The world has seen its fair share of supply chain challenges for several years now. Where historically supply chain management has been all too reactive, artificial intelligence (AI) is enabling a more proactive approach. “AI’s role in supply chain resilience is really about turning uncertainty into intelligence,” says Vaidy Raghavan, chief product and technology officer for Xometry.

Providing an on-demand manufacturing marketplace that connects businesses with vetted production services gives Xometry incredible insight into a range of manufacturing trends around the world. So, for this episode of Manufacturing Matters, Raghavan joined TECH B2B Marketing’s Aaron Hand and Dan McCarthy to talk about how AI is changing the face of supply chain management, reshoring, and trade uncertainties — moving from reactive to proactive, automating decisions, and leveling the playing field. Small and medium-sized manufacturers can operate with the same level of intelligence and agility that was once reserved for larger players, Raghavan notes.

Aaron Hand: [00:00:01] Hello and welcome to this episode of Manufacturing Matters, where we talk about the technology and trends that are shaping the global manufacturing industry. I’m Aaron Hand with Tech B2B Marketing, here with my colleague Dan McCarthy. Hey, Dan.

Dan McCarthy: [00:00:14] Hello.

Aaron Hand: [00:00:16] We’re joined today by Vaidy Raghavan, who is chief product and technology officer with Xometry. Welcome. Thank you for joining us.

Vaidy Raghavan: [00:00:25] Thank you. It’s great to be here.

Aaron Hand: [00:00:28] So, for those of you who are not familiar with Xometry, they have a platform for custom manufacturing on demand. They connect engineers and businesses with a network of suppliers offering custom parts production. So, maybe to kick us off, you could tell us a bit more about that. Also, tell us about your role with the company.

Vaidy Raghavan: [00:00:49] Sure. So hi again everyone. My name is Vaidy Raghavan. I lead product and technology at Xometry. I’ve spent my career building technology platforms that simplify really complex systems, and manufacturing is perhaps one of the most fascinating of them all. At Xometry, as you mentioned, we are using technology, data science, and AI to connect engineers and manufacturers in smarter, faster, and more resilient ways. Our mission at Xometry is to transform how the world makes things, and my role —— and the rest of the product and technology team’s role —– is to turn that belief into a scalable global reality. To now get into the second part of your question about Xometry’s place in the manufacturing ecosystem … Xometry sits at the center of the modern manufacturing ecosystem, connecting demand from engineers and buyers with supply from thousands of qualified manufacturing partners. The best way to think of us is as an AI-powered marketplace for custom manufacturing. When a customer uploads a 3D model or a technical drawing, our platform instantly analyzes it using machine learning models trained on millions of parts to generate an accurate quote, predict lead times, and match it with the right supplier.

Vaidy Raghavan: [00:02:07] Now, one of the interesting things about custom manufacturing is, if you think of any other marketplace, say, Amazon or Wayfair, you walk into the store and you point at either a physical shelf or a virtual shelf and say, “I want that.” Whereas when it comes to Xometry, the customer walks in with something that’s in their hand, a 3D model or a technical drawing. And then they say, “I want this.” And our job then becomes being able to spot what the drawing is about and analyze and give them a quote within seconds. Now, behind the scenes, what we have is thousands of small and mid-sized manufacturers, CNC shops, injection molders, sheet metal fabricators who use Xometry’s Mitre to keep their machines running at high utilization. We handle coating, payment, and logistics and quality management so that they can focus on what they are best at, which is making precision parts. In that sense, we are both a technology company and a manufacturing enabler. We’re not necessarily competing with manufacturers. What we are doing is helping them grow, modernize, and participate in a digital supply chain that’s faster and more resilient.

Aaron Hand: [00:03:15] Okay. All right. Well, that sounds like a very interesting model. I would assume on a day-to-day basis, you’re dealing with tons of data and that gives you incredible insight and a vantage point into the manufacturing trends with all those connections. So tell us a bit about what you’re seeing right now in terms of how manufacturers are adapting to changing tariffs to supply chain disruption or reshoring efforts.

