5 Things to Check Before Rolling Out Manufacturing AI

Ed Ballina (00:00)

Hello team, I am Ed Ballina.

 

Alvaro Cuba (00:02)

Hello guys, Alvaro Cuba here.

 

Ed Ballina (00:05)

And we’re bringing you our new Manufacturing Meet Up episode number three of season two. So welcome to Manufacturing Meet Up. This is the podcast of the show where we kick back. We talk about our downtime both physically and in the plants, share all kinds of really fun stories and get real about efficiency on the shop floor. Welcome.

 

Alvaro Cuba (00:41)

So guys, welcome to the Meet Up. And last episode, in episode two, we start talking about AI in manufacturing. And first, we talk about the huge investment that is going into the supply, billions of dollars in all companies, in all industries. But also, we talk about demand and…amazing that it’s predicted that next year, 96% adoption across manufacturing. More important, 50% is saying that their implementations are giving better than expected results. So it’s here. It’s not about “if” anymore, it’s about how to do it in a most efficient way. And this is the the episode for today. We will focus on that. We started last in episode two with Rick Wojcik, the senior manufacturing engineer manager for Fiberon. And he was sharing with us his experience implementing AI in predictive maintenance. And he was talking about how they…get ahead of the problems and what’s their experience in all the stages. But at the end, he reported that he reduced costs significantly and had a dramatic reduction of downtime. Now they are doing the rollout of the implementation ahead. And his last words of advice were, “trust the system”. So once you go in AI, just go for it. Trust it. So what we wanted to do in this episode is it’s not only about predictive maintenance. The first thing that comes normally, because it’s the first one that came into the plants, is predictive maintenance. But now it’s spread in all the other areas. And Ed and I will share something that we are seeing in our day-to-days in the plants, in AI, in different things. And then we’ll go to two very important things when you are thinking about AI. First is a checklist before you implement to make sure that everything is ready and aligned. And second, some tips for when you go all the way in. So with that,

Let’s start. Ed, you wanna share the first experience?

Ed Ballina (03:35)

Absolutely, no, great great tee up there and Again, big shout out to Rick. He did an amazing job. I thought that was one of our best podcasts and I just came back from giving a short presentation to Standard Industries and the title was “Trust the Process” so a little bit like what Rick was saying. But getting back to how expensive AI is becoming and where it’s reaching all parts of our of our business, AI for quality control, it just kind of quickly follows predictive maintenance and machine reliability. And we can see signs of it everywhere. There are companies that are using AI connected to vision systems to predict, for example, how a bottle is going to blow in a blow molder, whether it came out with the right specifications. And if not, it’ll go back and adjust the settings so that you always stay in spec or if it’s beyond the level, it’ll reject and cause a stoppage. But there is so much that’s available there for quality control. Things like being able to take advantage of the massive computing power and have, for example, the AI tell you, I see you didn’t change the ratio of this mix.

And by the way, I checked the original batch and it’s lighter than what you normally use. And I really would suggest that you change this ratio by a tenth of a point to get to your target. Right. How great is that? Because what happens today is you take a sample of your beverage, right. And you check it and you go, too much, too little. I tweak it. Well, you’ve already produced. You’ve either given product away or you’re getting really close to the quality. Right. So you’re like, Hey, what if some a system could tell you, based on what I see, right, I really think you should start up at a different point. Or here’s a real crazy one, every year when winter came, we used nitrogen to blow up our, to put pressure on our bottles, people will forget that it’s cold outside, it’s gonna take longer to start up the nitrogen system. Simple but effective.

So Alvaro, you’re gonna hit another very, very favorite point of mine which is safety.

 

Alvaro Cuba (05:55)

Yes. And a lot of improvement in cost and waste and all what you said. Safety. I want to share with you three things that I have seen two years ago. No, when I was working peer to peer was a challenge. Near misses a challenge. So much data. Everyone was volunteering for that. Not enough capacity to gather all the data. Now you go into your phone, you put this person, this thing, everyone in the plant can do it. All the data flows immediately. The trends, the hazard, anticipation, analysis, all predictive analytics comes to you. So what before was unmanageable, now it’s a reality. Another one is scanning, searching for threats.

