Jimmy Carroll: [00:00:02] Hello, everybody. My name is Jimmy Carroll. I’m the vice president of operations at Tech B2B Marketing. And welcome to this episode of the Manufacturing Matters podcast, where we discuss the trends and technologies reshaping the manufacturing industry today. Today I’ve got the pleasure of being joined by Naresh Ram at AAXIS, and Naresh, for those who don’t know, please tell us a little bit about your company and what you do there.

 

Naresh Ram: [00:00:24] Hello Jimmy. Thanks for having me. As far as AAXIS goes, AAXIS helps manufacturers with digital transformation. Our solutions optimize the digital ecosystem while strengthening customer relationships. For example, we are helping various customers digitize their product catalogs so their customers can buy with confidence. We’re making the purchasing and reordering process a lot smoother with e-commerce and surrounding technologies, like product information management, master data management, customer relationship management, other things. We do technologies and surrounding technologies around digital transformation. Personally, I play the role of chief science officer. It’s an interesting title. My job is the exploration of emerging technologies, with specific emphasis on using it to solve real-world problems for our customers. So I bring new tech to our clients essentially.

 

Jimmy Carroll: [00:01:18] Sure. And AI is obviously a primary technology and one that’s being talked about everywhere, whether it’s in the industrial manufacturing space or in the general public, with LLMs and generative AI being so ubiquitous now. And I want to ask a little bit more about that. So we get a little bit of an overview of what you guys do at AXXIS and the services you provide. But like I said, AI is a ubiquitous term today, especially in the industrial spaces. And when you first think of AI, what comes to mind and what does what does AI mean for AAXIS and how are you guys using it?

 

Naresh Ram: [00:01:57] So over the last five years, my views have changed. I think everybody’s views have changed so significantly. These days when I think of AI, I think of eliminating manual and repetitive tasks that no one wants to do or should be doing in the first place. Of course, there’s supply chain optimization with demand forecasting, predictive maintenance, machine vision for quality control. All those things exist, and those are still very important. Bread and butter so to speak. But on the shop floor, I think we’re talking about eliminating manual tasks. And let me give you a quick example. Converting emails to orders. Previously some of our customers would get this email. They’d look at it and then people would type it out. It just has a lot of errors and repetitive tasks — who wants to do that? And since 2021, we’ve been helping our customers extract information out of the email and place it into the ERP systems. And these days you can convert about 100 emails for two cents. So it’s very, very affordable, right? Look at maintenance and repair operations. When the technician walks to a machine, he gets all the data that the machine has to offer, in terms of what are the records, how we did. We even get the guides, the fixing guides and the manuals for it. They can query the document, even use AR to kind of look to see where to hit with the hammer, so the repair technician gets all this information in his palm, and he doesn’t have to spend too much time doing it. Even things like tariffs. Tariffs, they’re just changing. Every day there’s a new announcement. We don’t want people to go just looking on the site and figuring out if they need new suppliers, if their products are being affected. We want AI to automatically look at tariffs and see what the implications are for that specific industry. So these days when I think of AI, I think about automating manual and repetitive process and making things faster and smarter.

 

Jimmy Carroll: [00:03:59] Yeah, that’s really interesting. Because for our podcast, and obviously you can see the topics listed behind me. And a lot of the things we talk about involve robotics and like you mentioned machine vision, but it’s very similar. And in the manufacturing space, it’s almost like companies like yours and others are taking a holistic view of the manufacturing floor and saying, what are the jobs that people don’t want to do or are repetitive or dull or the “dull, dirty, dangerous” term that gets thrown out with robotics. But a lot of times when people think of AI and automation replacing these jobs, and I’m going to get to that in a second, but there are tons of different ways where these technologies are helping to fill those gaps, and what I was going to get to is that I think we need to dispel and perhaps to some extent we have dispelled the notion that robotics are taking people’s jobs. Part two of that, though, is now there’s sort of a new fear of AI, and you might not encounter it so much dealing with the customers that you do. But maybe in everyday life, how do you explain to people that, hey, AI is not to be feared. It’s here to help us. It’s here to better humanity. How do you have those discussions?