Vaidy Raghavan: [00:03:44] Yeah, you’re absolutely right. The beauty of operating a marketplace like Xometry is that we get a real-time pulse of what’s happening in manufacturing. So every quote, every part, every sourcing decision that we make essentially creates a data signal for us. And when we aggregate over millions of those, you start to see macro trends well before it shows up in journals or government reports. And what we see is —— I can share three trends. One is near/reshoring. Second is focused on data, and third is navigating trade uncertainties. Let me start with reshoring. What we are seeing is a regionalization trend that’s quite strong, which is not necessarily deglobalization, but smarter globalization. Manufacturers are reevaluating their supply chains through the lens of risk. Previously it used to just be cost. And this means nearshoring and reshoring activity. But it tends to focus on critical and high-precision components where lead time quality and intellectual property protection matters more. So that’s one. The second is, what we are seeing is suppliers are getting a lot more data savvy. Shops used to traditionally be focused a lot on tribal knowledge, and now they are shifting to leveraging digital coating, automated scheduling, and analytics to win their business, which means AI isn’t technically replacing, but it’s definitely helping them compete globally. The third is around tariffs and ongoing trade uncertainties. What we are seeing, the impact of that is a reinforcement that the importance needs to be on resilient capacity. We are definitely seeing companies don’t want to have all their eggs in one basket. Our data shows that there is a shift from single-source dependency of a manufacturer in one particular region to a distributed network, and fortunately, that’s what Xometry enables as well. So the big picture as we see it is this: manufacturing is becoming more local, more digital, and more responsive. And the companies that combine these three dimensions are definitely gaining share right now. And the competitive advantage we see isn’t about the cheapest supply chain, but the most adaptable and most resilient ones.

Dan McCarthy: [00:06:07] That’s a good high-level view. I’m curious, though, what kind of role does artificial intelligence play in helping manufacturers mitigate the risk of supply chain challenges? And not only that, but I mean, obviously there’s always been a reactive component to this, to mitigating this. Does it allow people to become more proactive somehow? Does it actually anticipate trends in some way?

Vaidy Raghavan: [00:06:29] Absolutely. I think you hit on the key points already and give me give me a moment to elaborate. I think AI’s role in supply chain resilience is really about turning uncertainty into intelligence, much like where you were going. And I would say, I’d highlight four things here. One, as you brought up, moving from a reactive approach to proactive. Second is automation of decisions. Third is insights for forecasting and simulation. And the final one is leveling the playing field. So I’ll start with the first one. Traditionally supply chains tended to be reactive, which means you found something was broken after it broke. But what AI allows you to do is shift from a reactive mindset to predictive. And I’ll give you an example at Xometry. We have models that analyze millions of data points through our quoting activity, through material availability checks to lead time checks, capacity, and pricing, which means we’re able to detect early signals of stress in an ecosystem. So if a lead time in one particular process for one region starts to creep up, the system can automatically reroute demand to another region where the supplier does have available capacity. And this is something that’s happening right now in action. This is not abstract and it’s deeply operational.

Vaidy Raghavan: [00:07:48] The second is beyond just reacting and rerouting. AI helps manufacturers stay ahead of the curve, and it does so by guiding smarter material choices, so you’re able to forecast demand patterns ahead of time. As long as you have a ton of data, you’re able to figure out what’s likely going to happen in the course of the next week, next month, and next quarter, and you can start simulating production scenarios even before you’ve cut a single chip. In other words, I think manufacturers are making better decisions faster as a result. So that’s the second one.

Vaidy Raghavan: [00:08:22] The last one I would probably point out is the broader opportunities as AI is becoming more of a co-pilot for manufacturing operations. We don’t particularly see it replace human expertise, but it is absolutely augmenting it. So what we’re seeing is small- and medium-sized manufacturers, they’re able to operate with the same level of intelligence and agility that was once reserved for larger players that had a massive infrastructure. And so net net, I think AI doesn’t eliminate risks, but it definitely gives you a little bit more time and more options.

Dan McCarthy: [00:08:57] I know you can’t name names, but is it possible to kind of give an example of a success story that somebody had with a particular client? Maybe a particular sector, for example?