Now you can, an alert tells you, hey, there is a problem in the floor. There is water or there is something that you can trip or the guards are not putting in the right place. No, or those kinds of things that generate accidents or if the safety gear is complete in every person that comes into the plant. So how amazing is that? And the last one is go with your phone. Check a person doing something that requires ergonomics, the software immediately tells you what is good ergonomics, what is medium, what is bad, and gives you the recommendation. You have to change heights to do it in a different way. So it protects people and the wellbeing of everyone in the plant.

 

Ed Ballina (07:54)

So following on the people side, we’re already starting to see quite a bit of use in training, right? And it’s gone from where training used to be these manuals and these books, and now it’s on computers and you see videos, and that’s great. That’s technological improvement. But with AI, imagine training that is so immersive that you gamify it. You turn it into a game. So if you’re going to be running a chemical plant, and we joked about this a few episodes back with simulations, right, where I blew up a chemical plant on a computer. Imagine, right, they’re teaching you how to operate this piece of equipment, and you have the ability now to go in there and actually start up the equipment or react to an upset or react to an issue. It’s so much more powerful than you trying to keep yourself awake, even the new training systems.

They’re still a little bit boring, right? And there are a big difference in terms of retention from being able to get in there and actually almost play with the equipment. We’ve all seen the potential of augmented reality eyewear, the Google glasses, the potential to enable people on the floor to look at a problem at the same time a technician in Italy that knows it inside and out, seeing the same thing, and then they would have the ability to actually draw on the screen and say, hey, you see that screw over there that’s kind of vibrating? Stop the machine, lock it out. That’s what I want you to adjust. And they can even watch the technician do this. I mean, the power of that is tremendous because the skills required to run our equipment these days is significantly greater and much more specific. You start talking about electronics and and all this, it’s a far cry from your shade tree mechanic that could take a wrench to something and fix it. Now it requires a bit more than that. So lots and lots of potential in the AI space of training to really catapult people’s knowledge and capability and confidence.

 

Alvaro Cuba (10:07)

And all of the above. Yes, all of the above. And one last example is customer service. How important for revenue, for cash, for customer satisfaction. Just two examples. One that is my favorite of all of them because I’ve been trying to work on this 10 years at least. I remember my pilots, with two top retailers and two top CPG companies trying to get instead of forecast, you all know, forecast is a nightmare, too many things. And once you have the forecast one minute later, it’s no good. Exactly that. So what about if you replace forecast by automatic replenishment?

So we were piloting 10 or 12 years ago, get the POS data every time you scan anything on the store and the perpetual inventory. If you have that, you can go to just-in-time and do it. Why we couldn’t do it? Because it was sheer amount of data and we couldn’t manage all that.

Well, now with artificial intelligence, I already seen pilots going for promotions that you can do just-in-time. So a lot coming in that way. And the second area for customer services, remember when you had to do or still most of us doing it, the production scheduling once a week. But one minute later, everything changes. No good. Now you can use artificial intelligence and it will keep reconfiguring. You plan for and you’re rescheduling to meet your deadlines and changing materials or lines or people. So those are the capabilities that AI is bringing in the real life today.

 

Ed Ballina (12:29)

You know, the one that you just mentioned was so interesting because, you know, demand type, right, or pull type production is ideal for all of us. And also to make some good decisions. You know, you may have a change in demand and all of a sudden the temperature drops 15 degrees and the AI says, hey, don’t worry about making it up because you’re not selling anything for the next couple of weeks. It’s going to snow. I just made that up. But you know, these things do happen, right?

 

Alvaro Cuba (12:56)

Can you believe it was Sam Walton’s dream when he created Walmart? His thought was, do just the marketing and the suppliers, every time I sell something, they bring it. Now we are starting to get to those times where this… Now it’s going to happen. So guys, we saw all these opportunities there are in our plants. AI is there.

 

Ed Ballina (13:14)

Now it’s gonna happen.

