How AI Will Transform the Chemical Enterprise Over the Next Decade

Anthony Schiavo: Hello and welcome to the webinar, How AI Will Transform the Chemical Enterprise Over the Next Decade. My name is Anthony Schiavo, Senior Director and a Principal Analyst here at Lux Research, and I’ll be moderating and presenting during today’s session. Presenting alongside me today is my colleague Akshay Chaudhari. He’s an analyst here at Lux Research. Throughout this webinar, you can type any questions you have in the questions box on your screen. Time permitting, we’ll answer all the questions that we can. If your question does not get answered, please do not hesitate to email it to [email protected] and we will respond. If at any point you experience technical difficulties, such as a frozen screen, simply refresh your browser and check that your internet connection is strong. Now, before we start the webinar, just a quick word about who we are. Lux Research helps organizations make more confident innovation decisions about what matters next. As an independent research and advisory firm, we work with many of the world’s leading companies to identify emerging opportunities, evaluate technologies, and make high-impact innovation decisions. Our scientists, engineers, analysts, and industry experts combine original research and methodologies with practical decision frameworks to help organizations separate signal from noise. Every day, we help clients answer three critical questions. Where should we focus our innovation efforts? Which technologies deserve investment, and which partners can accelerate innovation? And today we’re going to share some of that thinking with you. And before we get into it, I want to start with what is maybe my favorite chart over the past year. This is from the Federal Reserve Bank of Dallas. It is a chart of their AI scenarios and the impact on GDP. You can see they’ve mapped out four scenarios here. There’s a baseline case, there’s the AI boosted GDP scenario where GDP goes from 1.9% growth to 2.1% growth for 10 years. And then they have two alternative scenarios. The singularity benign, in which GDP goes infinite, money ceases to exist, we’re all living in a utopia. And then we have the singularity extinction, in which by 2040 or so, GDP goes away because humans have gone away. So with predictions like these, and I see you cracking a smile, Akshay, it’s a little challenging to know where we stand with AI. I think today our goal is to provide a slightly more grounded view in terms of the impact on the chemicals industry. So we’re going to start with a little bit of an introduction, talk about what AI is, what it really does. Then Akshay and I are going to talk about some of the applications in the chemicals industry. And then we’ll give a bit of a broader outlook to that bigger picture question of where will the chemicals industry be in 10 years with its use of AI? I want to start with just a bit of a definition, and if you’ve listened to our webinars in the past, you’ve heard this before, but I think it’s important to stay grounded. AI is a probabilistic tool. You can see here an example from GPT-2, which is quite a simple model. You can see it’s making a probabilistic prediction about the next word in a sequence, so the best thing about AI is its ability to learn. That’s the most likely at 4.5%. And you repeat this prediction and you can build sentences from it. More advanced LLMs work on very similar principles. Now they’re making much more complex predictions, but ultimately they’re the same. And there’s a degree of randomness baked into this. So you can see most of the time it’s going to pick learn, but some of the time it’s going to pick predict or make or understand or do. And this is described in that temperature chart. AI can approximate complex functions. If we want to use this AI, we can approximate the function on the left, which outputs 1, 0, or negative 1. If we use a more complex neural net, we can make an increasingly accurate prediction about this function. And we can use data to train AIs to make these predictions better. If we want to recreate the function on the left, again with outputs of negative 1, 1, or 0, we can use a neural net, and we can improve the quality of that prediction by providing it more examples and adjusting the results. So with 10,000 examples, our neural net does a pretty bad job. With 10 million examples, it does a pretty good job. So you put all this together, and you have that definition. AI is a probabilistic tool. It’s trained on large amounts of data, and it can replicate the outputs of complex tasks. So that’s what AI is, but what does AI really do? Well, we think about it two ways. First, AI is a technology of automation. And then secondly, AI is a technology of seeing. I’ll talk about each one in turn. So as a technology of automation, AI is a tool that mediates human labor. This is really not new. You have tools going back to the invention of the assembly line. You mediate that human labor. You take a complex single task of manufacturing a good, and you break it down into individual steps. And in this way, you can automate the labor, you can fragment the labor, turn it from this big complex thing into many simple things. You can reduce the skill level needed to successfully build a car, and you can casualize the labor. Someone can learn to work an assembly line in just a few weeks. And we’ve seen this happen with manual labor through things like the assembly line. We’ve seen this happen to skilled and service labor, things like Uber, automating, fragmenting tasks of service, and many other types of services have now been uberized. And AI is continuing this pattern for knowledge labor. You can take a big complex knowledge task and you can automate it, fragment it, reduce the skill needed to complete it. But AI is also a technology of seeing. It makes knowledge visible. Technologies of seeing do this by simplifying information, discarding certain information and scaling that information. And this is true of the subway map, which discards information about, for example, the position of buildings or geological features. It simplifies these complex train lines into nice, clean, readable lines, and it scales it down so you can see the whole system at a single glance. You know, without this map, it would be very difficult to navigate the subway system. You’d have to talk to someone and get directions from an expert who really knew the system. And when you combine these functions, they become more powerful, right? AI can automate decision-making by surfacing knowledge, transforming it, scaling it, and then acting on it. With this in mind, I want to turn it over to my colleague Akshay and talk a little bit about some of the applications that we’re seeing in the chemicals industry itself. Akshay.