Vaidy Raghavan: [00:09:07] Absolutely. I’m actually able to share a story that we’ve … We have a very successful story that I’d like to share with you all. It’s a company called SentriLock. They are a leader in electronic lockbox solutions for real estate; essentially it gives real estate agents flexible access to properties via codes or the SentriKey app. Now, they were faced recently with a major supply chain disruption while trying to onshore manufacturing operations. So what had happened was their design of a lockbox component failed validation, and there was an urgent need for new designs, prototyping, and reliable suppliers. SentriLock partnered with Xometry. Now they were able to set very specific parameters of cost, lead time, specification, tolerances, and so on. And because Xometry has two things —— one, we have an extensive supply network. And second, we offer a wide selection of manufacturing technology —— it meant that we could do rapid prototyping using one approach additive and then efficiently transition to production using urethane casting. And so we were able to minimize lead times. In addition, Xometry also managed logistics. So what we did was, we had the tools and the initial batch of units manufactured in China. We shipped them to us. And then in parallel we used ocean freight, and we moved the tooling onto US soil so that ongoing production could happen. Now, by doing this centrally, SentriLock was able to achieve tremendous results in this partnership. One, they reduced part count by 25%. They shared that they saved annual material costs by about $780,000, and they were able to double daily production from 700 to 1300 units. So, (a) they achieve lower cost, (b) they improve their availability, and (c) they made their whole assembly and production line far more simple and reliable. The big thing they gave Xometry a shout-out for beyond all of this is having a single point of communication. That way you’re not having to coordinate with multiple suppliers, multiple stages. We’re able to be a one-stop shop.

Aaron Hand: [00:11:13] Wow. So when you’re talking about, you know, the different approaches that they could try, like additive approach versus casting, did they end up going with both or were those choices that they were making, and were they making decisions then about what materials to use in either of those cases? Or…?

Vaidy Raghavan: [00:11:31] Absolutely. So in this particular scenario, they needed rapid prototyping for the initial phase to validate what they wanted. And so they worked with us on what’s the fastest way to do it. Now we, of course, have expertise and we will guide the customer, but we will start with what are the parameters that they have set on materials, tolerances, lead time, and cost. And we will work within those parameters and give them the guidance. And we were able to achieve that sort of split-approach where prototyping was through additive and production was through urethane casting, and it was able to hit all those parameters.

Aaron Hand: [00:12:05] Okay. All right, great. I just wanted to make sure I was understanding that correctly. So, we’ll bring the view out a bit. I know that Xometry has put out a global market report. I believe this is the first time you’ve done that with a manufacturing outlook. So, there’s a fair bit of focus on artificial intelligence as an operation enabler. So tell us from a broader view —— I believe that you showed that 82% of manufacturing executives view AI as a core growth driver. So how does that perception translate into the kinds of actions that factories are actually taking or how they’re treating their supply chain?

Vaidy Raghavan: [00:12:48] Yeah, that stat actually really stood out to us as well. 82% of manufacturing executives now see AI as a core growth driver. What’s encouraging is that this isn’t just optimism. We’re seeing it translate into concrete actions. And I’ll share three ways that we’re seeing AI being deployed. One is insights. Second is automation and third is enablement. So on insights, what we’re seeing is companies are using AI to mine their production and sourcing data, so they’re able to better understand where the bottlenecks occur and how costs are moving and where there are hidden inefficiencies. So that kind of visibility used to take months before, with manual analysis and so on. Now it’s happening in real time, as long as you’re able to wire in your data points into a constantly learning system. The second is automation, which is about streamlining repetitive, high-friction processes. And Xometry is a classic example where we have an instant quoting engine, which analyzes parts and drawings within seconds and gives you a quote. But we also have other areas where we have deployed AI, and we are seeing the same thing in the marketplace, whether it is quality inspection or scheduling. The last one I would share is enablement, where AI is helping manufacturers make better, faster decisions by giving them these predictive insights that we spoke about a little while ago about planning tools and faster feedbacks. So net net, what we are seeing is companies are definitely able to move faster, even if it is very small steps —— and they’re able to gain measurable advantages, whether it is faster request for quote turnarounds, lower rework rates, or higher on-time delivery. So the message we tend to share with folks is, it’s clear that AI adoption is no longer just a strategic thing. It’s quite operational. And the winners are the ones that have moved from talking about AI to starting to train their data and their models with the information they already gather.