 

Alvaro Cuba (13:26)

people are starting to implement. Today we have more than 80% adoption and it’s going to be practically everyone next year. So we wanted to hit in a couple of things. Now it’s like I said, now it’s not if anymore, it’s you know, how. And in that “how” there is two parts. Before you go, no, and some, and a quick checklist and then while you are doing some tips. So on the first one and on the first checklist, just five things to think about when you are thinking, OK, I’m going to implement AI in the plant. The first, it seems obvious, but to all of us, it has always a challenge.

What is the problem to solve? You don’t want the technology and that trying to fit the problem into the technology is the other way around. So you need first to understand what the problem is and then bring the technology to solve it. For sure, it happened to you, spending hours and days trying to do the control settings of a machine and doing all the controls and fixes and all that. And then nothing happens. Then you realize that your problem is material variability. Right? So you were acting on the symptom and not on the problem. Or the process is not working, guys. Okay, change the process.

 

Ed Ballina (15:09)

Yeah.

 

Alvaro Cuba (15:22)

But what happened is you didn’t train the people. And then the process was not bad. This is just some examples. Or you are doing a lot and then you realize that data is wrong. So first, go deep. Why, why, why? Go down and identify what’s the root of the problem. Then think on the solution.

The second is about data.

Data is the raw material for AI. AI is the one who captures data, brings it all, and manage sheer amount of data to bring you analytics and things you can act on. So do you have the data? Number one, the data is accurate. By the way, parentheses, I have seen AI

that now you put it into your system and tells you which data is accurate and which data is not. So run that before thinking on running AI in the rest of the plant. So.

 

Ed Ballina (16:28)

Right. No, you may have a mechanic over there that gets upset when the AI flags him because he’s been pencil whipping PMs for the last four years. How can this be 90% PM completion and your line is running at 40? AI will sniff that out in a heartbeat.

 

Alvaro Cuba (16:56)

Exactly. So you have the data. Is it accurate? Is it real time? What we used to do, the paper coming, doing all the putting into the computer one week later, not helpful here. And it’s enough data or not. You don’t want to put AI if you don’t have enough amount of data to get accurate results, no? And finally on data, it is secure. Your IT people, I’m sure is all over this, no? Your data has to be secure. So that’s the second part. No, first is what is the problem? The second, what the real problem, the second is do I have the data? Otherwise start working on the data and they come. The third is, what you are planning to do, fits into the strategy of the plant and it fits into the strategy of the company. It is aligned with the goals and the priorities that you have for the year or no. It happened to me, I’m sure to Ed. No, you are working in a line and you are putting the line just to learn some months later that the line has been discontinued because the product is…marketing decided that no longer good or they want a different product. No.

Ed Ballina (18:31)

Or painful, painful, man. I’ve seen us invest. I saw them invest $10 million on a line while I was begging for capital to replace a roof. Right? Within six months, the line was dead because the product was an absolute flop. I’m like, man, for a little bit, I could have replaced X, Y, Z.

 

Alvaro Cuba (18:51)

And this is worse, and I’m sure you experienced this. You are automating a process completely to discover that the process is wrong. And now you only accelerated the wrong things.

 

Ed Ballina (19:08)

You’re making a lot more of bad stuff. Yeah, you know, when I talk to people about the concept of, you know, center lines, and, you know, a sweet spot, every operations has a sweet spot, right. And sometimes people try to run, they think I gotta run faster, go faster. And I’m like, No, you’re running past the sweet spot of this line. And what you just accomplished is making more crap faster. You didn’t help. Go back to the sweet spot and we’ll figure it out from there.

 

Alvaro Cuba (19:33)

Yeah and will work much better. This last one, you work, you are talking about quality and you are doing all the quality and making all the changes and all that. And then R &D comes and says, change your specs. Right?