Akshay Chaudhari: Thanks, Anthony, for the overview. So let’s get specific. Let’s look at the applications of AI, especially in the chemicals enterprise. So we see four main use cases here. They are R&D, manufacturing, customer support, and sales. So starting with R&D, this is about speeding up materials development, reducing the overall resources required for materials development and achieving a particular goal. And again, that can be material discovery, optimizing materials performance. So let’s look at some of the examples of it. So one of the most common uses of AI, especially in the chemical industry, is the use of AI as an alternative to classical simulation. So a company like Preferred Computational Chemistry, PFCC for short, it has launched a tool called Matlantis, which uses neural networks to accelerate density functional theory (DFT) calculations. So Matlantis uses a database of around 60 million DFT simulation results. And it’s used that data set to train the neural network and predict the outcome of DFT simulations. And this is much faster. You can see the scale of acceleration it can achieve. And that’s because of the linear algebra needed to make these calculations. Basically, it’s very fast. And at the same time, by increasing the number of atoms you could simulate, it actually increases the scale and gets better results with this. We also have GPUs, which are specialized pieces of hardware designed to do these types of calculations very quickly. So here you can make predictions with roughly around 98% accuracy, much, much faster. This is really powerful. All of a sudden, you can simulate, scan almost millions of molecules, which you could not before. So there are, of course, some limitations. This is difficult and expensive. Generating 60 million simulation results requires access to supercomputers. So this is not something that just anyone can do. And it’s not necessary that something of this sort will be done by a chemical company. So that’s why these tools will be very specialized and will be used for very special use cases, especially accelerating the simulations. Now, when you look at the next use case, AI is now being used for discovering new materials. And it’s not just the molecules, but also the complex materials and their structures. So companies like Cusp AI use a combination of AI as well as lab automation to accelerate discovery of materials like metal-organic frameworks. So Cusp AI has developed this multi-agentic platform that can generate more structures. And the company also aims to kind of combine this generative discovery with self-driving laboratories. You can test the materials, develop the data, and feed that data back into the models and make those models better and better as you generate more data and train it further. So Cusp AI has raised a lot of money. But again, that’s necessary because the whole process is pretty expensive. You have to invest a lot of money in building those self-driving laboratories. The challenge is that the discovery is only the first part of the puzzle. You can discover these complex structures, which are really great, but can you actually produce them? The AI can accelerate discovery of MOFs or semiconductor materials, but manufacturing, including synthesis, shipping, stability, cost, and also achieving that scale is still unaddressed. Now, similarly, we have seen AI use cases in manufacturing. So the main purpose of using AI in manufacturing is of course reducing the downtime and improving the planning. So when it comes to predictive maintenance, especially that directly impacts the downtime, it’s not something very cutting edge. It has been tried for a long time, and there are very mature use cases, especially for the chemical industry. One of the examples we have here is C3AI and its use of process-specific models to spot issues before they cause downtime. So, some time ago, we interviewed one of our current customers, a major US chemical company. The company had experienced several issues with stoppages due to furnace failures, and it had evaluated C3AI’s solution some time ago, but decided not to work with the company at that point. But after these failures, it came back to C3AI. During this intervening period, C3.ai had developed process-specific models, including furnace-specific models, which the company could directly take and deploy and check with their historical data and see what they can achieve. And they found that with this backtesting and then the models that C3.ai has developed, they could have identified those furnace failures ahead of time. So the person we interviewed told us that the slowest part of adopting this system was actually getting the paperwork signed. The legal agreements, NDAs take a lot of time. So once the engineer and the management saw that the backtests were positive and achieving the intended results, everyone was convinced. So the deployment has been very successful after that and the company is now rolling this more widely across the organization. So the key point is when the problems are well understood and there’s clear data to back-test all these different modeling tools, adopting AI can be relatively easy