Dan McCarthy: [00:14:48] But are there specific AI applications that you’re seeing adopted specifically for manufacturing? You’re talking, sorry, high level terms here, but I’m curious if we can dig down into that a little bit further? And specifically how they differentiate, how they’re differentiated from traditional automation or digital tools. What’s different? What’s changed?

Vaidy Raghavan: [00:15:07] Yeah. So maybe I can share one story from the customer side and another from the partner on the customer side. The expectations have changed. A few years ago the norm was manual quoting. So you would give a drawing or you would give a CAD file and you would wait for hours, days, sometimes weeks. And when Xometry started to give instant quoting, that used to be considered cutting edge. Today that’s baseline. Now what we are trying to double down more on is, take that instant quoting mode and give far more capabilities and more processes. So recently we in Xometry launched instant quoting for injection molding. It’s the first time we have rolled that out. Injection molding is a more involved, complex manufacturing process where you have to quote for the mold or the tool, and all the parts and the units that you can mint from it. And these things change and vary depending on the material that you use for the mold. And it also varies based on where you choose to manufacture the mold versus the units. And the customer sometimes can choose and say, “look, I want you to manufacture the mold in one place, but in between production runs, I want it to be moved to this other location.” All of these are variable elements. We’re able to do instant quoting for all of that because of all the information that we already have.

Vaidy Raghavan: [00:16:28] So that’s one application. The second I would probably share is on the partner side, where we are seeing AI models determine the best manufacturer to match. Now we also do the same thing where we have a machine learning capability, where we take for every manufacturer we have the machine capability, the part complexities they are more comfortable with —— which we track —— and their performance. So every time they take a job, we track their performance on how many times they ask questions, how quickly they’re able to manufacture, how much is there on time. And so we have a proprietary method of calibrating partners as they make their way into our network. So we’re able to quickly take new partners, and we’re able to drive jobs so that they can start to get practice in our ecosystem. But we’re also able to maximize value for our established long-term partners by ensuring that they continue to have a large portion of the business as well. All of this isn’t happening through linear logic. It’s happening through modeling different behaviors and different data points, so that the system can automatically do all of this. And the net result that we drive for is a high-quality and on-time shipment, but ultimately customer and partner satisfaction. And so that’s really … I think those are the two examples that I would call out.

Dan McCarthy: [00:17:45] I recall something you say in your report that mentions AI enabling ideation, design, and production to happen in a single system rather than a sequence. You’re talking about, basically, Xometry-specific ordering system. Then we’re talking about AI changing the parameters of what what’s out there in manufacturing. They can take the modeling all the way through the production and the AI model, all of that, when they’re looking for a partner. Am I understanding that correctly? You’re talking about a single system that does all that?

Vaidy Raghavan: [00:18:21] Yes. And I can actually expand a little bit more, because what we are starting to see is this space is rapidly evolving and we’re entering fairly exciting areas when it comes to AI application, where it’s collapsing the sort of linear approach of idea, design, and production, which all were happening in a traditional site. They were happening in three distinct systems. And now we’re starting to blur the lines. And I’ll sort of interlace two different things here: one on Xometry, and also what we’re seeing in the marketplace as a whole. So I’ll start with Xometry. Xometry has always offered multiple digital channels for customers to purchase custom parts. So you can go to Xometry.com and upload a part. Second, we have plugins embedded in CAD software so you can be a designer. You design the CAD software and there’s a plugin that you can click and you can immediately buy from there. And the third is we have punch out integrations in enterprise procurement systems so that we’re able to come to the enterprise directly. Each of these channels feed into the same central Xometry back end, which is about orchestrating where the job needs to go, which is the best partner for it. So that’s one aspect, which is more Xometry-centric.

Vaidy Raghavan: [00:19:33] The second is, what we are starting to see is an evolution in the design process itself. So we’re starting to see designers and engineers begin to use generative AI and LLM to tweak and even generate new part designs from natural language prompts. So we’re seeing a good portion of startups come into the space and also establish CAD software providers that are offering these gen AI solutions to be able to quickly use just prompts to tweak or make new parts. Now, as this is starting to evolve, we’re absolutely tracking it and our role is going to continue to be: how can we maintain a digital walkway where you can have a generative design tool straight into quoting, sourcing, and production as a single click? That’s why our partnership with CAD plugins continues to be strong. We continue to engage with procurement platforms and integration players so that we are able to be there as each of these advancements occur. Now, that’s one. I’d love to also share with you another area where we’re seeing AI and intelligence make a huge dent, which is designed for manufacturing/DFM automation. I’ll talk a little bit about what we’re doing in Xometry.

