Turning Manual Manufacturing Variability into Data‑Driven Control – with Sebastian Dykas of Smith+Nephew

**Marilie Fouché** (0:12)
Welcome everyone to the Emerge AI in Business Podcast. Today's guest is Sebastian Dykas, Director of Manufacturing, Engineering and Maintenance at Smith and Nephew. Smith and Nephew is a global medical technology company focused on surgical devices and wound care. Sebastian joins us on today's show to explore how medical device manufacturing still runs on paper workflows and limited machine visibility, and what leaders can gain by moving toward tighter data capture and targeted automation. The conversation centers on how clearer real-time insight into processes can reduce variability, stabilize output, and set the foundation of more reliable long-term operations. Just a quick note for our audience that the views expressed by Sebastian Dykas on today's program do not reflect that of Smith and Nephew or its leadership. In this episode, we cover precision-driven process control in manufacturing. To go deeper on AI topics and learn how leading organizations define the right data and why cross-functional collaboration is essential to getting it right, download our free PDF report Beginning with AI at emerge.com/aik1. That's emelj.com/aik and the number one. Now, the conversation with Sebastian.
Sebastian, welcome to the show.

**Sebastian Dykas** (1:40)
Hi, it's good to be back.

**Marilie Fouché** (1:41)
I feel like this is going to be really interesting. Medical device manufacturing has seen big advances in technology and regulation over the years, but when you look at how day-to-day work is done, it still seems to be a clear gap between what's possible and how most operations function today. A lot of teams know they should be more digitized and more AI enabled, but they're still working with all the ways of organizing processes and information. You've been close to that reality. When you look across medical device manufacturing, where do you see the biggest gaps that are keeping the industry from operating at a modern digitally enabled level?

**Sebastian Dykas** (2:14)
Yeah, there's several things that come to mind when it comes to gaps. There's still a lot of paper. We still have a lot of paper routers, paper processes, paper sign-offs, log books or checks and signatures. We have data collected in ways that it's impossible to digitally and actually be able to analyze it. Another challenge we have is that we don't actually, in a lot of cases, understand what is going on with the equipment during the time of the equipment.
When it is no longer working properly, but we're not monitoring it until that condition comes so we can prevent something like that. Then finally, I would say in the inspection world, we still have a lot of attribute gauges where we're checking go, no-go conditions, but we're not collecting actual data of measurements.
And then looking at that and seeing how processes are breaking down over time. So really, in a lot of ways, we're blind in the ways of what's going on with our product as it's being made. We're very good at looking at the yield, first-pass yield scrap rates when it comes to the end of the process. Maybe even certain checks in process of certain parts of the product where we can scrap mid-process and we can understand, we catch it because we know at this point, it's not going to be good if it continues along the way. But actually looking at what's happening at the point of every machine, every part of the process, what is the last piece look like that just came off the machine? And does the machine know? And are we recording that to trend it over time? It's that lack of data that is actually helping, are not helping us to be able to solve some of our larger problems.

**Marilie Fouché** (4:06)
It sounds to me like, you know what the what is? It didn't work out. The part is failed, I've got scrapped.

**Sebastian Dykas** (4:11)
Correct.

**Marilie Fouché** (4:12)
But you're still missing that why? At what point did the breakdown come? And I'm wondering, is this an edge data capturing problem? So is it there on the machines themselves that you need some data captured? Obviously, it seems like you are saying, we don't need humans to capture this on paper. We don't need more humans capturing numbers on paper. This needs to be an automated capturing. Is that on the edge, on the machines themselves, that you really need that captured?

**Sebastian Dykas** (4:39)
Yeah, so it depends on the process, right? In the medical industry, believe it or not, we still have processes that are highly operator dependent. So there is that situation. But we do have very reliable machines, especially in the world of CNC. But you get drift, right? You have different ages of machine, you have different conditions that occur.