and of course it can achieve better results after implementation. However, there are many emerging areas where the use of AI is not mature. And one such example is production planning. So chemical companies have large, complex logistical and forecasting problems, but they still rely on relatively straightforward and rule-based approaches to forecast their production requirements. And to my knowledge, no one at this point is using AI to directly forecast production requirements. But AI is very powerful in its ability to write code. So BASF has partnered with Google and is using AlphaEvolve. It’s a coding agent. So BASF provided its existing rule-based approaches, its forecasting software and historical data, and asked the agent to iterate through alternative modeling approaches. And BASF says that this has improved the outcomes of its modeling and forecasting capabilities. What is interesting is that, for a company like BASF, the advantage of rule-based approaches is that they are fully explainable. You can understand exactly how the forecast was developed. There’s no black box. By using AI as a programming agent, BASF can access a much broader and more sophisticated range of modeling approaches. You don’t really have to go for modeling the whole supply chain directly. In this case, AI supplements BASF’s programming expertise, and it gives additional engineering capabilities to the team. So with that, I’ll hand it back to Anthony to talk more about customer support.

Anthony Schiavo: Thanks. And I think a lot of the interest that we’ve seen coming from our clients is in R&D and manufacturing. But in our view, it’s also very important to consider these other applications in areas like customer support and sales. In customer support, it’s really a lot of basic automation. We see that with a company like Brenntag, which is a well-established chemical distributor. They are leveraging Salesforce’s Agentforce, which is the AI platform built on top of Salesforce, to automate customer service requests. I think what’s important to recognize here is that these LLM agents, they’re really acting as a human and machine interface. They’re really functioning to take these customer requests in, interpret them, and then ultimately trigger some type of automation that’s already been well defined and established within that existing customer support automation platform. Whether that’s triggering a refund or a reorder, these are established routines that now can be activated with a lot less human oversight. So it allows this automated system to handle a lot more cases, but it’s not a replacement for existing customer support or automation capabilities. Really, it’s an additional capability that needs to be built on top of those platforms. And we also see this with Covestro. They have partnered with Lloydway to automate document retrieval with AI. I don’t know if anyone’s ever tried to go on the website of a chemical companya chemical company, but it’s often not the greatest experience, if I’m being very honest. Chemical companies have tens of thousands of products often. They have very complex codes and systems for identifying these products, and they have a lot of complex documentation. And if you’re an end user, you need to access that right documentation, but you don’t have the skills or the knowledge to navigate a whole chemical company’s product portfolio. It’s often quite a daunting task. But this is the kind of task that LLMs are really quite good at, the complex search and knowledge-retrieval capabilities of LLMs make it much easier for non-experts, and in that sense I mean people who are not experts in Covestro’s product line, to find data sheets. And so these small, automated, simple tasks of the company’s own knowledge are really a great first target for automation. And you can see that with the speed with which a company like Covestro can move forward here. But it’s not just in customer support. We also see this in sales. Reaching new customers and providing new services, it’s all enabled by AI. There’s a couple interesting examples. The first is material search. Evonik has developed Cotino, which is actually a materials informatics platform. It’s kind of what we talked about in the R&D use case. But here it’s being deployed in a much more customer-facing sales type of way. Cotino is essentially a materials informatics-informed search platform. It allows end users to identify, hey, these are the challenges I’m trying to solve with my coating, such as defoaming, needing different dispersing capabilities, and then easily match that to Evonik’s product platform. What is interesting is that Cotino is free. All you need to do is register an email, sign up, and anyone can use this. You don’t even have to be a customer, a paying customer of Evonik. Anyone can go and play with this. And by giving it away for free, Evonik is essentially making its own products much easier to buy because its customers can, or potential customers really, can very easily identify which is the right product that’s going to solve my problem. And interestingly, Evonik has continued to iterate on this. When it originally launched, it was really just that