Yeah, no, no, no. They don’t have it. this is important. You are going to invest time. This is going to solve really the problem. It’s going to go beyond your imagination, but it has to be aligned with the priorities and the strategies of the company. So you want to have all those discussions before you start going. The fourth is ability to monitor. No, I sometimes I went to a line. Yeah, we are fixing the line because the efficiency is not there. Okay. How you are measuring operation efficiency. Well. we measure parts of that. Well, if you are not measuring that, how you will know what is happening, no? Or if you cannot…make uptime or downtime. So, and it’s not only the idea of monitoring, but it’s the idea one step farther to take the learning. Right. No, I did something, it went wrong. How I, what I learned from this. Right. And how I applied that learning into the next. And then the fifth is the ability to act or to make adjustments. You can put all the data, do all perfect. And then an alert comes into a mobile and you don’t have the person or that alert goes into a mobile that nobody’s using. So you have to have the ability. Ideally, you automate everything and you use AI to help you to adjust. But if don’t, before going into AI, make sure that you have the ability to do something with what AI is going to give you. So those are five parts of a checklist before you go implementing.

 

Ed Ballina (22:23)

No, great great points You know you, you make me think of my short-lived career as a private pilot. I did actually solo an airplane. But the reason I bring that up is you know before you fly, you go through a checklist. You know you check your plane out statically, you then run it up to the flight line. You run up the engine, you check, you know, that you the magnetos and all this stuff, right?

You have that plane as well as good as you can possibly get it. You take off on the runway and once you get altitude, what is the first thing that you do? You trim it because it never flies exactly level and you know, exactly, so what you do? You trim. You have to use trim control. No different in a situation like this. You’re implementing major, major change. This is huge.

Would you buy a piece of equipment or a line for $10 million and then roll the dice that you’re going to install it right and you’re going to have a perfect startup? No. Treat this with the same level of importance because actually this can make an even bigger difference than investment in a new piece of equipment. It has the potential. And then to reiterate the watch that you had at the beginning, most people that the watches show are technically curious, right? And we like shiny new objects.

Okay, we’re technology geeks. We tend to be first adopters and all that and that makes us really really prone to become, to fall in love with the technology and then figure out how to get this cool technology to help you. As much as I have done that, my advice is to our, to Alvaro’s point, okay, what is the problem you’re trying to solve and then go out there and find for the right technology solution to that problem.

Okay, not the other way around. Don’t love the technology and find the use in your facility. You can do it, but it’s not very efficient. And the other thing is, buy technology that is as close to being perfectly suited to your operation and not something that you have to make work.

Alvaro Cuba (24:28)

Yeah. the temptation is bigger in this case, everyone is talking about it. So everyone wants to have it now, it’s the new toy. So go for this and not because it’s not a $10 million line, forget all these five steps. Okay. Ed.

 

Ed Ballina (24:38)

Exactly. Cool. So now you have launched this journey. You’re going down this river at a real fast pace. What are some things you ought to be thinking about? First of all, if you’re in a river, you’re going fast. You should have a life jacket. Getting back to the business, first and foremost, especially with AI, it is about people and process.

Going back to the $10 million investment you make in equipment. Yes, it’s about people, it’s about process, but a lot of it’s just about get the equipment right and get it to run right. With this, this is about making sure that you think about your people, how they’re going to accept this, and then the technology has to complement both the skill set of your customer base, right, as well as their willingness to use it. So if…you know, we’ve said a hundred times, that every one of us has this mini computer in our pockets, right? The closer you get the interface to that, right? The less esoteric it is, the greater chance that people are going to take it on and adopt it. And then culture is a big deal. The culture needs to change because for a long time we’ve rewarded that mechanic that is the knight in shining armor that comes to your line and 45 minutes later, he or she has his line up and running and, hey, Josephine did an amazing job, right? And that’s good. We need to continue that. But we need to start also rewarding the people that use technology to identify an issue before it happens, gets the part shipped in, like you mentioned earlier, right? Before you need it and you plan the downtime and you’re up and running. And you never took that unscheduled downtime. So, so important to start rewarding that predictive, forward thinking approach to maintenance instead of, hey, I can fix it when it breaks really, really fast.

 

Alvaro Cuba (26:45)

And one quick thing there is if AI is not improving the life of the person, something is not right. Because it has its in safety or in ergonomics or in calm and more time to think or whatever. But it has to impact the people to be sustainable, as Ed was saying.