Vaidy Raghavan: [00:20:48] In Xometry, over roughly the decade that we’ve been around, we have roughly 8 million parts. And the speed at which we are growing as a business, we tend to double the part count, unique part count, every two years, which means we have a ton of data. And so our enormous data set means that we can take the part information and extract geometry, quality issues that I’ve run into for that particular geometry and the material, the risks, analyzing for export control. And so we’re able to use all of that to automate several DFM checks. So a part gets uploaded the moment it’s quoted. We used to do a higher amount of manual checks. Now it’s automated checks where we’re able to check for all of this before it goes into production. We’re able to flag, we’re able to communicate with our customers and our internal DFM experts to be able to step in. And that’s another place where we are seeing, AI start to play a huge role. So an integrated system looking ahead is just going to bring all of this more and more to each other, which means you can design using AI and refine it. Then you use Xometry to instantly quote it, validate manufacturability, routes, the right supplier to orchestrate production, and then you have all of this happening in one connected ecosystem.

Dan McCarthy: [00:22:06] It’s an interesting idea to use an LLM to leverage a design. I’m curious —— maybe I’m speaking out of school here —— but I’m curious, how do you train an architecture like Xometry for your … you said 8 million parts? I mean, obviously that’s a data set. There must be other data sets. I’m just curious how you trained it to do what it does. It seems like a Herculean task, even for, you know, the age of AI.

Vaidy Raghavan: [00:22:34] Yeah. So I think there’s two elements. I have one clarification before we proceed: are you referring to the LLM portion, or are we double-clicking into how we are building data science models?

Dan McCarthy: [00:22:46] How you’re building for when someone comes to Xometry and uses your system to find partners. Just kind of curious how you would train a system like that. And you also talked about the design for manufacturing component, which I thought was interesting.

Vaidy Raghavan: [00:22:58] So maybe let me take one of those. And that should sort of give a general sense of how we do this. I’ll give the design for manufacturability checks. So we have over several years now, we have millions of parts. And for every part we have the drawing and we have all of the annotations that are available inside that drawing, and all of the remarks that the DFM has written in our database. So this becomes our starting data set. We’re able to take this data set and create a data science model, which says, “if this, then do a proximity analysis based on this data set of drawings and the annotations that people have written.” And so it starts to draw patterns of what are potential risk drivers. And when we identify a new risk, what we tend to do is we go back to the older files that we already have, and we will create a logic where we’re able to extract that information that previously we had not. We would do a one time pass to extract additional information so we can create an expanded data set and add one more layer of check in the DFM. So, what we end up doing is, essentially: we have the data, now you have to go create these models which say, “if this, if this, or if this, then it likely means this.” And that equation is something that you can train using data science and machine learning.

Dan McCarthy: [00:24:27] So is there a human in the loop at any point there? Where they sort of do a double-check on that process?

Vaidy Raghavan: [00:24:34] Yeah, that’s a great question. So I’ll give you a concrete example. We never introduce data science models directly into production. I’ll give you the latest and greatest of what we’re doing. So we’ve talked about, you know, even in our press release, we’ve talked about improvements we’re doing in DFM automation, and we’ve just recently launched injection molding. Let me talk about injection molding as a contrast example. This is just a few weeks ago that we rolled out injection molding instant quote. Now, we’ve deliberately launched the instant quote as a quote estimate. It’s not a direct purchase. It’s a quote estimate. In any other process that we’ve already had, you check out, which means you’re done. You pay for it and you’re done. Whereas with injection molding, what we have done is, the customer gets an estimate and they can confirm if they would like to proceed with verifying that estimate. So we don’t swipe the credit card at that time. What it does is, we have a high-priority queue of DFM engineers and manual quotas and there will be a human in the loop. So we’re introducing that as the interim step so that we can verify. Now their job in this case is not to run through and do the whole thing, but what they’re going to do is, they’re going to take the estimate and verify if the estimate and the data that it considered is valid and what it came up with was good. It’s a sanity check.