materials informatics platform with a somewhat complex user interface. It still required a fair bit of expert knowledge. But now they’ve layered an LLM interface on top of that, and that makes it much easier to use. And now you can chat very naturally with the system and get answers to the product questions that you have. Again, making it much easier to buy. But you can even go further. Dow has partnered with Luuna, a company in South America, to sell mattresses. And they’re leveraging a foam formulation materials informatics tool. This develops plausible formulations for foams with targeted properties. So you can have a foam that’s stiffer, softer, has the right kind of support. And this tool, they claim, allows them to cut the time it takes to validate these formulations from weeks to hours. And they can very quickly get to a working formulation. And with this, now all of a sudden, it’s cost effective to target much smaller market segments. You can have custom foams for different product lines, as opposed to maybe just having one foam that goes in every mattress or maybe just a few types of foam. This now, combined with this sort of market channel partnership that they have, allows them to bring differentiated competitive products to market much more quickly and at much smaller scales than what would have been required previously. We can think about all these applications in our Lux AI framework. This has two elements. We think about the value and the value model and then the prioritization. The value framework really, as it says, tries to capture the value of these applications. So knowledge value captures the impact the application has on company performance, the availability of knowledge, the impact on other parts of the business. Basically, is this AI application going to make us better at our jobs? And then automation value captures the value in automation, the standardization of the tasks, how frequent the tasks are, how much effort do they take? Ultimately, how much value do we get out of automating this task? But prioritization is different from just saying, all right, what’s the most valuable application? We’ll do that one first. We hear very consistently from our clients that CEOs are expecting results with AI in six months or maybe even less. So something that’s going to take years to develop is really challenging. So in the prioritization framework, we have that time to value. That’s our first key metric. And then we also have a measure of risk. This tries to capture the likelihood that these things are going to fail due to internal barriers, misalignment, employee rejection, technical difficulty. When you map out all these different applications for the chemicals industry, what you can see is there are a lot of valuable approaches, first of all. There’s a lot of really interesting tools here. There’s a lot of capabilities that are potentially valuable for the industry. But it’s those customer-facing approaches that really stand out in terms of value because they’re often combining a nice automation feature, you’re automating some elements of that sales funnel or automating some elements of that customer support funnel, combined with a deeper knowledge of your business and of your customer. If your customers are using your platform for their own materials development, that provides an incredible source of insight—an incredible view into what your customers want. The R&D tools, those score very high in knowledge value, but not quite so high on the pure automation value. And then things like document retrieval are a little bit more like nice-to-haves. What’s interesting is, okay, these customer-facing sort of approaches, they stand out in terms of value, but they’re also some of the more near-term opportunities to deploy AI. Things like document retrieval are very easy, very fast to do. A lot of the customer support, if it’s built on existing infrastructure, it can be relatively quick to roll out. And even something like material search, that’s actually a relatively basic implementation of MI. At this point, it’s not that challenging to deploy, whereas things like discovery and simulation are really quite a bit more challenging. So I think we see this situation where we actually really see AI having its biggest impact in the chemicals industry on that customer relationship. And then materials, the core R&D, core production, that type of impact will take longer to really develop. So, Akshay, what do you think about, as we go forward with the chemicals industry here, how are you really thinking about how the chemicals industry is going to be looking in this longer term?

Akshay Chaudhari: Yeah, absolutely. And I think this quote here from an Eastman Chemical executive in 2026 actually summarizes it quite well. It talks about how an intelligent AI linked to all data-generating equipment can actually help gather the knowledge that is generated by the team and how it will allow progress toward the sophisticated experimental design and modeling. And I think it really captures the spirit of what people want to accomplish with AI and how they want to move the R&D process forward.

Anthony Schiavo: Yeah, definitely.

