Jeremiah Woodford with Verusen AI

SUMMARY KEYWORDS

00:00

Hey, industrial professionals, before we get into the show, I want you to be aware of a free ebook that we have out on industrial talk. It takes and expands upon the five. These are common five elements that make companies successful. This is it. One, they educate. Two, they collaborate. Three, they innovate. Four, they invest in the culture of the organization. And then five are able to, in an effective way, communicate to the masses of what they do ineffective way. Those are the five elements. That's what the book expands upon. Go out to industrial talk, download it, and while you're at it, get the free workbook too as well. Now on to the show.

00:58

Welcome to the industrial talk podcast with Scott. MacKenzie. Scott is a passionate industry professional dedicated to transferring cutting edge industry focused innovations and trends while highlighting the men and women who keep the world moving. So put on your hard hat, grab your work boots, and let's go all

01:15

02:24

name? Verusen. Verusen. It's Latin for truth.

02:31

Whatever it takes. This is true. It's over my shoulder. They're in the same neighborhood on this floor. Your people are very nice. Just nice folks

02:41

there and a great team. We have a great team. Yeah, no

02:45

complaints. I wouldn't complain. I would have sold them. Why would I do that? Absolutely all right, for the listeners out there, Jeremiah, give us a little background on who you are.

02:56

Yeah, been in industrial, heavy industrial maintenance, and been in and around it for 20 plus years. Born and raised in Southeast Texas, still reside in Texas. We worked in the oil field, worked out I was a roughneck for a number of years, and then was an outdoor machinist in refineries and chemical plants. And finally went and got my computer science degree and got in the software and tried to run away from the plants as soon as possible, but ended up getting into maintenance software and got back into the plants. So yeah, been in and around it for 20 plus years, traveled the world, worked with some of the biggest industrial companies in the world. And, you know, I don't know, I'd like to say best practices in some of these things, but good and bad?

03:41

Yeah, we've been having this conversation forever. It just seems like, you know, but, but I think the technology as a whole is as able to keep up, and now we're, we're at a point where we can truly leverage these solutions in a way that makes sense, yeah? I mean, beforehand is like, I don't know,

03:59

yeah, you know, coming in and around been mainly around EAM software to 20 plus years. And, you know, helping people buy and implement those software systems and listen to maintenance people complain about how bad they are.

04:17

And yes, yeah.

04:19

And it is seeing, you know, work arounds and trying to figure out how to make the thing work, and ultimately creating a ton of data. And I think what the perception has been is bad data, ultimately. But yeah, so now, you know, I'm at an actual AI company, and I say actual is because it's such a buzz word that everybody uses today. It is and, you know, and I think it's overused, but what we are, what we do is, it's a large language model. Okay, so it was a joint venture with Georgia Tech Institute of Technology, and they took Stanford's large language, my open source large language model, and started training it on MRO data. And what's really cool, I found. Found these guys couple years ago when I was at an unnamed three letter EA, own provider and gotcha and really wanted them to acquire this company for their technology. And then the more I learned about it, and more excited I got about it and said, I got to come work for you guys. Very cool. Yeah, that's how I got started with them, and it's some really cool tech

05:23

for that listener, just from a level set perspective, give us a little What does MRO stand

05:27

for? Maintenance, Repair, some companies, depending on what's the industry and operations or overhaul

05:33

Good deal. Want to make sure that everybody's on the same boat. All right. Take us through the company. What? From a basic perspective, what are you trying to accomplish?

