may be bottlenecking capacity constraints just based off of throughput data. It's quite good at synthesizing multiple sets of data and giving us analysis that you saw more of it. It's also quite good at doing things such as simple inventory management, allocating inventory to where it needs to be based off of demand or forecasted demand as well as storage capacities. We even saw a hint of that even when we were doing that center of gravity analysis and it prompting us for inventory needs. So really good at doing these very discrete options that you may spend a dramatic amount of time either creating initial versions or updating as [indiscernible] and forecast changes, generative AI can react very nimbly to that and constantly update you on things like inventory levels and process analysis, things that might not be immediately updated within your own ERP systems or your own existing management side. Scenario analysis is, we think, another expansion of something that generative AI is quite good at already, but can be implemented today in a lot of businesses. You could give it your current [ state ] analysis and ask it, hey, what if we tweaked it like this, if we moved this particular manufacturing site here to here, how does that affect my net miles? How does that affect my greenhouse gas emissions. Again, all of these things, these specific examples, keep in mind that the -- that generative AI not only has the ability to provide answers and responses, but you can also flip it on its head. You could give it data, you can give it the model of how your current supply chain works and ask it to give you questions. Tell it to audit how you've been doing things, have it asked you questions. It's not just necessarily a one-way street. We saw examples of this earlier in the year when ChatGPT was query questions off of a final MBA exam. It did all right in some of those answers. But I think what was missed in the headlines was as these professors were trying out these generative AIs in terms of what it could answer. The real key that they found that it could help them save time was not necessarily trying to answer their questions, but to help them create new questions for their exams. So very cool ability to do that to answer and to create questions. And again, the multi-modelness of it, text, images and now we're starting to see video at least be able to be analyzed in terms of generation. We're getting to that point as well. So you can very quickly see advancements even in the modalities of these technologies. Okay. What do we see then in the future as particular as we're dealing specifically with supply chain applications? We talked a little bit about forecasting inventory management, that's only going to get better. But also in the short term, we can see generative AIs being useful for content creation or marketing, things like creating slides and things that look very visually pleasing, but still have the accurate information that [indiscernible] underneath. We already see some lending usage of AIs in customer service. Think about chatbots that you may have already encountered when you're trying to get assistance. Those AIs are going to get better. They're not only going to be and enabled with customer-specific data. So instead of just a really fancy way of providing information that might already be on a website for a particular company. It's going to be able to go in and pull customer-specific data and provide that information on an automated basis. Again, there's going to be probably some challenges to that as we go through. But we generally see this as an improving sector. And again, initially as a tool to help practitioners either in supply chain, customer service, more content creation as a tool to make it more efficient. What about the midterm then, well, we really see generative AIs being instrumental in helping logistics, planning and the optimization, already touched upon some areas where it could be good at that, everything from DC placement to kind of manufacturing design, we see that just getting better, being able to incorporate, again, things like cost, the current state of global supply chains. And as part of that, if we get good at supply chain risk management as well, can be able to dive into your data, look at the outliers and really look to try to reduce their prevalence in future states as well as incorporate kind of some of those outliers into the overall branding. Now the really far out areas where we might still be 5 years plus of some of these functions being aided or kind of assisted by artificial intelligence. Artificial intelligence would be things like product design into element. Generative AI can do many things, but at the moment right now, it's not necessarily going to be designing jet engines for us. We're starting from scratch very complicated product design and duration. It doesn't have those physical tools actually go out and test certain things. Another areas where we feel that generative AIs little bit far off is actually being autonomous in its supply chain management, design and decentralizing part of that activity. Ultimately, we -- even though supply chains may eventually head in that direction, it's part of a resiliency effort we've seen over the long term. Generally speaking, supply chains will at least in the short to mid-term, still be relatively centralized and at least how it's designed and implemented with kind of that on that horizon. So these are areas where we feel there's opportunity, immediate opportunity, things that are in the mid-term and things that are a little bit more further on the horizon. Things to think about as you think about some of its uses would also be the challenges that are involved with generative AI. We're a little bit beyond the initial challenges of generative AI maybe in the beginning of the year, you've heard a lot about generative AI's hallucinating answers, while not still can be the case, especially with cited and documented sourcing in a lot of these newer iterations of general AIs, it's not becoming as much of an issue. Challenging generative AIs with conflicting information can still be an issue. But generally speaking, the algorithms are much better with sourcing where that data is coming from. So data validation still potentially an issue and can be incredibly labor-intensive if you get a result out of generative AI, but you know that something is not quite right, the problem could be in your source data. But to correct that could be a problem as well. Again, that's something of a challenge that you could put a generative AI to help you solve, but it's not going to necessarily at this stage, sort out those data issues itself. If you feed a generative AI, a bunch of junk data that hasn't been validated, chances are you're going to get some junk results out of that. Another issue is that data within the supply chain as you guys all know, is still oftentimes very fragmented. What you have within your company may not be the full picture, your suppliers may have another big chunk of your data, your customers may have a truck. That goes back to the circular supply chain that Manuela presented earlier. And the goal is for us to be able to have that circular supply chain, not just with physical goods, but with information related to data as well. Raw material suppliers ultimately still, to some degree, need to know end consumer behavior so that they can better plan and make sure that in periods of increased demand or decreased periods of demand that they're better able to provide that information. We don't want to be caught up again in the shortages that we saw on the pandemic as it related to everything from microchips to enough capacity on both. Other challenges would be cost. That's kind of the elephant in the room that we haven't really addressed yet. These are great solutions. Many of them are actually free to use. But if you want something where your data is protected, your actual customer data instead of demo example data that we're pulling off of publicly available information, there is a cost component to it. For example, Microsoft copilot, even though it's been made available now has a fairly significant license count and costs associated with it. There's the expression. There's no such thing as a free lunch. I think it applies very much with AI. At the end of the day, you're either helping these AI models become better with training it with examples and providing feedback or you're paying for a solution where it's kind of been customized for you were in your data or information that protected. Finally, explainability and transparency issues, even though generative AI's been better about citing work, some of the nuts and bolts, some not secret sauce, a lot of these companies don't necessarily want to explain exactly what is going on behind the scenes. They will start to show work. We've seen examples of it already today. But if you really wanted to type down and understand how it solves the specific problem depending on the problem, it might not be so easy to peel back the layers of the onion and understand that because that's kind of -- that black box effect is a lot of how these companies really see them protecting their IP. There's also the issue of compliance issues when you're facing decisions and there's a compliance layer involved, how do you go back to your compliance groups, so either internally or externally and explain how you feel about your decisions. You always have to be able to back up your work, especially when it comes to high compliance issues. All right. So let's synthesize a little bit of everything that we went over today. We talked about some of the areas where you can implement generative AIs. And really some of the things that we talked about, the holy grails of supply chain independently managed supply chains, they're quite a bit out. According to Gartner, even though we're going to see more full-blown mainstream utilization within the next 2 to 5 years, a lot of the more complex areas that we addressed earlier within supply chain specifically, are more likely to be 5 to 10 years out. And that could change. We could have more advances, we could have more competition. But as far as we see it progressing, they're quite a bit out. Now even though it that may be true, that doesn't mean we as practitioners in supply chain to sit around and do nothing. Just like a fine wine or a whiskey, we need to understand them now. We don't want to be surprised with a lot of these modules become mainstream. We have to prepare our data, we have to prepare our organizations. So that work really has to [ strengthen out ]. And that's the content that we have prepared for you guys today. I'm going to turn it back to Rhonda to talk a little bit more about other events that we have.