Akshay Chaudhari: But well, actually, I lied to you. This is not a quote from Eastman Chemical in 2026. It is not about AI. It’s a quote from an Eastman Kodak executive in 1998. And it’s about electronic laboratory notebooks. The original quote said that all scientists would use an intelligent electronic laboratory notebook linked to all data-generating equipment. And the point I want to make is that chemical companies have been pursuing this vision for a long time. Along the same lines, chemical companies have been optimizing R&D and production for a long time. Many of the ambitions we now associate with AI are basically a continuation of much older efforts to make research and production more systematic, connected, and efficient. And this is actually quite well summarized by the case of ammonia. So if you look at ammonia production, in the 1920s, it took around 100 gigajoules to produce a ton of ammonia, whereas today it takes about 28 gigajoules to produce the same amount of ammonia, and the theoretical limit is around 21 gigajoules. So we are already getting quite close to what is physically possible through optimization, and there is limited scope for any further improvements within the existing infrastructure. So where I really think AI is going to be transformative is in enabling chemical companies to change their customer relationship and chase the long tail of customers. So historically, it has not made sense for multi-billion dollar companies to chase these $100 customers. When you look at reward relative to effort, it would take a lot of effort for the company to kind of chase these customers, but the rewards simply are not there. So what AI is doing, especially when we talk about reaching this long tail of customers, is actually reducing their efforts by automating tasks and enabling customers to self-serve, which was also evident in some of the examples Anthony shared. It does not really increase the reward from each small purchase, but it dramatically lowers the efforts needed to capture customers. And all of a sudden, it becomes much more viable for chemical companies to target these smaller potential purchasers. And what it is basically achieving is with this interesting approach is that a huge amount of knowledge is being generated. And that knowledge is becoming increasingly visible to executives. And even as parts of sales roles are being automated, this knowledge is being captured through these customer interactions. There’s a huge amount of tacit understanding, basically what we can call metis. And that lies in the heads of salespeople, customer-facing employees. And their understanding of customers’ problems and how they solve those problems basically comes from that. And as you start to automate part of this work, you also make that knowledge more visible to corporations, particularly to the executives, managers, as well as planners. If a customer uses your materials informatics tools to answer questions on an online platform, all of that data becomes visible to you. You get an incredible pipeline of understanding about what customers want, what problems customers are facing when it comes to materials, new materials. Instead of all of this remaining in employees’ heads, where it is difficult to extract, AI makes it available as large-scale data through these platforms. You can use it for forecasting, targeted R&D, and product development. And that can be really, really valuable in this case. However, there are definitely some risks. So corporations are really creatures of techne: rational, systematic knowledge derived from universal principles and first-principles thinking. They love their Excel spreadsheets, memos, formalized knowledge, record keeping. But corporations also rely heavily on metis, the practical, specific local knowledge developed through practice and experience. So salespeople have it, the R&D engineers have it, the people who work in plants and refineries have it as a process know-how. So organizations are often very good at managing techne, all the process knowledge, but they are very bad at managing metis. And this leads to some risk, and I think Anthony will be able to share an interesting story about it.

Anthony Schiavo: Yeah, absolutely. Because I think whenever you have this new type of technology of seeing or technology of knowledge, if you want to call it that, you transform metis into techne. You take information and knowledge that was only from this hands-on experience and you transform it into something that comes with universal principles, right? But when you do this, something is always lost in translation, right? It is never a one-to-one. There’s an interesting story. Back in 1765, they had developed sophisticated new methods of assaying and bookkeeping in Prussia. And this is essentially the first development of those spreadsheets that we know chemical companies love so much, right? And the Prussian government had a big problem, which is that timber yields were very low. With these new bookkeeping methods and combined with some surveying methods, they were able to identify areas of forests that were high value, high yield, and successfully cut them down. And then when they went and replanted those forests, they used those same spreadsheets to calculate exactly how many trees they’d need per hectare to produce the yields they wanted. And so they went and planted these new forests exactly conforming to those spreadsheets, right? They had these nice rows of trees in very straight lines. And it was very wonderful for the surveyors because all of a sudden you could just walk through the forest and see that what was literally on the ground matched what was in the spreadsheet. The map had transformed the territory. And all was well until the forest died. Because it turns out that a lot of that undergrowth, all that messiness, all of that other stuff that the spreadsheets didn’t put a value on, it was actually really necessary for keeping the forest alive. So as chemical companies go through the next decade and start to really automate tasks using AI, they have to ask themselves, do they really know what they’re replacing? Do they know what knowledge, what skills might be lost when they do that automation and when they adopt AI? So with that, we just want to leave you with a few final thoughts. First, we really think AI will unlock new types of customer relationships for chemical companies. It’s going to enable chemical companies to chase this long tail of small customers, right? It’s going to enable more services for buyers, right? The uses in R&D are valuable for sure, and in manufacturing, they’re going to take time to impact the chemicals industry. And it’s simply just a fact there’s a lot of business processes between R&D and the bottom line, right? So it’s really going to slow the time to impact. But it’s important to keep in mind as the chemical industry adopts these technologies that when you automate tasks, you lose metis; you lose that practical knowledge. And it’s very difficult to know what kind of practical knowledge you have and what’s at risk of being lost. So we really advise companies to tread slowly and carefully as they deploy AI and monitor those results so they can really chart a safe path through the next decade. So with that, we are going to now begin to open it up for questions. If you have questions, you can type them in your questions box. If we don’t get to your questions, we will be in touch. Someone will be in touch after this webinar. Akshay, we’ve seen a few questions come in during the call here. And there’s one I think you might have a perspective on. You mentioned Cusp AI and sort of GenAI generally, but there’s also, this is a space where we’ve seen companies like Microsoft get involved. Google has its like DeepMind GNoME project, IBM. You know, we got a couple questions about these types of bigger companies, bigger tech companies getting involved. What do you think of these sort of bigger, you know, generative AI efforts? They’re big headlines, you know, 20, 200 million materials identified or a bajillion crystal structures. What do we think about this? I know this is something you’ve looked into.

