← The AI in Business Podcast
The AI in Business Podcast
Capturing Tribal Knowledge to Solve the Manufacturing Skills Gap - with Sebastian Dykas of Smith+Nephew
00:34:11
Loading summary
Transcript22 lines
- [00:14]
A
Welcome everyone to the Emerge AI in Business podcast. Today's guest is Sebastian Dijkas, Director of manufacturing engineering and maintenance at Smith and Nephew. Smith and Nephew is a leading portfolio medical technology company across orthopedics, sports medicine, ENT and advanced wound management. Sebastian joins us on today's episode to explore how manufacturers can strengthen operations as experienced workers retire and production demands rise. He shares how teams should capture expert techniques, standardize training and use real time process control to stabilize output. He also unpacks how machine connectivity and automated feedback loops reduce scrap, tighten control limits and give leaders clearer visibility. The focus is on practical steps any manufacturer can take to modernize workflows and build more resilient operations. Today's episode is sponsored by Poker. Just a quick note for our audience that the views expressed by Sebastian Duikas on today's program do not reflect that of Smith and Nephew or its leadership. For our solutions partners, position your brand alongside the Fortune 500 leaders defining the enterprise AI roadmap. For the opportunity to showcase your solution to the executives currently funding and scaling global initiatives, partner with Emerge. Secure your partnership@go.emerge.com partner that's go.emerj.com b a r t E R Now the conversation with Sebastian. Sebastian, welcome to the show.
- [01:40]
B
Thank you for having me. I'm excited about this conversation.
- [01:43]
A
There's a lot of talk about digitizing the factory and we've had some interesting conversations around digital twins and advanced automation. But we hear one thing across manufacturing and that's that there's a more fundamental problem, a basic problem still lingering where there seems to be a gap in knowledge. So the a lot of knowledge is still in, in the hands and in the minds and in the hearts of a very few experts on the manufacturing floor. And the question is, how do you get that into the workflow where in an industry where a lot of workers are either falling off, not interested in going into manufacturing anymore, we see a lot of job changes. How do we capture that? And so before we get into AI and where all of this is heading, I want to start with what you're really seeing across operations you've worked with. Where are the biggest points of friction around training, knowledge capture, or those paper based processes that still seem to shape
- [02:44]
B
frontline work, there's several issues that you brought up that is plaguing the medical device industry. In my opinion, one of them is operator challenges. As the older workforce retires out, we're losing the machinists, we're losing some of the experts who've been in the industry running certain processes for sometimes decades. And the challenge is, is how do you create processes that don't require that level of expertise and how do you transfer that knowledge in a way that someone can easily absorb. The other thing is, in certain areas of medical device, you also have a lot of craftsmanship. So there are machinists who are running CNC machines and making offsets to processes. You have hand finishers that are doing processes to finish the final form of an implant in some cases. And then you have other just tribal knowledge of how do you run the cleaning system, Some of the older equipment, when it glitches, what do you do? You know, that's from the maintenance perspective as well. So there is a challenge as the younger, newer workforce comes online. There's just not that depth of knowledge that we're seeing in the older workforce that's been seasoned and has been decades in the industry. And so what we have to do is figure out how to capture that knowledge and then how do we then take what is done by these people best practices and then standardize it? For example, in a company I work for, we were finishing, hand finishing, polishing certain medical device implants. And some of the older, more senior workforce was able to basically provide double the quantity and the shift. As someone who was only doing it for a short period of time, they developed their own best practices. They knew just because of quantity and time and hours on the equipment, they could, you know, produce almost no scrap, how they could produce higher throughput. But it's very difficult to put that and ingrain that into someone who's starting off. So those are challenges as well as, you know, the way we inspect equipment, a lot of it is attribute focus, box gauges, go no go gauges, things like that. So we have a tremendous gap of gathering the data of our processes that can help us to adjust equipment or understand what is happening in the manufacturing port. Examples would be we may use box gauges as go and no go to pass certain features on parts. But if they're just off or if they're just big or just too small, what is it that we need to adjust in the previous process to fix that issue? And that's just the data we don't have and then we can't trend it over time to see what's going on and really understand what might be causing some of our issues. Yeah, we have a big challenge with the paper based system. Paper device history records, paper routers, paper copies of sops on the floor, sops being updated constantly because of Some process optimization, retraining, you know, some of these more cumbersome things. And then in certain parts of the industry, you have a large time to learn a craft. I can give an example. We have a very specific hand finishing process in a product and I have talked to operators who've started on the journey of learning how to do that. They say could takes six months for them to feel comfortable doing it alone. And that's just not sustainable and it's not scalable. And so those are some of the challenges we have. We make very good products in the medical industry, but some of our processes are very old. And the way we understand what's actually happening on the floor and during the process, that data is not being gathered or might not even exist at the moment. So it doesn't give us something to look at for further insight.