Dan McCarthy: [00:25:54] Yeah.

Vaidy Raghavan: [00:25:55] And so we plan to run this as the shadow mode or the initial beta test mode for the next period of time. And once we feel like we’ve achieved the right level of accuracy, we’re going to go back and we’re going to enable the customers to be able to do full checkout, which means you’re not just getting an instant estimate, you’re able to do instant estimate and checkout all in one shot.

Dan McCarthy: [00:26:16] Very cool.

Aaron Hand: [00:26:17] Okay. Something I’m curious about that I’ve thought about a couple times as you were answering these questions, is the idea that you have the ability, with all of the data that you have, to maybe make changes in the product itself to better adapt to circumstances. And I wonder how viable that is, whether it’s a change in material, for example, is one key that I’m thinking about. Do you have customers that would make that kind of leap to say, “the supply of this material isn’t something I can handle right now, but if we change this material or would that just mess with production too much”?

Vaidy Raghavan: [00:26:59] Yeah. So I think what we’re going into now is how easy is it to take a data science model that is focused and optimized for a set of parameters and add a few more parameters. So this is one of the problems that we constantly are trying to address so that the system … I’ll give you very concrete examples today. If you go to Xometry.com and you try to upload a part, you will see a list of manufacturing processes and a list of materials for each of these manufacturing processes. Now the system has been fine-tuned for a subset of materials and for other materials that we don’t see as high-traffic materials. We didn’t prioritize them in the first pass. Now, what we are doing is, we’re starting to take the second and the third tranche of material selection that the customers have started to ask for, and we are going back to the original data models to extend the initial model to be able to predict for these materials. Now the approach doesn’t vary too much, which is you essentially have a human do a set of these and that becomes your training data. And then you go back to your original data science model, the equation that you have developed, the mathematical model that you’ve developed, and you tweak it so that you can now do the second, the additional tranche of materials as well.

Aaron Hand: [00:28:24] Okay. It just seems like there are so many possibilities that this might open up in terms of how manufacturers might adapt, whether that’s to the supply chain or to their manufacturing operations. I know in your report you talk about the fact that nearly half of manufacturers have already seen significant ROI from their AI projects. And you have gone through a lot —— and I really do appreciate all these examples. Are there particular AI implementations you’re seeing? That would deliver, you know, the quickest value?

Vaidy Raghavan: [00:29:04] Yeah. I think that’s one of the most encouraging findings from our 2026 manufacturing outlook, where we’re seeing nearly half of manufacturers are already seeing ROI from their AI investments. And the quickest returns tend to come from targeted applications rather than full, massive overhauls. I’ll probably share three that I’ve seen recur over and over again. One is quoting, where you can take the historical knowledge and the data that you have gathered from your manual quoters and turn it into something that’s an instant quote system. For suppliers, this doesn’t just save time, it accelerates their ability. And for us as well, it accelerates our ability to do order conversion. Every customer wants to come to your site and they want to be able to buy right away. Their intent is high, which means you have a moment to engage with the customer —— and the last thing you want to do is come back to me later. So quoting is absolutely one of the places where we are seeing this, both on marketplaces such as ours, but also with the suppliers. And it’s freeing up engineering capacity as a result for higher value work. The second I would say is sourcing. We, of course, do this. AI is helping companies think beyond just cost. Our models, for example, don’t just try to optimize for cost. It is a factor, but what we look for is supplier performance, lead time, reliability, and geography. And we identify risk signals from within the network. So the intelligence allows sourcing teams to make smarter trade-offs between cost, speed, and resilience. The third one __ and I think I mentioned this already, but it’s worth hitting again —— is quality and manufacturability. AI models can automatically flag Xometry risks. As soon as you’re able to sort of train your models, you’ll start to see, you know, which Xometry and tolerance combinations are going to be higher risk, and how you can prevent costly rework and scrap. And those are all bottom-line impacts as a result. So, I think AI doesn’t just cut costs. It definitely improves decision quality.

Aaron Hand: [00:31:12] Okay. And, you know, you’ve really hit on this with supply chain, about not just finding the lowest cost. So, can you give us more examples about how they can really manage those sourcing risks on a daily basis?