 

Ed Ballina (27:13)

Absolutely. You guys have used, I’ve heard me use this term before, WIIFM. What’s in it for me, right? And you gotta think about that mechanic and that mechanic wants to do a good job, right? They want to be part of the team and be altruistic. But at the end of the day, what is it that you can give them? Let me tell you a few things you can give people. If your line, first of all, nobody likes to run a crappy line. I think I say that at least once every podcast. I’m at least constant. I am consistent if nothing else. You have the ability to give people control over their lives again, okay? So that when you come in the door, you don’t have to worry about when you’re going to be able to go home to see your daughter’s soccer game because unexpected breakdown happened and now you’re stuck in the plant for 12 hours fixing a problem. Or every day you come in, it’s like, what is going to break today?

Right? What sharp part is going to be there, you know, ready to cut me? So, you know, there’s a lot that needs to be considered when we’re in this journey because you’re changing people’s lives. You’re changing their work, what they do every day, and it’s really important. The last thing that I’ll mention here, understand you have a lot of stakeholders, right? And you need support from across all business functions. So you need to have, when they teach you sales training, they’ll tell you, there’s the technical buyer, there’s the economic buyer, there’s a this buyer and all that. Very, very similar to here, right? Yes, your customers, first and foremost, is a shop floor. It’s that mechanic, it’s that operator, right? But don’t forget, there’s a finance person looking at this and saying, yo, you promised me a return, where’s my money? And then the supply chain planning people are like, hey, you said you’re gonna be more predictable, so I should drop my inventory. And anyway, you get the idea.

It’s holistic and you need to approach it that way. So Alvaro.

 

Alvaro Cuba (29:13)

Yeah, it’s critical areas, it’s recognition, no? Make it fun, accountability, it’s critical, share best practices, no? So in that order, people first, process. Once you have those two aligned and well, then automate just to make it speed and better. Yeah, you have there some ideas about something to think your checklist before implementing some things that Ed and I are seeing right now when we visit plants, when we talk to customers about we are all learning. This is learning process in AI.

And that’s the conclusion. Anything else, Ed?

You were going to say something!

 

Ed Ballina (30:19)

I just had to throw this one in, right? Please. Just to give you a sense of like, how wide ranging and impactful this technology is, right? I just read an article about this woman who had a stroke 20 years ago that through the use of an implant, an AI is now able to speak again. They’re literally

 

I mean, loosely said speak again, right? But they’re essentially can translate her thoughts forming a sentence to an avatar on the screen that is using her own voice that was captured from years ago and her facial expressions to work an avatar where she can actually speak. I mean, talk about transcending, you know, capabilities and providing people an amazing improvement in their quality of life. This is what’s coming, folks. And Alvaro. It ain’t coming, it’s here.

 

Alvaro Cuba (31:13)

It’s here and a lot to take advantage of. So one step at a time, let’s go. So with that, friends, we come to the wrap up of today’s episode. We hope that you enjoy it. Please share with us what you think, what you want to see in the next episode. We love to hear from you.

And if you are seeing this on YouTube, please like us, give us a review on iTunes. And more important, if you like it, talk to your pals, invite them, come to the meetup with them the next time and we’ll have a coffee or a drink together.

 

Ed Ballina (32:08)

Fantastic. So this has been another fun episode. If you want to keep the conversation going, you can email us at mmu.augury.com. You can also find us on the Endpoint. It’s a free online community for manufacturing pros like you and I at endpoint.augury.com. Now look, folks, we’ve been telling you this for a year and change. We’re easy to get a hold of. So come on, fire up. We want more questions.

And if you get some really good ones, we’ll feature you in an episode, because we are bringing more guests. That’s going to be one big change you’re going to see for this year. We’re going to have a lot more guest speakers, because truly, they’re so much more interesting than Alvaro and I. We do the best we can, but…

 

Alvaro Cuba (32:48)

Yeah. A lot of fun. time life episodes. guys, stay tuned. Take care. Bye.

 

Ed Ballina (32:52)
See you.