05:43

Yeah, so it's really straightforward. We're trying to help you determine if you're if you're a big industrial company that's buying and stocking maintenance spare parts. We're plugging into your ERP and your EAM, and helping you determine the stocking level of those parts. There's a lot more to it, but at the base, that's what we're doing. So we're we're leveraging all that data that's been created for years and years and years, and most people think a lot of it's bad. And the reality is we can make sense of it, and that's where this AI large language model is really powerful. Just give you an example. So we're pulling your master data catalog. So the master data catalog is your internal parts list that your maintenance teams need for to keep you know, the bullet, the boiler, the chiller, or whatever. You know your whatever industry you're in. We operate a lot in industrial manufacturing, but we have oil and gas and mining and and so forth. They all have same problems, and a lot of the crux of those problems are, is the poor naming conventions in those catalogs. I'm a maintenance guy, and I you rely on me to tell you what I need to keep this machine up and running and and, you know, I'll do my best to describe that bearing and that the size of that bearing, or that pneumatic cylinder and the stroke and bore of that pneumatic cylinder, but everyone's going to describe it slightly different. Yeah. Slightly different. They're going to abbreviate it slightly different, so that everyone views this is like an insolvable problem. And this is where this large language model is really cool, in the sense that it knows that air and pneumatic is the same thing. It knows that the bore stroke on the nematic cylinder, point two, five or one, dash four is a quarter in those that's the same measurement. It's been trained on it, and it's constantly learning. So every time we put more data into the system and users are interacting with it, it's learning. And to be clear, though, this is not some, you know, nefarious AI that's gonna, you know, break loose one day. It's still a dumb it's still a dumb model in the sense it's very pointed, purpose built to solve a very unique problem.

07:48

Yeah, every time when, when I was deploying systems, it was always, how how do you scrub the data? How do you normalize that data? How do you make it relevant? Because there's a lot of lot of bad data in there. Shall we say, are you? Are you indicating that the solution can do that and scrub it and go back and do the heavy nobody wants to clean data. Yeah. Nobody, you don't. And nobody,

08:17

time and time again. You know, being in the am software world. You're ripping and replacing some other older legacy system, and you got to go down this data cleansing process, and then you talk about, we've got a, you know, how much, how many years of history are we going to migrate over? If we migrate over, we we don't, we don't. We're not a data cleansing house. We normalize that data on an effort to help you determine that this bearing at this plant hasn't moved in five years, but you've got the same bearing at this other plant that's moving fairly you being used on work with, and you should transfer it to that other plant or optimize at a network level. So our goal is, you know, we will identify duplicates and merge those duplicates in our system, all in an effort just from so you've got eight plants in our system. You're buying the same parts, but they're all named, eight different name numbers, different part numbers. But we we can our large language model identifies these are the same parts, and if you agree there are the same parts, then we can help you start looking at your suppliers. Is a supplier, you know, who's the best supplier for this part out of those eight plants, this supplier is giving you the best price. They're delivering consistently on their lead times. And this part is the most reliable from a usage perspective.

09:39

That doesn't happen overnight. I mean, if I came to you and I said, Yes, I am a manufacturer. I'm Acme, a manufacturer, and I and I hear what you're doing, and I think I'm excited, and I want to get better handle on at all, whatever it might be I have, I have problems. And, I mean, where do you even begin? Give us your day. Right? So I even have problems with that.

10:03

So we hear it time and time again every time I'm training a new team right now. And is, you know, what's really unique about this? There's not, I don't really have competent I compete. My biggest competition is do nothing. Most people look at this problem as unsolvable,

10:21

or, look, I'm I'm there. That's a tough one.

10:25

12:27

How do you interact with E mate? What? What? What does that relationship look like?

12:33

Yeah, so we're a strategic partner. Of you made some you know, we, we, we have an integration, standard integration for x4 and x5 and essentially what we're doing is we pull all your master data catalog, your work order history, your procurement history, any of the purchase data. And to be clear, too, if you have most of your email customers, gonna have an ERP that you may have some integration with. Will in certain cases, depending on how the integration is structured, we'll pull from the ERP as well. So, you know, we have customers that are in SAP and Maximo, you know, we're agnostic, but the Fluke partnership is strategic, in that sense that we have a tighter integration. And so ultimately, the goal there is, is that, as ours, we pull all that data and then we start making stocking recommendations. So if you log into our software, the AI is going to say, here's a bearing. You've got 10 on the shelf. Your min max is 10 and 11. But you only, you only use it once a year, and your lead time is two days. So it may say the new recommendation is three and four. And if you agree, then it's going to go back into email and change the min max, but it's always on listening. So in that example, we're also tying that bearing to the asset. So if that bearing is used on a really critical pump, most critical pump in operations today, and if that you know so that is maybe, if you're depending on what your criticality metrics for your asset is, maybe it's ABC, the Baron is going to adopt the asset criticality. So it's really important for us to know how important, how critical that spare part is. So that's where we're bridging the gap between procurement, supply chain and maintenance. And from there, what we're going to do is procurement knows that. They don't know how important that bearing is. Maintenance knows that, but procurement knows I can get this bearing from Granger McMaster center three times super short, so it's not critical to source it. So from that perspective, what we we can flush out is, okay, this is really critical to operations and maintenance, but it's not hard to source it. In that example, we're going to downgrade the criticality, but we're constantly monitoring it, so if something changes like now, making this up now, you got to go to Germany to get this bearing from the OEM, and your lead time jumps to six months. Then it's going to pick it up immediately and change the stocking strategy and the criticality.