- [06:44]
A
And that raises. Actually, for me, it seems like there are two problems here. It's on the capture side and then also on the consumption side. So it's one thing to capture tribal knowledge. What is the best way, especially as you said in these, I'd say very detailed processes that sometimes I guess would actually come down to instinct and just over time, the experience of how things feel, how things look. So I would say, if I'm hearing correctly, it's on the capture side, it's one problem, but then also on the consumption side of, let's call them SOPs or trainings. I've heard in the past as well, it's everybody's used to, especially the younger workforce, they're used to screens, they used to consuming things in a certain way, and it's not on paper anymore. So am I correct in saying that it's actually a two pronged problem?
- [07:37]
B
Well, yeah, but there's two aspects to the capture side. One is, is what are the best practices as you put out, or as you had mentioned, and what is the best way to learn and teach and train and spread the tribal knowledge. That that is true. But the other missing element of the capture is what's actually going on on the shop floor in the process. So if you're using certain type of gauges, feeler gauges, plug gauges, you're not capturing variable data to see how the process is changing over the course of the day or over the course of the life of the machine, or over the course of the shift or the time of year, you know, whatever the case may be. And so what happens is, as we know, engineers love data. And the more data that you can provide from inspection that you can Provide from monitoring in the machine that you can provide from trending over time. Well, that allows them to make decisions or to actually understand what is actually going on in the process floor and then perhaps actually uncover issues that might cause scrap or that might cause long process time or whatever the case may be and take those learnings and actually improve the process or completely reinvent it. So it's really twofold on the capture. It's how are you training the operator, how are they learning, how are they proving that skill set, what is the ramp up rate? And then on the second portion of it, what's actually going on in the floor, how are we getting a look of what's going on minute to minute on the floor from machine to machine? What's running in what machine? What was the last piece made in this machine? What is the feedback coming from the inspection equipment? What do we do with that data if we're even capturing it? And so you know, we are not in the state where we have the data where we can trend things, things and then actually see what's going on. How are tools wearing, how is the machine wearing when it comes to spindle time or uptime of the machine? These things are other operational data that we need so that we can really understand the pulse of the shop floor and the equipment.
- [09:44]
A
And as we see a lot in, in AI, it's so much information, it's impossible for a human to consume all of this. There needs to be some kind of system that's running this, but then giving the insights, giving the analytics to, to a human in a way that they can act on it. It's one way to have all of these dates, all of this data captured. How do you actually make decisions on it? And I guess that's, that's another point where they'll need to be more training and then adoption, of course, because now you have that split between the older generation that have done it in a certain way for a very long time and they know how it works and, and this new wave of employees that might be, once you are able to capture all of this, actually train and learn in that way and then you'll be sitting with the adoption because now you have to get both of these groups of employees working on the same system and improving it as you go along. So very much an intricate and a complicated system that you need to work with here. And I'm assuming within regulated industries like medical equipment, it even gets more serious because this is life and death situations when you're creating these types of equipment. You cannot have flaws. It really is critical to get this right.