Akshay Chaudhari: Yeah, I think the amount of funding that’s going in this area is also significant. More than a billion dollars have gone into these companies, especially new startups that are working on material discovery. And one thing that is increasingly becoming clear is materials discovery in itself is not as much of a challenge now. So when you actually discover those materials and molecules, how do you synthesize them? How do you basically do characterization? How do you do application-specific testing? And that generates a lot of data. And, of course, there is a risk of failure, and failures in subsequent downstream materials development increase the cost. So it’s very easy nowadays to come up with a new material, but can you really test it, design it for this particular application, and manufacture it at scale and at a reasonable cost? That’s the major issue. And I think currently there’s too much attention on material discovery. What we need to also do is focus on these downstream steps as well. And perhaps there’s more opportunity for the startups and also large companies to invest and also apply AI in those subsequent stages.

Anthony Schiavo: So, a healthy dose of skepticism for the Googles and the Microsofts of the world here. And one other area, we don’t have a ton of time for questions, but we had a couple of questions here on China, specifically, in a couple of different ways. It’s hard to talk about the future of the chemicals industry, I think, without talking about China. So basically, one of the questions was, will this allow companies to compete with China, chemical companies in the West to compete with China? And I think the short answer is no, unfortunately, in part because we really see that Chinese chemical companies in a lot of ways are leading here. You know, when it comes to areas like e-commerce, for example, this is an area where the Chinese chemical companies have really been historically very strong and there’s a lot more adoption of e-commerce as a chemical sales vehicle in China. And I think what is critical to think about is that in the West, however you define that, we have a lot of pre-existing relationships with our customers and with even companies like distributors. When I talk to chemical companies, they often say we really don’t want to disturb our relationships with distributors. Maybe selling directly to customers is a risk we’re not willing to take. But I think in China, as that industry has grown over the last decade, they aren’t as burdened with those established relationships, with the old ways of doing things, right? And so they’ve been able to experiment and I think adopt a lot more quickly and a lot more freely a lot of these technologies over the past, this sort of past wave of digital technology. And I think you’re going to continue to see that in China, that the adoption of things like agents for these more customer-facing roles, even buying and selling, it’s going to be able to be adopted a lot more quickly. Not necessarily because they have access to much better AI tools, but because I think that there is a business culture and a lack of these sort of really established relationships and established ways of doing things that hinder implementation. So I think it’s going to be incumbent on the Western companies to leverage AI more quickly here to help them catch up and maybe pull even with China. But I don’t think you’re going to see AI as something that the Western companies are really uniquely capable of leveraging to move forward with competing with China. And so with that, I think it’s about all the time we have today. This is going to conclude our webinar. The slide presentation and the recording from this webinar are going to be sent to all attendees today. After leaving this webinar, you’re going to be prompted to complete a survey on today’s presentation. We really appreciate that feedback. We always look at that. It really helps us inform our webinars. It helps us improve future webinars. So we genuinely appreciate it. Also, I encourage you to take a moment, check out our upcoming webinars. You can go to our website here. And we thank you for joining us and hope you have a great day. Thanks. Thank you.