14:59

You. Are you asking me to be in another system? Or can, can, can I use email? Or what? How do I do from a user perspective? What?

15:09

Yeah, it is another user interface to the system, to and typically, the users of the system are going to be your material managers or sourcing procurement people. And it, and it's usually in the beginning. In the first six months of our software going live, they're in the system about 20 minutes a day. And maintenance people will get notifications, because typically, like critical spare parts. You don't want your sourcing, procurement people to be able to prove they're you know, they're going to immediately accept all decreases in recommendations. So you want, so we have workflow in the system that will know, you know, anything that's a critical A or B in the system have to go to these maintenance people for approval. But it's what's really cool about our system and the AI, we're using explainability, and it's using natural language processing to explain why it's making this recommendation, meaning, like, you know, so maintenance guy get a notification that, hey, they wanted, they want to decrease this critical a part. The immediate reaction is like, no, I need this part. But if you read the explanation and you see this thing, you've got 10 on the shelf. You haven't used one in two years and and even if you use one, you've got nine more so, and the recommendation is maybe five and six. I mean, making this up, right, right? And so you're not really putting anything at risk. So it's really, it's, it's like it from when you see it in that perspective, it's, it's pretty basic understanding. And then so if they agree to it, then immediately that goes through a workflow approval process and pushes back.

16:44

Do I have a dashboard? Hate to say it, but do I have a dashboard with these recommendations? And if I log in and it's like, hey, we realize that we need to increase this inventory, decrease this inventory, we've got a part over here. We you're trying to just optimize that whole supply chain that's right, in such a way that I'm getting real, real, accurate bottom line value, and not not incurring costs where I have to incur costs, and incur costs where I have to and so on. And so you're optimizing.

17:14

That's right, absolutely. So we, you know, from my perspective, we talked a lot about decreases, but just as important, we identify the increases that are needed as well.

17:24

Let's say I'm i The your system identifies an increase, saying, for whatever reason, these bearings, you're going through these bearings a little bit faster than you should. Is there some logic or or analytics that say, Well, why the heck are we doing that now, why are, why are we going through more bearings? Why? You know, what is that? Why?

17:47

Yeah, so we have key data points that we're flushing out, and we're constantly improving this our data scientists, and we're 75 people globally, and half that team are data scientists, engineers that come from like Georgia Tech and others that are mechanical engineers that are looking at, how, how do we improve these data models? Yeah, to really help tie the reliability and uptime. Give an example. I did a project with a super major in the North Sea, and one of their their big initiatives were like, we want to optimize inventory. We don't really care about reducing inventory. It's the production platform. Is $700,000 an hour if it goes down and you're out in the middle of the North Sea, and if you don't have the right part when you need it, they have to helicopter or fast boat it out to the platform. It's a no brainer. You have their part, right? And so through this process of helping identify these are your critical spares. And on those platforms, you have very limited space. So one in this process, we helped them clean up so they get a bunch of stuff that was on the platform that shouldn't be on the platform, that doesn't move, and get it back to the shore based warehouse, and get the really critical spares on the platform, and then help and so in correlation. So they they were actually able to improve their liability and uptime from a number of measures. I'm not going to claim that we we were direct, but we correlated that, you know now we reduce their helicoptering and fast boating parts out. They for that the two year period that we're tracking it, there were none that were spent fast boated out or helicoptered out, and the splatter chart actually correlated directly with their reliability and uptime improvements

