15th of April 2020


AI in Steel Industry and Warehousing

Manufacturing vs. Tech

Artificial Intelligence is widely used in online B2C services such as Amazon, Netflix and social media. Even though the foundation of today's most disruptive technology was laid more than half a century ago, it took even the earliest adopters - the tech giants - till the end of the last decade to bring AI into production. The key enabler of AI is a strong data foundation to train models on, and so it is not surprising that internet companies, which by definition are data collectors, were the first to adapt this technology. Whereas private user data is comparably cheap, collecting data in other industries tends to be more complicated. So, it is not surprising it took the manufacturing and industry sector years to build up the right data foundation which was established in tech basically overnight. We see a shift of approx. 6 to 10 years for many industries (especially manufacturing) in their adoption of data-driven technologies such as Machine Learning and AI as compared to IT and tech corporations. These days, we see early adopters in manufacturing and industry getting their feet wet by developing prototypes where they can gain value from this new technology.

As the industry made progress in data collection processes, companies were able to gain value by using this data for Business Intelligence. BI and visualization tools provide rich information to the experts working in warehouses and in the shop floor. Still, companies are heavily reliant on industry experts to evaluate information, make decisions upon multiple factors and monitor processes, just to name a few of their tasks. It is time to move to the next level with tools that provide not only information but also directly help operators and experts in decision-making by using AI in manufacturing and warehouses.

Specific Requirements for AI in Manufacturing and Industry

After years of successful delivery for the industry, it is clear to us that delivering solutions tailored to the needs of the companies is the way to go, instead of trying to adjust internal processes to existing off-the-shelf AI tools. We see that solutions are very specific to each industry and application area. Also, because of the data available to train the algorithms most solutions are unique. We worked together with voestalpine solving a very specific business problem in warehouse optimization for the steel industry.

Not only is there no tool for smart warehousing with this particular business problem available on the market, but a tool would also have to be adapted to the user's specific data collection processes and infrastructure.

AI Pioneers in the Steel Industry

While warehouse optimization for private consumer goods is still a fairly generic problem, the storage of heavy and sometimes massive parts in the steel industry for business customers is a niche application. With voestalpine, we developed an AI Solution to support decision makers with a complex forecasting and recommendation tool for multi-warehouse stock optimization. This tool represents a cornerstone on which to build further AI applications within the company. voestalpine is therefore ideally equipped to further automate its warehousing and to cost-effectively optimize its stock.

Crayon's AI CoE Vienna was engaged in the development of the tool to optimize multiple stocks in parallel for the client. A close collaboration was key to develop a satisfying solution to meet the requirements of the user, as well as delivering excellent results from a technical and from an expert perspective. It took a team of data scientists, data engineers and consultants from Crayons CoE working closely together with the client's industry and IT experts to develop the product.


In the future, we will see AI playing a key role in manufacturing and warehousing. Imagine robots loading or unloading pallets, bringing items to the stock or searching pieces on stock. Entire production plants connected via IoT solutions and optimized production planning and decision-making using machine learning. Fleet, route and travel time optimization, demand forecasting and more. The list is potentially infinite, since for every need satisfied by a smart solution there will arise a need for a new solution. For some players like Amazon and Alibaba this is partly reality already. The AI journey for most businesses starts with a very specific use case until a level of confidence is reached to reorganize entire processes with the goal, or at least the possibility, to operate fully automated.