 

Naresh Ram: [00:05:21] I’ve been at a conference recently, and I’ve had discussions with several people, more specifically in the audience, asking these questions. I pull them out, and I’ve had this discussion to try to explain it. Whenever a new technology comes along, there’s often fear it will replace people, but history shows the opposite, right? Take Jevons paradox, for example. Jevons paradox is that when something becomes more efficient, we actually use it more, not less. So let me explain. This is from the economy. This is an economic principle from the steam ages. So when steam engines became more efficient, one would think that you would use less coal. But the exact opposite was true. How did that happen? There were more steam engines. There were more rails. The technology was more accessible, and we ended up using more and more coal. So in manufacturing, let me give you a manufacturing example: predictive maintenance. So if you don’t have predictive maintenance, you have big repair jobs and you need a lot of people. Now if you have predictive maintenance, the jobs are a lot smaller, require a lot less people. But that doesn’t mean that we need less people overall. We need more workers because there’s less downtime and there’s more production, and people can focus on higher-value tasks.

 

Jimmy Carroll: [00:06:40] Yeah, exactly.

 

Naresh Ram: [00:06:42] My advice to most people in the conference was upskilling. So AI will take on repetitive and manual tasks. And people should learn new skills and take on more interesting jobs, like robotics. So when robots came in, robots do most of the work, that didn’t replace people. We need more robot operators and technicians. So it’s not about replacing people; it’s about enabling them to do better, faster, and more interesting things. With the right training, AI can be a great tool for growth I think.

 

Jimmy Carroll: [00:07:12] Yeah. For sure. And now in the last four or five years, we’ve seen the ongoing labor shortage. Well, we’re not replacing jobs now. There are more open jobs now than there are unemployed people. We’re solving tasks that nobody wants to do, whether that’s summarizing emails that you’re talking about or picking oranges from a conveyor and putting them into a box. People are not wanting to do these jobs, at least in North America anymore. And now we have these technologies that can kind of fill that gap.

 

Naresh Ram: [00:07:46] Absolutely.

 

Jimmy Carroll: [00:07:47] One question I want to ask about AI, and I don’t think I’ve asked this one on the podcast before, but it seems like it would be a good one for you is: What do you say about the concerns that over-usage of AI has got sort of a negative impact on the environment in terms of the compute power and the data centers required. Have you had those conversations with people at conferences too?

 

Naresh Ram: [00:08:12] Yes I have. And I think what will happen. AI is an emerging technology, and with any emerging technology we’re going to have some ups and downs. Right now, the way we are doing compute is at a massive scale. We’re training all these things. Now, if you look at the major companies that are bringing these models to the market, you’ll see that now there’s a little bit of focus on smaller edge models. So the smaller edge models can live near the computer. And we don’t need all these wireless connections and internet to route to it. And they run quite efficiently. So when you look at it, when we look at tasks, we don’t just go take the biggest hammer and start chopping down whatever we see. We look at what is the best and most efficient model that can exactly solve the problem. I just talked about extracting orders from PDFs. If you look at that, we use the tiniest model we can, like GPT-4 is a good example. It’s very small. Or Haiku from Claude, right? We use those models for extracting named entity recognition. So we use it to extract the entity, and we use it to enter it into the ERP system. The next step for that would be: What is the impact of this on my production, on my inventory, on my suppliers? That is a higher-value task and requires higher reasoning. There we bring in the bigger models like either Claude, Anthropic Claude, or we bring in GPT-4.0 or 4.1 right now or 4.5. So when you look at it, I think as we go more and more into the future, we’ll start seeing more efficient models, and we should settle down on something. We should settle down on something where the training cost does not create a huge impact on the environment.

 

Jimmy Carroll: [00:10:00] Yeah. And you mentioned ChatGPT. And I do want to ask about  about generative AI in general in a minute, but I want to follow up there. One of the things that’s interesting is you hear about these concerns about using AI, but the thing is that you don’t always need the largest device, right? Like you’re talking about edge computing, like FPGAs or small boards that can solve these problems or some of the smaller, the Jetson devices from Nvidia. These are small edge devices that are incredibly powerful, that have really emerged in the last several years. You don’t always need the biggest GPU integrated into a large device, never mind the cloud. And I’ll get to the cloud too. But, yeah, I figured it’d be a good question to ask, and that was a good response.

 

Naresh Ram: [00:10:56] No worries.

 

Jimmy Carroll: [00:10:56] So before I get to that generative AI question I wanted to ask, just in general, from your perspective, both in terms of what you do and how you interact with your customers and in the general space, what are some trends that you’ve noticed in AI when it comes to improving processes and manufacturing? You mentioned predictive maintenance, but there’s also visual inspection, supply chain optimization, and so on. What are some interesting trends you’ve seen?