19:27

that they had. We talked about criticality, of course, of the asset and ABC, or 123, or whatever. The the approach is, is that also fed to your system. Because if I have the CMS or, you know, asset management software, whatever it might be, that's the system of record. Yeah,

19:52

absolutely. But we want that data Absolutely. So we absolutely want to see the asset criticality is vital. So. Way we've built our system, and the way the AI models are working is we're going to ingest the asset, the asset strip, the action, the equipment, name, the criticality of that asset, and then we're constantly monitoring if anything changes. You know, typically equipment criticality doesn't change. But you do? You do obsolete pieces of equipment, right? Oh yeah, you do. You work it out. You get the newest, latest, greatest compressor from Siemens, and that replaces it. So any of those parts associated with that asset, it's really important for us to have that purview. So we have customers today that are tracking as they obsolete older equipment. They want, they want to make sure those parts that were just used for that they want to identify the parts that were just used for that equipment, and do they need to transfer it to another plant that does still use it? And so that's part of our system. Is we have a networking model that looks at all the plants and shows All right, so you've got these parts that haven't moved in five years, this plants using those parts, you should transfer it before they buy more.

20:59

Yeah. And I would be looking at why, you know, why? Why are they consuming more of the part versus this particular location? Is it? Is it regional? Is it because of there's just so many other signs of variables you can just keep on going on it. One last sort of, put your put your future hat on, where you see this stuff going, where, what are you looking at?

21:26

Yeah, we want what our goal is, and what we're trying to build is what we call, it's a buzz word, but this is we have a large language model, but we're building AI agents to use that large language model, ultimately. So when you buy the software, there's we, our end goal is limited interaction with this thing. So it's management, management by exception. Yeah. So right now we make stocking recommendations. You agree disagree, if you disagree, it asks a series of questions, right? Training this model, and it's like, I would think of it as like a junior supply chain analyst. But as this thing gets better and better and more customers, and so right now, we've got, most of our customers are fortune 100 fortune 500 big industrial companies, and we're sitting on, I think it's around 60 million master data SKUs with our customers, and it's around 10 billion transactions tied to that total inventory sitting in our system is around $20 billion of inventory globally. So it's getting better and better. But where this is going is this, like I said, this occupied space that we have is, is to automate, yeah? So yeah, because,

22:40

yeah, I want it to magically. I don't want to be notified, because this is what's gonna happen. I get notified. I get notified again. I get notified again, and I'm gonna ignore it like I do my email. Absolutely.

22:53

What we wanna build is resilience and trust in the sense that in the first six months, you're just validating that this thing's working right. And then you let it do its job and it's gonna and then it notifies you when it has a problem. And the last notification,

23:12

yeah, that's that's cool, by the way. Verusen, Verusen, verason, Jeremiah, how do people get a hold of

23:22

you? You look me up on LinkedIn. Jeremiah, Woodford, barrison.com, is where you can find you got go to there. And if you guys highly recommend get a demo, really cool demo software,

23:35

because you live, you live near here, you know you have automated

23:39

vehicles around. I'll leave you with one last that we're in the future. I was in downtown Austin last week, and I was at a red light in the way mo self driving car

23:50

I've been there, and

23:51

so it was sitting beside me, and then the little robot that delivers food was crossing the street, and then the Amazon truck ran it over. I

24:04

24:30

You're listening to the industrial talk Podcast Network

24:39

Jeremiah Woodford, the company, or a Sen AI, incredible technology. It's all giddy. I love it. Man, how AI, how they're leading the way and being able to sort of impact. MRO, yes, that was a great conversation. Information at Xcelerate. Really appreciate Jeremiah being on the show and sharing his insights, wisdom and really solutions. It was, it was fantastic. All right, again, we have an e book. We have a workbook. It's all out in industrial talk. It is all focused on your success. We want you to succeed. We want you to educate, collaborate, innovate. We want you to have that vibrant culture that that really separates you. And then, of course, we want you to tell your story in an effective way. Go out there, find out more. All right, be bold. Be brave. Dare greatly hang out with Jeremiah. Change the world. We're gonna have another great conversation shortly. So stay tuned.