Episode 4. Integrating IT and OT into the manufacturing industry: Best practices and common pitfalls

According to Remco Ploeg, CTO at Manufacturing Microsoft Innovations Hub, Microsoft Netherlands, IT and OT teams need to work together more closely to run data-driven processes such as predictive maintenance. Microsoft has a lot to offer in this area. There are also several ways to effectively integrate data processes into manufacturing.

Remco shares about the pitfalls of data and AI pilot projects.

Alex: Good afternoon! Our third guest today is Remco Ploeg, the CTO of manufacturing at the Microsoft Innovation Hub. I’ll be discussing some interesting topics with Remco, including the integration of IT and OT. We’ll also be talking about digital twins, predictive maintenance, and, of course, how Microsoft is positioning itself within these areas.

Alex: Remco, I've mentioned the integration of IT and OT in previous conversations. Could you tell us a bit more about it?

Remco: Yes, within Microsoft and with our clients, we see a lot of challenges around integrating IT and OT. We’ve noticed that these two worlds are coming together more and more.

OT specialists are increasingly using IT technology, while IT specialists are being required to implement OT technology. This demands collaboration because, right now, there are often two separate teams managed by different people.

These teams must work more closely together to succeed with use cases like AI in factories and production lines. The IT and OT worlds must truly merge to enable innovation and efficiency.

Alex: Thanks for that helpful explanation. What do you think Microsoft can offer specifically for the manufacturing industry? This is obviously why we’re here today.

Remco: Yes, we are now developing solutions that can be used by both OT and IT specialists. Whereas before, it was always about different systems from different vendors, now we offer systems where IT and OT can work together. Both teams can work in their own way but with the same technology.

This is cheaper for the manufacturing industry. Additionally, security is one of our top priorities, especially in these times. Systems must be properly secured so that hackers, for example, can't take over machines.

With our integrated tools, we combine both worlds, making them cheaper and safer while providing more space for innovation. This helps our clients operate more efficiently and safely.

Alex: Well, you’ve already mentioned that cost-effectiveness and security are key aspects. Of course, these are also the main points we hear when talking to different parties in the manufacturing industry.

Alex: I've also heard that digital twins are being used in manufacturing. Could you elaborate on this term?

Remco: Yes, a digital twin is a virtual replica of a person, process, asset, or a combination of these. You can think of it as a kind of digital copy of your entire factory or production line.

The goal of this is to bring all sources of information together. This includes data from the production line and machines, as well as IT systems such as ERP. This way, you get a completely virtual image of your factory, which allows you to realize various use cases.

For example:

• Preventive maintenance
• Optimization of overall equipment effectiveness (OEE)
• Other data-driven applications

So, a digital twin is a virtual copy that combines data from both IT and OT systems and serves as a basis for innovation and optimization.

Alex: And is this something that can be rolled out across the entire manufacturing industry? In other words, is this something every company would be interested in pursuing?

Remco: Yes, when we look at digital twins, we see two main applications.

1. Product Twins

These are digital twins of individual machines or products that manufacturers build and deliver to production companies. These digital twins contain data and insights specific to a particular product and are often provided by the supplier. This is something we’re seeing a lot of, and this type of digital twins is already being used throughout the manufacturing industry.

2. Factory-Wide Digital Twins

In this case, the factory itself brings together all individual product twins and creates a complete digital copy of the entire production line or even the whole factory. This allows companies to integrate data from different machines and systems and ensure that they communicate with each other. This provides a holistic overview and opens the way for process optimization at a much higher level.

Both applications are interesting for the manufacturing industry, depending on the needs and scale of the company.

Alex: So, this means that they’re being used on a very large scale.

Remco: Exactly, yes. We’re already seeing this quite a lot in the market. It helps bring data together, which makes it easier to, for example, do preventive maintenance.

Alex: And which Microsoft technologies could be used to implement this?

Remco: Various technologies. One of the products we are about to launch is Azure IT Operations. This allows us to easily bring data from factories into the cloud, and in the cloud, we have a product that our customers and partners can use to create a digital twin of their product, production line, or even the whole factory.Alex: And, as you mentioned, data from factories goes into the cloud, which is, of course, a hot topic right now—hybrid solutions and security when using digital twins.

Remco: Yes, that's correct. What we often see in the market is that whether you want to process your data locally or in the cloud depends on the use case. This may be due to privacy or security considerations. That's why our products run in a hybrid model: They can run locally, in the cloud, or somewhere in between.