- [11:02]
B
Yeah, that's correct. To be clear, catching the flaws is not the issue. We're able to catch those when they do not leave the facility. All medical device manufacturers are very good at that. The situation is, how do you get into prevention? How do you prevent the flaw from ever happening? And as you said, that all starts with the data. What is going on with the machine, what is going on with the operator and things like, we can get into it if you want to talk about the strategy for that, or we can talk about the current problem. Not sure what you want to do, but, you know, it goes from machine connectivity. Now you're gathering data, now you're analyzing that data, you're pushing it somewhere. Perhaps you're using AI, you're using some kind of algorithms, and you're presenting that data depending on your audience. So, for example, the data that a supervisor or an ops leader is interested in is going to be very different than that of an engineer. We can collect all that same data with the same machine connectivity and the same type of sensors in the same general architecture of that Manufacturing or Industry 4.0. Then it's creating the data analysis in ways that are useful for whoever is looking at it. So, for example, the leadership in operations are going to want to know what the cell output for each shift, each person, each machine is, and why, what's down for the equipment, what is the maintenance team doing with that piece of equipment? You know, understanding the pulse of the shop for from that standpoint and how that might affect their ability to make their components and hit their numbers. From an engineering standpoint, we're going to want to look at what is the data doing? How in control are we, how do we tighten our control window by changing parameters and that it might include things like what is the data on this machine as it runs? What is the tool life for certain tools? Are we changing them too often? Are we waiting too long, long? What are the things that we're doing that affect other processes downstream that we can fix in this station so that we prevent, you know, a scrap part or a bad faulty product to be made which will be caught in final inspection. So really it's about prevention and understanding what's going on, both in your process and with your people.
- [13:26]
A
And now you've pricked the ears of our audience because that's exactly what they interested in. Where are we going to start saving money? Because if you're creating the products and you can't use them because as you said you're very good at capturing when it cannot go off the shop floor, but preventing it is where the money saving starts. And I guess on that note, if we put it back to the frontline worker, you just stated that we need to present the analytics from the data to each consumer in a way that it actually helps them to do their job better, better. And for the frontline worker, it even takes a step further. You don't just want the analytics, you might want to ask it questions and get responses from the data on the questions you're asking. So it even, it takes it a step further. But again, it needs to start at the capturing. If you haven't gotten all of this captured, there's no data to ask questions of. So let's zoom out a bit. If a team actually gets this right, if they modernize the work they are taught new, new onboarders are onboarded correctly, what does it look like on the floor? What has changed and who, who has taken the lead on this to make that change?
- [14:35]
B
So yeah, if we talk about the onboarding culture for the operator, you know, with the capturing, the best practices of what happens in processes, especially in more manual processes, which still exist very much in medical industry today, and having that standard training for kind of raising the bar and saying that this is the best way to do this process, whether it's finishing or grit blasting or whatever the case may be, creating that baseline, that's the key. If you have that upper baseline where you know that this is the best practice and everyone trains to that standard, you're going to naturally have the higher yield, the lower scrap, and it's just going to be an easier process for people to train to because there's not going to be a difficulty in trying to figure things out or be puzzled when there could be a failure. As you're being trained currently too, you know, the onboarding and training process. We have trainers in the medical industry. You have people who are training coordinators and then you have, have techs who might train. Very comprehensive. You can't just have them say, here, make one lot of parts. I'll show you how this is done. Do you have any questions? Okay, let me watch you. Oh, that's pretty good. Let me watch another one. And then you sign off early. Not to say that that happens in every company. I'm not saying that. But you know, you can't shortcut the interaction with someone who's new to take the time and show them the correct way to do things because oftentimes when we've worked with the people on the floor in the human error prevention avenue of engineering and improvement, you'll often find that people misunderstood parts of their training. And you also find too that the sop, especially older ones with lots of words and not pictures, might not be very clear. And everyone has different ways to learn and come comprehend. So I would say the first step is to have a very comprehensive training and sign off procedure for what is the true level of how someone is ready to walk away from a person they're training, knowing and being satisfied that they fully understand the how and the what and the why of that process. And that requires patience and not rushing, you know, spending the time teaching them the things and then understanding the error points of a process and then working through