 

Naresh Ram: [00:11:25] Yeah. So I think that use cases like machine learning, visual inspection for quality control and predictive maintenance and forecasting, demand forecasting for supply chain optimization. They are like bread and butter. And they’re still very, very valid. The major trend I’m noticing is a shift from isolated applications to more connected, end-to-end solutions. So some of these generative AI models, they can bring reasoning into the mix. And then instead of the visual inspection just creating a report, we can use generative AI to look at the fault tolerance to see how often is this happening? Is it a separate case? It can even create a trouble ticket and send the technician to fix the problem. Also, increasingly this is being done in real time. I was at a developer conference, just developers, right? And they were just hacking around. And the guys were using generative AI to look at these drones. They were drone manufacturers, but they also did maintenance and repair. So when the drone came in, a machine scanned the drones, figured out which parts were broken, then started to look at if there were supplies, then started to place orders with suppliers. And I’ve seen it even call suppliers and negotiate with them for better pricing. So it can check stock. It can do everything. So nowadays I think the trend that I’m seeing more and more is real-time and next-best action. So it’s doing those things.

 

Jimmy Carroll: [00:12:57] Yeah, super-practical uses of AI there. And you touched on it a little bit, but I wanted to ask, beyond what you’ve talked about, what do you see as the role of generative AI in the factory floor? I’ve heard a lot of different things. Robotic programming is one of them. But what are some other ways that you see it being viable, because even a couple of years ago, maybe there were too many hallucinations that would emerge when people tried to use the models, but they’re just improving so rapidly. Even if you look at, it’s really kind of silly actually, but there was a video that that one of the generative AI websites out there a while ago did of Will Smith, the actor Will Smith, eating spaghetti, and you could tell it was fake. And then the most recent one that came out, it’s really pretty good. It’s scary how fast the technology has evolved. So anyway, now that it’s at that point, what are some ways that you’ve seen it add value?

 

Naresh Ram: [00:14:00] So I think, like you just said, generative AI has been truly, truly revolutionary in every space. And manufacturing and shop floor, it’s no different. They are hugely impacted. When we think about generative AI, especially in the developer conference I went to, we were told, just don’t think of what generative AI can do. It’s not an agent like a chatbot. Like most people when they think of generative AI, they think a ChatGPT interface where you talk, and it gives you advice on doing something. But why stop there? Think of what would a human do next, and see if the agent can actually do it. So when computer programmers, when we look at agents, we look at things that can impact the environment around them, read something from their environment and impact the environment. Let’s take an example. So there’s a huge purchase order that comes into an email. And we’ve done this for four or five years now. The agent reaches in and grabs that email, grabs that procurement PDF, parses it out, looks at the order, and then enters it into ERP system. Now what would the human do next? They would look at the order and say, Is this a big one? Oh, yes, it’s a big one. Now, do we have inventory to supply to this particular order? No, we do not. We should start scheduling. Before we start scheduling, do we have parts? An agent can look at the inventory of parts to see: Do we have supplies to produce all these things? No, we do not. So why don’t we place an order with the supplier, so it can reach out and place orders with suppliers? Now you must be thinking, some of the suppliers don’t even have websites, right? How do we do that? But no problem. I think one part of generative AI is meeting customers where they’re at, right? So these agents can make calls and start to place orders with suppliers, even get confirmation and expected delivery times. It can take these delivery times, work it into the schedule, and even start calling some employers’ workers to schedule them so that they can come in at that time. I’ve even seen agents start negotiating with suppliers to get the best price. So all of this is very possible and very affordable with today’s current technology. So when we talk about going beyond chatbots and looking at looking at generative AI and what is happening. I think of agents having the ability to interact and make changes in real life and being able to schedule manufacturing, call people, and to collect responses and to put it all together, I think is the next step.

 

Jimmy Carroll: [00:16:33] Yeah, that’s really interesting. I haven’t heard much about that. So that part of it is,  again, very practical. From my perspective, my background in this industry began in machine vision. So a lot of the discussions I had early on with AI involved deep learning and how it was overhyped and, initially anyway, being kind of pushed out there as something that’s going to solve everyone’s problem. But in the last several years – AI’s been around for a long time — but in the last several years, especially the last maybe two, three, four, there’s just kind of been an explosion in terms of the ways that AI can add value and beyond the ways that I traditionally thought about it. And that one’s just very fascinating to me, super-practical. Cloud computing — I wanted to ask about cloud infrastructure in general, and we talked earlier about people fearing AI and people fearing robots taking jobs. And we’re kind of over that hump, at least with the robots part. And hopefully we’re getting there with AI. But I know that for a while there was — and maybe we’re over the hump too, but I’m curious on your perspective — I know that there were a lot of companies initially that were hesitant to store their sensitive data in the cloud. Do you feel we’re over that? How do your customers go about handling large amounts of data like that?

