Interview – Syntegon’s CIO Johan Nilsson on digitization in pharma

For digitization, data, and AI in pharma – the customer’s perspective is important

Syntegon CIO Johan Nilsson at Achema exhibition

At the recent Achema exhibition in Frankfurt, Syntegon’s CIO Johan Nilsson gave an insight into the digitization activities of the company. The company presented a small box at the booth that can collect data from its machines for insights into the process. PSA’s Susanne Bluml talked to Johan Nilsson and asked for his views on digitization and the use of AI in the pharma industry.

Susanne Bluml – Is the digital transformation already going forward in pharmaceutical machines and pharmaceutical production?

Johan Nilsson –There is one answer to the question when it comes to the development of drug assets. How can we do the drug development process in a much faster way with the help of digital means? In the production process, you manage huge amounts of data and you can do it faster.

However, one can imagine that there is quite a lot happening there. When we come to the manufacturing of drugs, I think it’s an industry that is a little bit conservative – it could be for reasons such as regulatory compliances. My feeling is that perhaps in comparison to the food industry, they wouldn’t move as quickly into something they have an interest in when the tools are not yet intensely proven.

There is such high profitability on drugs and with hundreds of millions of euros invested in a factory, a constant desire to optimize the utilization of the assets. The question of the output of that factory is fundamental to the business success of companies.

In the pharma industry, that pressure is not there in general yet. And maybe contract manufacturers who don’t own the brand themselves can optimize the profit with constant small steps of improvement, performance, the utilization of assets, and so on. The entry point for digital transformation in the manufacturing process of drugs is certainly the quality side – the questions of how to secure the quality of the drug and how to reduce the waste of some of those very, very expensive drugs will gain more importance.

SB – Is serialization the step for digitalization to enter this market due to regulation?

The identification of the product and its uniqueness is one of the challenges we have. What makes it a little bit difficult is that there are no real global standards. For a company like ours, it’s sometimes a bit challenging. What should we adopt?

Being an equipment and service provider to the pharma industry, we are not massively investing, for example, in serialization and aggregation of data for track and trace purposes of medicine. There are separate companies that focus on that. You could, maybe, see a future where you see us partnering in between those. This is why one needs some data from equipment for serialization. It all depends on the level of aggregation: if you want to identify one tablet or a batch of tablets.

SB – Do you see already use cases where AI could help in the pharmaceutical industries?

I think one point is when you try to predict the failure that might occur. You use data from your machines and artificial intelligence that help you in taking a large set of data and analyzing that data. Through machine learning, for example, and with prediction with 75% probability you will have a problem in this part of the machine within the next three months and then the closer you get the higher probability. That is an aspect of AI that we invest in.

It depends a little bit on exactly where you draw the boundaries of AI – it’s a bit of an abstract and the technology giants like to focus on use cases, which has been very driven by the Covid-19 pandemic. For example, how to provide remote assistance with video. Think of the use of different types of glasses and how to make use of knowledge and information that is typically in paper binders. With the use of glasses and the information we collect, we can make that available to people instantly. You can provide support in seconds instead of days.

SB – In pharmaceutical industries, machine learning is related to special use cases, which means that one cannot take out the data for machine learning and apply it to other use cases. How is this solved?

First, you can apply machine learning on a ball bearing, for example. It doesn’t matter if the ball bearing sits in a food machine or a pharma machine, it’s the same algorithm. It is relatively basic artificial intelligence that would fall in this field of machine learning and that can be very useful in many applications in pharma machines. For example, we discussed with one customer out here earlier, the use of AI in HEPA filter supervision. You have a lot of HEPA filters to keep a clean environment in drug manufacturing and some in food manufacturing. How do you know when to replace the HEPA filter? Do you wait until you have a problem, a quality problem with your product if you can measure pressure drop over the filter? Those types of use cases are going to be much more common in pharma because the cost risk in a balancing formula is very different from food. But the consequence is not as bad as in pharma, so the willingness to invest in pharma industries is greater in these types of applications.

SB – And do you have any discussion with the customer with regards to AI procedure?

The discussions we have are more in the packaging area, of course. And they are more driven by securing the quality of the product.

SB – How do they validate that AI works?

It’s typical to run parallel processes. So whatever method you have on securing the quality today, which normally involves a lot of human beings and human assessment, you have to run the AI process in parallel until you see results. And sometimes it works, sometimes it doesn’t.