Vaidy Raghavan: [00:31:34] Yeah. So I think AI lets sourcing teams, as I mentioned, move from just a price-driven model to a more intelligence-driven model —— which is, instead of just going after the lowest quote, you look for, like in our case, what we look for is a combination of factors. Every time we are matching a particular job, we’re not looking for “what’s the maximum profit.” What we are looking for is the combination. Is the partner reliable? Have they handled this particular complexity? So every time we get a part we have a geometric assessment of it. We do a mathematical compute of what is the geometry? Which means how much material? What is the bounded volume and the material volume? How much complexity will be there in the manufacturing? And so we are able to grade it. That grading is then matched with, what is the kind of jobs that we saw every supplier engage with, take on, and be able to deliver? And so we’re able to use those things when it comes time to matching a new job. So we’re looking for cost reliability, lead time, and resilience. And we can start flagging. We also are able to detect for a given process in a given region. Do we have sufficient supplier inventory or are we going to start running into risks, which can then affect on-time delivery? And those are the kinds of ways that we’re able to do an immediate decision, but also inform us that, hey, in the next quarter, we we should be doubling down on this region for this particular process.

Aaron Hand: [00:33:07] Okay. All right. Cool.

Dan McCarthy: [00:33:09] One of the areas we haven’t delved into yet is the talent gap. And that’s obviously always a topic for any manufacturing sector. But what about introducing AI? It’s a totally new platform, totally new technology for a lot of people using it. Is this a technology that —— as it applies to Xometry, but elsewhere as well —— is that a technology that requires a lot of upskilling for current workers, or is this something that could be easily adopted? Is it something that could be made easier? How are you addressing that?

Vaidy Raghavan: [00:33:48] Yeah, I think there’re a few different elements. I’ll start with this. One is, we’ve noticed that when we talk about barriers to adoption, people often assume the biggest hurdle ends up being the technology itself. But our data shows something slightly different. In the manufacturing outlook, we shared that about 44% of executives actually name “shortage of skilled workers” as the primary obstacle to scaling. So I think technology alone is never going to be enough. The challenge and the opportunity lies in people. The greatest barrier is going to be talent. And manufacturers are realizing this where they need to hire for a specific skill set, or reskill their existing employee base to be able to handle something like this. What we’re seeing is more manufacturers are recruiting technology leaders. First of all, leaders from Silicon Valley or digital industries, they’re recognizing that, number one, manufacturing leadership could look a little bit more like software leadership, which means agility, data driven, and an iterative approach. But at the same time, they’re also introducing new functions. In our case, for example, data science and machine learning have been a function that we have invested in from the get go for modeling different elements, whether it’s demand volatility or pricing or what have you.

Vaidy Raghavan: [00:35:13] And the other is, engineers that are more fluent in AI-driven design and machine operators who can work confidently. In our case, we’re seeing this much like anyone else, which is, as AI tools are becoming available, we’re having to upskill and reskill our workforce. Product managers now don’t need to wait on a designer to be able to give mockups. You have many design tools. The UI design tools are offering LLM prompt-based approaches to create UI mockups for your entire website, for example. Same thing for data analytics. For decades, competitive advantage in manufacturing was defined a lot more by capital expenditure, size of plant, sophistication of machinery. But I think as we look ahead, my sense is that it’ll also increasingly be about intellectual capital and the ability to deploy people and adapt to technology at the right speed. So, yeah, I think talent gap is a barrier. And so, investing in people, in being able to reskill, and bringing in talent from outside that already has the orientation —— the combination of the two can, you know, help really give a competitive advantage to the companies.

Dan McCarthy: [00:36:32] Yeah. Forgive me if I didn’t catch this in your answer, but do you see this as more affecting the AI supply side, where you have people doing the training, architecture of the neural networks, or is it also on —— or in addition to —— the user side, where you have prompt design, to develop the right prompts to do all the things that AI is capable of, learning the systems, and so on? Which side of that is inhibiting adoption of AI most?