 

Naresh Ram: [00:17:59] So I think that’s a really good point. Generally, when cloud infrastructure first emerged, there was a lot of hesitation with the companies. They didn’t want to store their data outside their walls. They were worried about security breaches and loss of control. And it’s their data, right? Who can blame them? But over time, I believe as cloud providers became more robust, they became more secure, more scalable, more reliable, more everything, the mindset I think started to shift. Today, with most of our clients, they don’t think twice about using cloud-based solutions. In fact, a majority of business applications that are now SaaS-based, right? And they run entirely in the cloud. Compute and storage, whether it’s CRM, HR, or even financial systems are running in the cloud. So the bulk of all implementations we do are in the cloud. And we do see a small part of the customers who want to keep their core system still kind of close. They have cages that they maintain, and they want control over it, but it’s becoming less and less and less. And an extension of that, I wanted to also say that an extension of that is I see this trend also going on with AI, similar trend. When it first came out, there was, like you mentioned before, there was hallucination, fear of hallucination, and there was a lot of skepticism around relying on external AI platforms. Nobody wanted to have their data end up inside of ChatGPT. You ask and you get your information back as an answer, right? But now privacy policies for the last two years, privacy policies with most of these commercial models like ChatGPT or GPT or Claude from Anthropic, they’ve all changed it. They no longer use information you provide for training. Also, there are several different options for private hosting, right? So you can get a self-hosted version of GPT with Azure Cloud, and you can even self-host in the cloud, hopefully not in a cage, but you can still self-host, things like Llama and for DeepSeek. DeepSeek can be self-hosted. And if you’re looking at extracting emails or something simple, Llama is a great option. The fixed-cost option. And then you can just process it, how much ever you want, and you don’t pay too much. And the other big thing is that these benefits are just too good to ignore, the benefits of cloud and of AI are just too good to ignore. So as organizations see the changes, they are now going from “Should we use it?” to “How should we use it better? What’s the cost-benefit analysis? How should we use it better?” And I think it’s just natural. And once the value and security are proven, adoption just becomes second nature. So to answer your question, I think larger organizations are definitely over the hump in terms of cloud and AI adoption. And I think smaller ones are getting there. And the competitive pressures, if it is not a very niche space, like you’re making a single precise screw for something and nobody else makes it in United States, okay maybe, but otherwise most people have to good options.

 

Jimmy Carroll: [00:21:09] I’m glad to hear that. I’ve got a question that’s sort of like a part two of one of the first questions I asked you, and with that first question or one of the first being: What does AI mean to you, what does it mean to your company? But another term that you used almost right away when you were describing what you guys do is digital transformation. And it’s a term that gets used a lot, and it makes sense in general, but it kind of means different things to different people, I’d say. So like at a high level, what’s it mean to you? And what do you see as the requisite steps that a business must take to successfully manage a digital transformation?

 

Naresh Ram: [00:21:52] So I think digital transformation for me is using the technology, any technology, to be more efficient and to stay competitive. Otherwise, why do it, right? What’s the point? We don’t want to just adopt a new tool just because it’s a new tool. It has to drive some value, and we have to rethink process, culture, and strategy by adopting this new tool. So digital transformation for me is not just about the technology, but it’s about the whole thing. And it has to be driven by value. So the way we do digital transformation is, I think the first thing is to have a vision. What does efficiency look like to you? What is the vision that you’re trying to achieve with this digital transformation? Do you want to handle five times more orders by digitizing your order process? Do you want to have your purchasers, buyers buy with more confidence because now all your catalogs are online and they can look at their specs without calling anybody? Do you want to cut down waste? Do you want to reduce the footprint, carbon footprint? We have to have this vision. So once you have that vision, I think you need to pick the tech that drives this vision and aligns with the goals of the organization. I would say that’s the second step. So the first step is to have a vision. The second step is to pick the tech. And this may not strictly be necessary, but in most cases, the third step would be to figure out how to get the data to feed the decisioning process for the tech that you have.