that. Now, on the other side, from the data side and from the engineering side, you know, we've set up systems that very quickly can catch when someone is not trained or has somehow misapplied something. Because we're looking at scrap rates by shift. And so we're able to say, wait a minute, something. There's an anomaly here from this as well as other issues that we might find that's coming from shift scrap reviews, cell scrap reviews, the daily tier 1, 2, 3, 4 meetings all the way up to make sure that all levels of leadership are involved in understanding what happened in production and having their finger close to that. So you know, we do have those safeguards in place. But nothing replaces good patient training. Being patient, taking the time and having a true expert understand what the baseline and correct way is. And that becomes difficult when you're talking about 24, 7 operation. Is your best trainer only on first shift? How's the quality of your second shift or your third shift trainer? How do you get those people on the right shifts to the right so that you can actually standardize? Because what we do see in industry is depending on the shift, you can have the exact same process. And depending on how much manual process a certain cell might be, you can have vastly different yields because of the shift in the mix of the people due to the quality of training, supervision and knowledge.
- [18:42]
A
And that actually raises an interesting question because it's clear to me from what you're saying is that that baseline really is the foundation of whatever you're working with. If you don't have that baseline, how do you even test if somebody's properly onboarded? How do you check against is there a problem here? So that, so I'm hearing from you that you first really need to establish that baseline and that needs to come obviously from capturing best practices from your best operators on the floor. And then what I hear from you, a second problem problem is that even if you have a really good baseline, it could depend from shift to shift who's actually training and onboarding someone. Even those finer details could make a difference. So to me it sounds like, and as we see in most AI processes, that part where humans are just augmented by AI systems makes the most sense. So you've got that baseline captured and you've probably got the best practices captured in. It could be on video, it could be per photos, it could be a system on a tablet where an operator could ask questions with a tablet. But then that doesn't exclude the fact that you need a human there to still walk them through. And I think I've noticed it before. It's one thing to watch a video, but to be able to watch a video and then ask questions of a person next to you who's gone through the same training that is vastly more detailed just watching a video or just watching a person who might be close to the best practice but just lacking in one spot. So if you've got that integration of the AI with a human, that might be the best spot to fall in.
- [20:18]
B
Yeah. So a perfect example of that, I worked in a company once where they had a whole training center that they had replicas of clean room equipment. So instead of trying to train someone in a grade A or a grade bag to be a clean room, they gowned them up and they trained them in a training center with identical equipment. So the pressure was off. Right. They're now being professionally trained in a simulation of exactly what they were doing. They get their feedback, they're observed, and they're not under the pressure of, you know, the production timeline, especially in higher volume stuff. So that's one approach that I've seen that's been wildly successful in a company that I was once at. And then they'll again, that's the thing. There could be video training, there was digital training, but that human expert training to walk someone through and teach them these processes, that's really setting up someone for good success. Especially if you're transitioning into manufacturing with no manufacturing background. Right. It might be simpler if you've already been in a manufacturing environment. It let's say you're going from medical device manufacturer A to device manufacturer B and there's a lot of similarities. You already have a good baseline. But if you're bringing in someone from a whole other industry or you're bringing in Someone who's graduated from high school and they maybe just don't have that, that baseline of understanding and experience. You know, everyone's going to have a different learning. So that's the human element of it, right? The digital element of that is you create easy training platforms such as, you know, digital routers or not digital routers, digital work instructions, perhaps videos they could watch. But nothing I think will replace them actually doing and being supervised and guided and taught. I mean as far as when there's a lot of manual human intervention in a process, the goal is to actually eliminate the need for that skilled craftsman, that operator by transforming processes, bringing in automated processes, new technology, letting certain decisions be made via whether it's an algorithm or AI or other things like, and take the guesswork out of that and take that human variability out of, out of it. Not because humans are bad, because the variable condition can be so bad. Someone could be coming into work, they're not feeling good that day, they didn't sleep well that night. It becomes very, very easy to lose concentration for just a moment and then you, you know, you'll create scraps. So how do we take that away from there? And in the true nature of human error prevention, how do we set our operators up not to fail and to really create the tools that they can be successful because they're adding the value of being the ones that make the product.