This ensures that we can support any use case—for example, in safety use cases, where you film people in a factory to see if they’re wearing a helmet or the right shoes. In this case, you probably wouldn’t want the footage to go to the cloud due to the potential risks of the data being accessed by others.

In this case, we perform the analysis locally. For example, we remove faces from the footage and only send the outcome—such as, "Is this person wearing a helmet? Yes or no"—to the cloud. This way, we combine the benefits of local processing with those of the cloud, depending on the needs and requirements of the use case.

Alex: That sounds very interesting! I also want to talk to you about predictive maintenance, a term that, of course, is often used in the manufacturing industry. We hear this not only from partners but also from customers.

What can Microsoft do to successfully implement predictive maintenance?

Remco: For predictive maintenance, you first need insight. You need to know what's happening with a machine. If a machine breaks, you need to be able to determine what the conditions were when it broke.

Microsoft offers various solutions for storing data from machines, both on the cloud and locally. The first step is collecting and storing the data.

In addition, we offer standard models and the ability to build your own models for predictive maintenance. With these models, you can predict, for example, when a machine might break down. You can analyze filters, vibrations, or other factors that could predict wear and tear or failures.

We make these models available to partners and customers so they can use them in their own environments to effectively implement preventive maintenance.

Alex: How long does it take to implement all this? In other words, what should a potential customer in the manufacturing industry do to get started?

Remco: Yes, I think there are two important steps:

1. Use standards: Ensure that you're using standards that are already available in the market. Without standards, it’s very difficult to extract data from machines and make it consistent.

2. Use or develop models: Make sure you're purchasing models, building your own, or using models that are already available in the market. Once you've collected the data, you can start predicting maintenance needs and failures.

Alex: And something we often hear when talking with companies is that they can be afraid of change. That's not only the case here but probably when it comes to other topics as well. What should we—or you and we—say to companies in the manufacturing industry about why this is almost a necessity if they want to keep up with the industry in the next 10 years?

Remco: Yes, that's also what we often see. I think it has a lot to do with adoption and change. Often, the workers on the shop floor aren't involved in these kinds of projects. Instead, a solution is suddenly introduced with the announcement that your job will be different.

It's important to involve people in the implementation of such systems. Think about their practical experience—they often know the machines inside out and can tell, for example, just by the sound of a machine, what might be wrong.

By actively using this knowledge and involving them in the process, you increase the chance of success, and the changes will be more likely to be accepted.

Alex: So, you're essentially creating an interaction. You’re leveraging the knowledge these people have built up over the years working next to the machines and linking that to everything related to data and AI.

Alex: Let’s talk a little more about Microsoft. As mentioned earlier, you work for Microsoft. Could you tell us a bit more about the technologies available and how Microsoft is currently positioning itself in the manufacturing industry?

Remco: Yes, we invest heavily in the standards that are used globally in the manufacturing industry. For example, we've been part of the OPC Foundation for about 16 years, which is one of the key organizations for standards in areas like manufacturing.

These standards are a very important focus for us. We also implement them directly into our products. This means that if you're already using these standards in the manufacturing industry, our products can communicate directly with the machines. You can pull data straight from the machines without spending a lot of time transforming it or creating data models.

Because our products already support these standards, you can quickly move forward with your use cases and create added value. Instead of getting stuck on technical transformations, you can focus directly on the applications and outcomes that really matter.

Alex: And people always like to hear success stories. You've been working in manufacturing at Microsoft for a while now. Are there any success stories you could share? Ones where you think, This is where Microsoft made significant progress for this company.

Remco: Yes, I think there are many success stories. Years ago, we started with the question, “How can we easily collect data from factories?” What we often saw at the time was that factories reported based on Excel files they sent to the central organization. That was the report.

A big success we see now is the automated collection of data from various factories and reporting that centrally. This way, everyone involved has the same view, which provides a lot of transparency and efficiency.

A second success we see is the rise of AI in the industry. We make sure that data is made available to many more people and that they can naturally ask questions of that data. This is something we have seen grow tremendously with our customers in recent years.

Thanks to standards and AI technologies, like GPT models, customers can now ask questions like What was my OEE (overall equipment effectiveness) last week? or How can I optimize that? The technology can then suggest things like Pay attention to this or Adjust this.

These kinds of applications make data more accessible and help companies take immediate action based on specific insights.

Alex: What pitfalls do you still see when it comes to data in the manufacturing industry in the Netherlands?

