Alfred Krupp founded his company in 1811. Now, over 200 years later, he is the namesake and inspiration for an artificial intelligence solution built by thyssenkrupp Materials Services (tkMX), one of Germany-based thyssenkrupp AG’s strategic business areas – and the largest materials distributor and service provider in the western world.
The “alfred” AI solution, powered by Microsoft Azure, helps the company analyze and process more than two million orders per year and better serve its 250,000 global customers.
Though alfred has been in place for just under a year, the solution is already helping tkMX optimize its logistics network – allocating materials to the right location much faster, minimizing transport volume and enhancing usage of the company’s transport capacity.
Transform caught up with Axel Berger, head of digital transformation at tkMX, to hear more about how alfred is changing the business.
TRANSFORM: Tell me about alfred and why tkMX developed it. What business challenges were you facing?
AXEL BERGER: We are a wholesaler, so data insights and data algorithms are possibly one of the strongest levers we have to improve our business. We had a lot of data that we weren’t using before, for three main reasons.
First, we didn’t really have the expertise to work on specific data science topics – we had the data, but it wasn’t always available. Second, data quality was an issue. And third, we lacked the technology to store data in different formats to use it and make it available in one central location on a massive scale. We also lacked the related tools to really analyze it, visualize it and finally, build algorithms out of it that could be deployed in different scenarios.
There are many possible use cases for wholesalers, and it took us a long time to pinpoint the use case that we should implement first. The major topic we’ve focused on is network optimization. How can we optimize, for example, transport costs or our supply network? How can we reduce the stock that is delivered from A to B without sacrificing our service levels? So the first project that we’ve worked on is network simulations within our German trade network.
It’s important to note that alfred is growing through its use cases. We didn’t create a huge global platform that could do everything. The first use case requires a specific amount of data, computing power and certain tools. But with additional use cases that we are now implementing, alfred is growing.
TRANSFORM: I understand you developed alfred internally. Can you tell us a little about that?
BERGER: Alfred came to life in early 2018. The biggest challenge was definitely data availability. You can have the greatest technology, the best tools, but the biggest challenge is to get quality data. Another challenge is to have the domain knowledge, the expertise in the specific topic to really make it relevant.
Everybody’s thinking that if you just use data and artificial intelligence, in the end this artificial intelligence will give you the insights that you don’t know yet. But that’s not happening. It’s about having the right data of the right quality, the expertise and the domain knowledge on a specific topic, and the technology to run it. Technology is the easy part, because nowadays there is someone like Microsoft with the technology. But to bring data and domain knowledge into the project and to understand the use case and the questions you are trying to answer, that is the hardest part.
TRANSFORM: Can you walk me through what alfred might do over the course of one day?
BERGER: There are so many things that alfred can do! Alfred dynamically tells us from which site we should ship which material to which customer. Alfred optimizes our stock levels. Alfred tells us what the perfect price for a specific customer for a specific product is. Alfred visualizes and tells us which customers are profitable, and which customers are not.
Alfred can help us build a predictive maintenance model for our machinery, and tell us which machine is about to break. Alfred also helps us to optimize our supply network in terms of physical sites – where should we open the next site or close it down, and which materials should be subbed somewhere else. It helps us to get better purchase prices because it helps us in negotiations and the bundling of materials that we want to purchase. These are all current or potential use cases.
TRANSFORM: I understand it’s still evolving, but what is the biggest benefit alfred has had on your business?
BERGER: We handed over decision-making to alfred (a machine) that relies on data. One of the taglines that we use for alfred is “intelligence in each transaction,” which means that we want to build decision engines. Alfred already delivered the first decision engine: The system tells us from which location the customer is to be supplied – taking into account all relevant frame data. That was our first decision engine, you could say.
TRANSFORM: What has been the employee reaction to alfred? Have they embraced alfred, or was there some resistance early on?
BERGER: People weren’t resistant to alfred, because right away we could show them how alfred would help them in their daily work, and the benefits we’d gain. With the use case we’ve been working on, alfred doesn’t imply any layoffs or redundancies. It is purely optimizing the way we are working, and helping to enhance the impact our employees are driving. So alfred is seen positively.
BERGER: Yes, absolutely. We helped them, trained them, involved them in the process very early. We trained them in the tools. What we are also planning is to deploy data labs, small versions of alfred, so people working on a specific data problem can use alfred to solve their own problems with just a push of the button. We teach them how to do this – how to use Microsoft Power BI, for example, to visualize their own data. That helped a lot because they started to work with data and to better understand what it’s all about and how it can be utilized.