- [23:13]
A
And that actually raises another interesting. I think so we've, we've spoken about capturing and then training but you mentioning the fact that it depends on even your background coming in, what type of onboarding you need and whereas in the past you might have all onboarders start in one spot and, and you take, take them through the training actually explicitly finding where what the skills are, doing a skills test and then personalizing the training from there. That would could win you a lot of hours on onboarding with, with employees that are actually further ahead and probably people lose interest when they feel like they already know the content. You start them on the level they're at and I guess that's where digital products are easier than a human because goes for a human to adapt. And I can start this class here now. Whereas if you can do a skills test and you stop the onboarding in, in the process where you need to start it from, you can save time on onboarding and then also not let them lose their interest in, in the training program.
- [24:14]
B
I, I would say, you know that there is a caution for digital training. You can add too much of it every medical and pharmaceutical device company I ever worked for had a digital platform, right? You did digital training on documents, you took a quiz, you signed, you passed the quiz, you signed off that you understand this document well. Again, if you're just reading it from a digital platform, if you're an operator and you can read through an SOP and answer some questions about that, having not done it, you're going to lose that. The retention is just not going to be there from just reading a digital document and seeing some pictures. So true digital sops and things like that are very important to be able to recall quickly. On the floor, help guide people. There's even, I've seen cases where the digital SOP walks you through step by step as you go station to station. This is, here's what happens in this station. There even it can be some animations and things like that. Those are all very helpful guides. But it's not going to be what will replace just the time and training of a human. The true real goal is to eliminate as much as you can of that need for high skill, high craftsmanship and to create repeatability and to bring digital, to bring the digital world into inspection with real time communication to the process, to allow algorithms and other software to adjust and not allow for the people to have to enter adjustments. So I can give you an example. In the CNC machining world, someone may make a part or a component and they measure it and go, oh, it's in, but it's on the low end. Let me offset it back to nominal. Then they accidentally put in the wrong number. Maybe they put an extra zero and they hit go, the offset's wrong. Next thing you know, the next part's ruined because it's something they had to do manually. How much better would it be? And this is the future is you bring in a digital piece of equipment, like for example, a cmm, you teach it that this is what you're inspecting, this is the nominal mark, this is how the part is from nominal. And as it inspects, it sends that information back to the machine controller and it adjusts to nominal on its own. And so every part, every part, yeah. No chance for flaws. And now your, your operation window, your SPC is so much tighter and your control limits are so tight. Now you're making everything a nominal. So if something happens and there's an event, you're well within your tolerance when the part's still good, you know, and then if something is catastrophic, if you're monitoring the machine like a tool breaks, all of a sudden, for example, let's say you're monitoring the load on your spindle. Well, load goes to zero, it's supposed to be cutting, it knows, wait a minute, I'm hitting error, something is wrong and it sends it alarm. Stop the process, don't go to the next tool and then crash it into the thing that it thought was cut but was never cut. And that's really the future of manufacturing and that's the application is like I said before is you need to get machine connectivity, you need to get the data, you got to understand what the data is doing and then feed it back into the machine and feed it into the cloud where then you can use AI and algorithms and, and other systems to understand what's going on. At a minute to minute situation then you should be able to have much tighter control.