You cannot replace a quality assurance process overnight with an AI that is led by people who use different types of equipment to measure things as well and check. You have to prove that. So very often you end up testing them in parallel.

SB – Where do you see the biggest hurdles when introducing AI into the market?

There are a few different hurdles. The biggest one is probably still a bit of a lack of standards in the world because of the automation platforms for machines. Being a big customer who buys equipment from 10 different companies, (probably all here in Achema), all suppliers have their standard for how you get data out of the machine. And then, of course, we offer our solutions to our machines. The others offer their solutions to their machines and in the end, the customer ends up with 10 providers of the same thing. And these things typically don’t work together. It’s a trap many of us fall into – we take our technology perspective instead of taking the user’s perspective of this because the user doesn’t buy machines only.

Even if we would love every pharma company in the world to only buy machines from Syntegon, the reality of the matter is not that. And then it is a big challenge to be the one who says I provide a solution that is machine-agnostic. ‘I don’t only provide a solution that is for our machines.’ Companies like us have to watch out for companies such as Siemens, Microsoft, Amazon, SAP, and others because they want to claim this space. Of course, they attack it from a completely different angle.

SB – Is there any workgroup where you are working on these topics?

There is not one big group, but for example, if you take Europe, SAP is trying to take the lead of gathering companies like us, from different industries.

They try to create solutions because they cannot provide everything – they don’t have what is referred to as domain knowledge and they don’t have industry experience. They know a lot about data and how you transfer data and such things, but they don’t know what is relevant data out of the pharma machine.

SB – Do you think that SAP is a partner for such production data because they are only in the administrative business?

No, they’re moving very fast. Yes, they want to be the kings of the office, but they want to become the queens of the factories as well. Of course, the more you automize the business, the more exchange of data there is between the manufacturing side, the sales side, the invoicing side, and so on.

SB – Where is the danger for Syntegon in such a structure?

I think the danger is that if we are naive and think that we can do what SAP tries to do or what Microsoft tries to do. I think we just have to be part of that and understand: where is our value creation? You know our value creation is, of course, not in the field of SAP. Sometimes you need to access the data in nanoseconds and that is where SAP is an expert as is Microsoft. We are experts in knowing what is important when you want to manufacture drugs with high quality – what the data is relevant to, what conclusions can you draw from what data, et cetera.

It’s also important to have the customers’ perspectives in mind. Customers are afraid of handing over the data to clouds. That is an important thing. Some companies went down the path – ‘We have our own cloud.’ And that doesn’t resonate with food and pharma manufacturers, of course, because they say where is that? Is it in Russia? In Ukraine or where? When you say that we operate on the Microsoft cloud or the Amazon cloud, or the SAP cloud, they know that’s safe because a lot of other data from their business is already on those clouds. They know cybersecurity will work. If we would go into these discussions and be naive and think that we can solve that ourselves, it wouldn’t work.

SB – Concerning India and digitalization, do you have any special requirements from Indian customers in terms of digitalization or their special market requirements?

Not that I’m aware of in terms of specific functionalities. If you take India and China, their expectations are always much higher speed from idea to reality. They work in a much more dynamic world. They have societies developing much faster, of course. If they have an idea today, they want to have it working in six months; they don’t want to wait three years.

SB – Where do you see Syntegon with digitization in the next Achema, in two years hence?

I think that you will see a significantly higher number of machines that are connected and provide data for different types of use cases. Everything from just displaying performance and being transparent with machine performance. You will also see the first cases in full operations of failure prediction. You will see much more use of remote support.

How you take knowledge, e.g. knowledge that might exist in one corner of Germany and that you make available within seconds in Bangladesh, for example. Very practically, it will not take five hours to set something up, but it needs to take five minutes to set something up and how people can communicate. But you know when you have a German specialist engineer talking to a Bangladeshi technician, they are not wholly native in English, not even all our employees are experts in English, and the same for our customers in China and every corner of the world.

We will see drastic changes in communication. And our knowledge can travel. I wouldn’t buy shares in airline companies because the amount of business travel for sending the technician across the earth now to solve a problem is going to be drastically reduced in the next two, or three years.

SB – Mr Nilsson, thank you for this comprehensive overview.


Please enter your comment!
Please enter your name here