TRANSFORM: How else has alfred helped your employees achieve more and optimized their work?
BERGER: Alfred has helped employees by enabling them to simulate tkMX’s network setup, which was extremely difficult before because our network is extremely complex. It has helped with data availability – the employees have much more data that they can now access themselves, without involving anybody from data warehousing. And obviously by increasing data transparency.
TRANSFORM: Have new roles or opportunities opened up to support alfred?
BERGER: Yes, of course. Roles like data engineering, data architecture, data science, solution designers – these are all new roles that we staff now.
TRANSFORM: What advice would you give other companies that are considering launching an AI initiative?
BERGER: I’d like to shift the focus away from the buzzword “AI” and better discuss what’s behind it. I don’t believe that there is an artificial intelligence as such. We have focused algorithms.
In other words, what I would recommend is to calm down and don’t be afraid of AI, because the methods are 60 years old. What has changed are the opportunities that advanced technologies such as cloud and edge computing provide and the pace at which they evolve. So, businesses need to get used to these new technologies, and use technology that is easy to handle – like Microsoft Azure. With Azure we can quickly launch applications that can be used for data aggregation, manipulation and analysis with the click of a button, with only a few people in the beginning.
To start, I would recommend taking data, searching for your first use cases, and just building them without engineering them forever. Clarify the questions you want to answer. Don’t believe in overarching algorithms that will solve the problem of finding the question, the use case. Because otherwise everybody is expecting results for something that you don’t even know is a problem.
TRANSFORM: Based on your experience, what concerns or rewards do you see for society as AI becomes more ubiquitous?
BERGER: Again, I would say calm down and get in touch with the methods and technologies behind AI. People are fearing things they don’t know. If you get in touch with it and understand what’s really behind AI, then I think it’s easier for people to understand that we are far away from real artificial intelligence. We see specific use cases, specific technologies to solve specific problems, but nothing like a mastermind.
It’s important to talk about AI and engage in the public debate, because with the evolving technology around machine learning and AI, there are questions to answer, including both ethical and legal questions. For example, the much-used example of an autonomous car. How do we cope as we give more and more autonomy and decision-making capacity to machines?
I studied mechatronics some 25 years ago. With mechatronics you were already talking about cyber-physical systems and programming and automizing machines. So IoT is nothing new. It’s just the technology has evolved and that gives us new opportunities.
When you look at artificial intelligence, the methodologies are out of the 1940s, 1950s – neural networks, for example. It’s nothing new. It’s all about cheaper storage, more computing power and better connectivity, but also about standardization and harmonization of data. And if you come back to that point, you realize it’s feasible to cope with it, because we’ve been able to cope with it for many years.
BERGER: Technology is evolving so fast, it’s hard to foresee. Do you know the saying, ‘The appetite comes with eating’? It’s like when you’re working on a project, you’re finding new data insights, new data points that give you the motivation to go to the next step. So I am convinced there will be so many more use cases in the future that I cannot foresee right now.
I will learn, we all will learn, the machine will learn. We will get more and more data created out of the data that we already have – other data sources, third-party data and so forth. So right now, I cannot foresee all the use cases we will see in the future. We will work under one paradigm, which is ‘Intelligence in each transaction’. Over time alfred will also take decisions in our ERP system automatically. In average transactions that we do, we would like to have more intelligence, and alfred will help us with that.
TRANSFORM: Is there anything you would like to add?
BERGER: I’m a great believer in removing the mystique of buzzwords like AI and focusing on what’s behind it instead – helping people and companies understand the technologies and methods that help us make our businesses as well as our personal lives easier and better.
It’s part of my role as the CDO, but I also believe that digitalization is a bunch of buzzwords. If you ask someone at a conference what you really mean by digitalization, most people will get very thin in their answers. Why?
Because they don’t really know, because they are looking at digitalization from a huge height. And I think if you really want to go beyond the buzzwords, you really need to go into the use cases and the business, and you really need to redefine the opportunities. So I am trying hard to get out of these buzzwords and really get down to the use cases.
Top photo: thyssenkrupp Materials Services receives around 14 million order items annually. With alfred, these can be efficiently processed and analyzed. (All photos courtesy of thyssenkrupp Materials Services)
The post thyssenkrupp Materials Services ‘keeps calm and carries on’ – with its new ‘alfred’ AI solution to optimize its logistics network appeared first on Microsoft Malaysia News Center.