- [28:00]
A
Before we close out the episode, I would like to ask you for the leaders. Maybe it's all still paper processes. Maybe a lot is paper processes. And they see a lot of the current workforce retiring out. They need to onboard new people. Where do they start without disrupting the.
- [28:19]
B
Yeah, that is the challenges is the workforce flips over. They have to. One of the things that I would say to the leaders is you're going to have to carry extra headcount to allow for proper training. You're going to have to create that onboard timeline and then you're going to have to set up processes that we know can onboard much quicker, get people used to the environment, move some of your senior staff to the more difficult jobs. You have to be very smart with your workforce. Again, this is where things like training matrixes come along. And so that can be digital. You can understand just by who shows up on your shift that day, who can be trained where or who is certified and trained where. And then how you move people around to cover your more difficult processes. So as this new generation of work workers come in, you know you have to create a path or an avenue to say maybe you come in as a packager and after you learn packaging and the general medical device culture and it's a whole other language, right. How do you learn how to access our systems, how to scan, how to sign off on things, how to do routers and all that, especially if there's still paper, they get comfortable with that, then you move into the next level and the next level. And meanwhile on the other side, one of your more higher craft processes, you're cross training those people who've been there and understand the manufacturing side of it. And then you know they're, they're becoming masters and owners of the more difficult or more, let's say more craft like processes instead of putting someone new in there, hiring them and expecting them to understand it. And it will be a benefit. Right? It allows people to grow in their career, it allows people to cross train, but it creates a pathway where people don't get frustrated and leave because they're overwhelmed with something they just don't understand that makes sense.
- [30:18]
A
So I want to wrap up the episode. It's clear that human, the tribal knowledge needs to be captured in some way as far as you can use it. But that doesn't mean that you are going to be able to lean on it 100%. You still need that human interaction. Creating that baseline from the best practices is really important because that's really going to point you in the direction when you're leaning on the side of more error than you can afford and where you need to add more training, where maybe you need your higher level experts involved in the training and then also pinpointing probably who your best trainers are and where you're really struggling and use the employees that are based at training in those spots. And then for leaders that need to start now, there is a place for digital digitization in a sense for those training matrices, figuring out where to put the people best before they get frustrated and also really leveraging those experts and in the higher end jobs and then knowing that you might have to bulk up on the, on the employee side for a while until you can train them up. Thank you for your insights. Anything you want to leave the listeners with that I've missed.
- [31:32]
B
I would say that we're in an exciting time in industry with the application of digitization, AI and even just software. In a lot of these challenges that we've seen over the last 20 years, there are solutions that are going to be coming and it really makes a lot of sense to invest into looking into them. It's going to be a little challenging to understand how to roll it out and apply it, but what it will, how it can transform manufacturing, it will be incredible. It will be incredible and it will solve a lot of these issues.
- [32:08]
A
A real optimistic way to end. Thanks Sebastian.
- [32:11]
B
Thank you very much for your time. Foreign.
- [32:28]
A
I think there are three key takeaways from our conversation with Sebastian. First, capturing expert knowledge and turning it into standardized, repeatable processes is becoming essential as experienced operators retire and manual craftsmanship becomes harder to sustain. Second, without real time visibility into what machines and processes are doing, leaders can't prevent errors early or tighten control which limits their ability to stabilize output across shifts and finally modernizing training and data capture together. Pairing human expertise with digital tools creates the conditions for faster onboarding, lower scrap, and more predictable production performance. If you have an AI solution, position your brand alongside the Fortune 500 leaders defining the enterprise AI roadmap for the opportunity to showcase your solution to the executives currently funding and scaling global initiatives. Partner with Emerge. Secure your partnership@go.emerge.com partner that's go.emerj.com partner for further executive level analysis and to join our network of leaders delivering with workflow impact with AI, visit emerge.com on behalf of the team at Emerge. We'll see you on the next episode.
- [33:55]
B
Sa.