Vaidy Raghavan: [00:37:03] I think there are two elements. One is the availability of tools that are immediately deployable. And there is a range of maturity. If you take a designer, what is the availability of LLM-based tools? That is very much in an infancy stage. Whereas if you go to analytics or sourcing, there’re far more capabilities that are now available to be able to use AI to do better decision-making in technology. We’re seeing the same thing, by the way. It’s not just in manufacturing. Our associated field, Xometry and technology, our workforce is a combination of folks who come from both a manufacturing  and software background. And we’re seeing both of them have to adapt and reskill and upskill.

Aaron Hand: [00:37:50] And certainly, how any company addresses a lot of these labor questions, I think, would feed a lot into how well they’re going to adapt to this whole change. So you talk about manufacturing leadership —— and certainly change management is a huge part of whether or not companies see success there —— so your report talks about the manufacturers who anticipate change or embrace digital transformation and can generally be more agile, those being the ones successful. So if you had to put together a list of tips for actually putting that into practice —— which of course I’m going to make you do now! So, what tips would you provide to manufacturers to adapt?

Vaidy Raghavan: [00:38:41] Yeah. I think, first off, I should caveat, I’m only about 8 or 9 months into the manufacturing world, although I’ve been in software, so I won’t be arrogant enough to presume too much, but I will share what I have gathered so far. I think the companies that thrive in 2026 and beyond are the ones that operate with transparency and agility as core principles. That’s what we’re doing in Xometry as well, not just using it as slogans. And let me elaborate a little bit. Transparency is about giving visibility to everyone in the company on data, on your data, your supply chain, your decision-making process. It’s about eliminating the blind spots that make organizations slow. What we have noticed is, every time data is available and is flowing freely across different decision makers, whether it’s in design, procurement, production, or with our partners, the teams spot problems early and adapt pretty quickly and avoid disruption. So that’s one. The second is about agility and incentivizing agility, meaning building organizations that learn fast, not necessarily about reacting to every change, but having the infrastructure and mindset and the culture to adapt continuously.

Vaidy Raghavan: [00:39:59] The best manufacturers and marketplaces are ones that treat digital transformation not as a one-time project, but as a slow muscle-building exercise for every quarter and every year. Experiment and learn and scale what works. So my input, I guess, would be this: if you want to future-proof the business, start simple, start small. Pick an area. Then be bold. For example, you could take quoting or you could take scheduling and make it more data driven and then expand from there. And finally, creating a culture of learning where information is constantly shared, where different tools that people can use without too much burden are made available, and bringing your teams along on that digital journey —— with the transparency of, “here’s why we’re doing it” —— and making AI part of your day-to-day operations …  The companies that make technology and transparency part of their DNA are more likely to be able to succeed. That’s how I see the change.

Dan McCarthy: [00:41:05] Very interesting. We’ve covered a lot of ground here. I think we’ve covered most of our questions, but is there anything we missed that you came here to tell us today? About either AI manufacturing or the direction of design procurement? Are there any additional topics that we should cover before we go?

Vaidy Raghavan: [00:41:28] I think you guys had fantastic questions, I appreciate it. I think the one thought that I might leave with is this: I think traditionally we have thought of automation as automating tasks. I think, looking ahead, automation is going to be about intelligence. We’ll start to see automation move up the stack where previously it was about robots on the floor. But now it’s increasingly about algorithms that make workflow decisions themselves. We’re doing it. We expect everyone’s doing this. And I think the factories that will lead are the ones where humans and AI can operate in partnership, where machines handle the repetitive and people focus on creative risk and innovative-oriented tasks. And as that happens, I think automation will start to shift from sort of this cost-saver mentality to a growth-enabler mentality. And it’ll let manufacturers do more with the same resources, and bring products to market faster, and respond more instantly to customer needs.

Dan McCarthy: [00:42:35] The shift to proactivity as opposed to reactivity.

Vaidy Raghavan: [00:42:38] Exactly.

Aaron Hand: [00:42:41] Great. Well, thank you. I really appreciate the insight. And, thanks a lot to our viewers for joining us on Manufacturing Matters. If anyone has any questions for Vaidy, please put them in the comments below, whether you’re watching on LinkedIn or YouTube, or wherever you might be watching. You can also see past episodes on our website, which is manufacturing-matters.com, or on your favorite podcast platform. So for now, do us a favor and hit like, subscribe, and keep tuning in. Thank you.

Vaidy Raghavan: [00:43:15] Thank you.