Remco: What I often see is that companies run a pilot project with AI. Such a pilot is often successful, for example, on one production line or machine. But the challenge lies in scaling it up to all production lines or factories worldwide. This happens because companies often don’t think through how to connect data to data platforms and AI in a standardized way. This is a big challenge and a common pitfall.

The second challenge is not using standards or working with closed systems. For example, when you want to implement something in your factory or on a line, you sometimes discover that a system is closed, meaning you can’t access the data. This can happen because the machine supplier has restricted access. In such cases, you either have to install extra sensors to collect data or pay extra to get access. This is complicated and expensive. So, when purchasing new machines, it’s important to check whether you can access data and whether the system uses open standards.

The third pitfall is related to security. In OT (operational technology), security is crucial. IT people also see security as important, but with OT, there is often another element to consider. Some factories can be controlled remotely, and if something goes wrong, it can have dangerous consequences. Therefore, it’s important to include security from the start of a project, even in pilots. This should not be postponed, thinking that it can be sorted out later. You have to make sure the entire system is secure from the beginning, end to end.

Alex: So, basically, what I’m hearing is this: Before starting a pilot, it’s important to make sure the foundation is completely solid. This means you may need to take a step back to ensure you’ve thought through everything.

It’s not just about having a helicopter view but also almost literally writing everything down so the foundation is strong. From that solid foundation, you can take the next step and start building.

Remco: Exactly, exactly.

Alex: I have a few statements here. I’d like to ask you to pick the first one. Please answer “yes” or “no” first and then give a brief explanation.

Remco: Statement: The benefits of preventive maintenance for the manufacturing industry are overrated, as implementing it often brings more problems and costs than expected.

Response: No, I don’t think so.

Alex: And could you explain a bit more why you disagree?

Remco: Well, yes, the benefits are definitely there, so I completely disagree with that. The implementation is actually fairly straightforward.

My advice is mainly to start with preventive maintenance. Collect your data and do it. Often, all sorts of problems can arise and reasons not to do it, but the technology is available now. We can collect data, and we already have all kinds of models at our disposal that you can use to implement preventive maintenance successfully.

Alex: And can we say that if you don’t start doing this, you could lose your competitive edge in about five years? Could you say that it’s becoming more of a must than just something that seems interesting?

Remco: I think so. I actually think you should have already started doing this. Especially with our aging population, and especially in the manufacturing industry, it’s essential to have these kinds of tools at your disposal.

These technologies prevent machines from breaking down and from needing repairs done by people who may no longer be available or who are in short supply. So yes, you should have already started, and if you haven’t, you should start now.

Alex: So, indeed, urgency is required. I’ll pick a statement too.

Statement: Work and skilling: Despite the efforts of companies like Microsoft, most workers in the Dutch manufacturing industry are not adequately prepared to succeed in an increasingly digitalized environment.

Well, as I mentioned earlier, we regularly speak with parties in the manufacturing industry, and this is a common concern. So, I think I would generally agree with this.

I think it mainly comes from the fact that people are often not aware of all that’s possible. For example, when we speak with family businesses that have been running perfectly for two or three generations, the focus is often primarily on their core business. As a result, they’re less open to potential changes, even though these changes could ultimately help the business take steps forward.

As for the part about “despite the efforts of companies like Microsoft,” we just touched on that. Microsoft certainly wants to play a supportive role here. And as a partner, we also have a responsibility to support companies in this. This could be through webinars or by actively providing information.

Remco: Yes, partners are super important for us to be able to scale in the market, especially in the manufacturing industry, so I’m glad you’re doing that too.

When we look at this subject, I think a major challenge is that people are often not included in the process. My advice to the market would be to involve people in your change process. Make sure they are engaged and understand what’s happening.

They shouldn’t just come to work one day and hear that we’re implementing preventive maintenance and that their job is changing or disappearing. Actively involve them in the change and make them part of the process.

Alex: No, I can imagine they wouldn't be happy about that. We always try to sit beside potential partners and customers, not across from them.

Alex: Well, I’d like to thank you very much for your time and, of course, for the open conversation!

Remco: Yes, thank you too. 

Alex: Today, we discussed the pitfalls that data pilots can bring, like being limited to just one machine or one production line. We also talked about how important it is to connect data in a standardized way and how to integrate IT and OT.

If you’d like to continue the conversation with us so we can start a free proof of concept for you, and perhaps explore with us and Microsoft what’s possible, we’d love to hear from you!