 

Naresh Ram: [00:23:27] Right. So every tech, if you’re thinking of digitizing your catalog, you need to make sure that all your catalog is there. So the next step is to process, collect, clean all this data and put it in a place that’s there. And generative AI is doing a great job with cleaning up data. So I recently talked to a customer who has paper from 60 years ago where they have specs of these chemicals that they first started selling. And he was afraid that they needed to get all these specs onto the e-commerce system, and we were using Grok to kind of scan these things, de-dupe the information, and put it on. So the third step was getting the data in, and then the last part: you can have the shiniest, brightest, best tool, but if nobody uses it, what’s good? So you have to create the right incentives. It’s the change management. It’s the upskilling. It’s the training of the users so they can realize the value and benefit of the tools you have. So it’s to manage the change and to provide the right incentives so that users can actually use the tool. I have some very interesting stories, but I don’t think my clients would want me to share how they incentivize, so I’m going to leave that out for now.

 

Jimmy Carroll: [00:24:42] I totally understand. Let’s keep them happy. I think you just wrote a whole blog on taking steps, the first steps to take for digital transformation. Right there with your answer, I think you got a whole blog ready to go. Let’s see what else? In manufacturing in general, what are you most excited about today? And part two of that is: Do you have any predictions for the next few years?

 

Naresh Ram: [00:25:06] So my job as a chief science officer is to predict what is to come and then to be prepared for it. But what excites me most is the things I don’t know. I want to see some exciting new things. So I was at this advanced manufacturing conference recently, and they showed this picture of this steam-powered industry. I’ve never seen a steam-powered plant. I saw the picture. They had these large spindles on the top, which were run by steam engines outside the plant. And they had these belts from the conveyors up top powering all the machines. That was how it ran. And now when electricity came in, most people thought, “Oh, yeah, we’re going to replace the steam engine outside with electricity, with a big electric generator outside, and we’d have the same spindles.” But no.  What do we do now? Today there are these tiny motors sitting in tiny places that are actually doing the power and doing these things.

 

Naresh Ram: [00:26:02] So it’s like you can’t even imagine what would new Industry 4.0 be, that industry that’s born in the age of AI. So I’m really excited to see that. However, I think in terms of predictions, what I think I’m going to bet my job on, I think factories are going to be more and more automated. I’m actually ready for auto-factories, factories that pretty much run by themselves, with very, very little manual intervention. And you might think, going back to your jobs analogy, before mechanization of agriculture, 70% of the labor force was on the farms. Now, today, that’s less than 2%. So we don’t have a 68% unemployment rate, right? The 68% of the people are now employed in factories that are making the tools that makes agriculture possible. And they’re doing other higher-value tasks. So as far as I’m concerned, I think I’m ready for people to move on to the next phase and for factories to be auto-factories.

 

Jimmy Carroll: [00:27:05] Yeah. Yeah. You think about the people that are in these jobs too. A lot of them will be aging out of the workforce at some point. Welding is one that they use often as an example. Most of the skilled welders in the U.S. are at or near retirement age. So these factories where most or all these applications can be automated will become necessary at some point, right?

 

Naresh Ram: [00:27:30] Yep. I was pretty horrible at welding. I did it during my engineering, and I just couldn’t get a straight line. I have a lot of respect for these guys who just weld it, and you see the thing and say, “Wow, this is like the Sistine Chapel of welding. It’s so nice. It’s a work of art sometimes.”

 

Jimmy Carroll: [00:27:47] Yeah, absolutely. Yeah, it’s a skill indeed. Naresh, anything else that we haven’t talked about today that you want to mention? I know that if folks want to learn more, they can go to AAXIS.io. But beyond that, if there’s anything else you want to mention, please do.

 

Naresh Ram: [00:28:03] No, I think we covered most of it, Jimmy. I think in general, we are excited. We want to help manufacturers realize the benefit that other industries are seeing. We want to help them transform digitally and be able to, in general, realize the great value that this transformative technology is bringing to the market by eliminating manual, repetitive tasks, get more precise, improve production, make everything affordable for people. So that’s that’s what I have to add.

 

Jimmy Carroll: [00:28:32] Well, I really appreciate it. I really appreciate you taking my time too. It was great to meet you and great to talk to you. Like I said, if anybody has more questions, they can visit AAXIS.io or check out our website at manufacturing-matters.com. We’d be happy to pass along any questions to Naresh and his team. Thanks, everybody, for joining and have a great day.

 

Naresh Ram: [00:28:51] Likewise, Jimmy, that was great talking to you.

 

Jimmy Carroll: [00:28:53] Thank you.