Thank you so much, Longjie and thank you, Manuela as well for your time. I think this is a very informational, it's an exciting topic that you have presented today. And I didn't even think about using the generative AI to look at -- try to create new questions with supply chain. So very neat, very neat. Thank you. We do have a question in the chat box that came up. And I just -- before we get into that, I did want to just mention that we have informative resources that you are welcome to register for. You can use these QR codes using your phone or when you do receive the presentation, there will be a hyperlink that you'll be able to utilize as well to sign up for these. Also, we have upcoming webinars, which you can register through the QR code as well. And we have some nice ones that are on the horizon. So I just wanted to mention a couple that are coming out. 12, 13 which is for the Americas, is the FDA import compliance update. As you see, on the 14th, there is an in-person seminar, that's going to be in Boston. So talking about the manufacturing in Southeast Asia and hubbing in Singapore. And then we also have the Northeast market update for air ocean and domestic, which is upcoming on the 18th of January. But there is one more that I'll drop in the box that's not listed here that I know of, which is on December 14, and it's for the Southeast region, and it's a Panama restrictions and capacity crunches. So I'll make sure that that's listed. Also, I'm going to drop in the chat to everyone before we get into the Q&A, just to how to manage your subscriptions, if you would like to have additional subscriptions, say you have a news flash, but you would like to get some of the others. I'll put that in there for you as well, and you're welcome to utilize that. I also wanted to utilize the chat -- or excuse me, the link for all of Northeast events in case you didn't have that as well as all of Southeast.

So I will address the question next. And please, if you have anything you would like to add, you can go ahead and do that, I'll list these for you. Okay. There's [indiscernible] for that. Okay. Speakers, we had a question that came through and it's -- that I'm interested in how this can be implemented to improve processes and accuracy of the supply chain functions and they would like to know if you could expand on this, please?

Yes. I can take this one, Rhonda. So a great question. I think we touched upon, but we didn't fully answer, let's say, you have an existing supply chain. And you're concerned about, let's say, either data accuracy or you're trying to look at improving your process. How can you do improvement functions with imperfect data? I think we touched upon already some of the functions that you could do. So for example, you could load all your data. And one of the great things about visualizations in general, dashboarding and taking data that's maybe in a tabular form of neutralizing it out is that those outliers will really stick out when you do that. So I would say a great way you can do this is take your data and visualize it first, go through that Power BI an example like well-adjusted and visualize your data. And instead of having it to show maybe a bar chart, have it chosen to box and whisker chart so that you can see where some of your outlying data plans are, right? And then after you've got that done, and hopefully, you've got that done relatively quickly by utilizing either Power BI or some kind of AI-enabled tool, have it then dive into those specific outliers. What happened there and validate that those events actually occurred. I know, generally speaking, from our industry experience, a lot of times, once we visualize the data and analyze it, it turns out that we have maybe a data point that was manually fed to us instead of electronically fed to us and someone might have fat fingered a month or a year even sometimes. And then all of a sudden, you have data in your data set that is a year of. Are we able to do this with AI yet? AI can take spreadsheets now. You can have it feed in and look for those outliers for you. But ultimately, unless your company organization has this huge data link, which to [indiscernible] it's still going to take a little bit of legwork to understand if those outliers might be true outliers as in maybe something within an FDA or customs hold and really just sits on work for a couple of months or opens the data entry error. AI is not going to be that smart to triple that down. But eventually, it will be. So that would be -- how I would deal with maybe some accuracy dealing with processes and steps to address that.

Thank you so much, Longjie. Very good. So we do not have any additional questions anymore. So I thank you everyone for taking the time. Be on the lookout for the survey and then for the presentation. And please don't hesitate to reach out to any of your Expeditors' contacts, if you have additional questions or would like to learn more, and we hope that you will have a wonderful day. Look forward to seeing you on our next webinars. Bye.