Be it in quality assurance in production environments, with chat bots, in customer service or in predictive maintenance of machines and systems: Companies in Germany are now using machine learning (ML) and its sister technology, artificial intelligence (AI), in many areas or have at least started corresponding field tests. This is shown by the study “Machine Learning 2020” by IDG Research Services.
But not only has the variety of “use cases” increased compared to previous years, but also the number of companies that actively use machine learning. “Most companies in Germany have recognized the relevance of machine learning and are increasingly using this technology,” confirms Dr. Susan Wegner, Vice President Artificial Intelligence & Data Analytics at the IT service company Lufthansa Industry Solutions (LHIND). “At the moment, this is particularly true in the financial sector, in order to use machine learning to optimize capital flow analyzes for forecasts.” LHIND is also increasingly registering inquiries in the cargo area. There, ML can be used to predict how many deliveries will be made in a certain time window. Companies in turn use this data to plan the use of labor and optimize pricing.
To the study “Machine Learning 2020”
According to the IDG Research Services study, almost three quarters of companies are currently dealing with ML. It is positive that this does not only apply to large companies, says Dr. Reza Bakhtiari, Senior Data Scientist and Consultant at A1 Digital Germany: “Of course, large corporations tend to be more willing to invest in new technologies. But we see that increasingly smaller companies are also recognizing the added value of ML.” However, from his point of view, speed is of the essence. If you don’t want to deal with machine learning and AI for a few years, you run the risk of losing touch.
But after all, around 30 percent of medium-sized companies with 500 to 1,000 employees already use several ML technologies. For large companies it is more than 43 percent. On the other hand, smaller medium-sized companies with 50 to under 500 employees still have a lot of catching up to do: Only ten percent are already using machine learning solutions on a large scale. According to Reza Bakhtiari, the technology can be used profitably in many sectors of companies of all sizes, “from logistics and retail to construction companies and service providers”.
The fields of application that are best suited for initial applications include personalized customer service and a “360 degree customer view”, says Susan Wegner from LHIND. The reason: “The marketing departments often have a strong affinity for evaluating data. In addition, data in this area is often available in the required quality.” The participants in the IDG Research Services study also share this view. Around 38 percent see the greatest potential of machine learning and AI in customer service, around 37 percent in sales and almost a quarter each in marketing and improving the customer experience.
But ML projects are not a “sure-fire success” even in these sectors – an experience that both LHIND and A1 Digital have made: Many machine learning projects get stuck in the proof-of-concept phase. There are mutliple reasons for this. “Many companies are not yet able to use functioning ML algorithms productively because other tools and skills are required,” says Reza Bakhtiari from A1 Digital. In addition, according to the experience of Lufthansa Industry Solutions, many users neglect basic requirements, such as “the ML development process, the database, the technology platform and the subject of data governance”, says Susan Wegner. This means that users of machine learning and AI have to do a lot of homework before they start with a corresponding project. A central point is above all the amount and quality of the data that are required for training the corresponding algorithms. “Machine learning stands and falls with the quality of the data. That is why a lot of time, expertise and money have to be invested in the acquisition and processing of data,” says Reza Bakhtiari.
Such preparatory work is made more difficult by the existing IT infrastructure: “Large, traditional companies in particular are faced with the challenge that, for example, user data is distributed across different legacy systems, not all of them are directly linked to a user or even have different user IDs,” explains Susan Wegner from LHIND. In addition, the statutory data protection requirements must be taken into account, especially for personal data. This can result in a company having to implement different solutions depending on the type of data. Against this background, it is not surprising that in the study by IDG Research Services, the item “data protection requirements” ranks second with 35 percent in the ranking list with the greatest hurdles in the use of ML.
To the study “Machine Learning 2020”
But such challenges can be overcome. For example, LHIND recommends introducing a clear governance model. It regulates who is the producer and owner of certain data and who takes on the role of the system administrator. In order to dispel resistance and reservations on the part of employees towards machine learning and AI, the support of corresponding projects by the management and the division heads is also indispensable. Comprehensive information for employees and regular training courses also provide assistance.
Companies involved in machine learning should also implement tools that enable close collaboration between departments and machine learning experts, advises Dr. Bakhtiari from A1 Digital. In addition, it is necessary that workflows are automated so that ML models can be used in productive use. This means that it is not enough to simply hire a few machine learning experts and provide them with ML frameworks, data and the corresponding IT hardware. Obviously, companies have to lend a hand in other areas, for example by first recording their processes – keyword process mining – and automating them. This is where another technology comes into play with Robotic Process Automation (RPA). It is also currently very popular with German companies, as confirmed by another study by IDG Research Services (“Robotic Process Automation 2020”).
In order to get started with machine learning faster and more efficiently, cooperation with IT service providers can also be considered. A considerable part of the user companies also see it this way: only about a third forego consulting external experts. The majority fall back on external support if necessary or have outsourced the development and operation of ML solutions. “We particularly come into play when sustainable machine learning solutions are sought and not just a specific product,” explains Susan Wegner. LHIND concentrates on business-relevant solutions for which operation and possible further development are secured for a long time – keyword future security.
A1 Digital, on the other hand, provides a complete package, “an ML-as-a-Service and Infrastructure-as-a-Service offering in combination with support from our data science experts,” says Reza Bakhtiari. “The user can thus focus on his technical expertise and the ‘business insights’ without losing time and resources for infrastructure and the construction and operation of a tool landscape for machine learning.”
In summary, it can be said that it is gratifying that a significant number of German companies already have machine learning on their agenda. This should also reflect the assessment of managing directors, specialist departments and CIOs that ultimately there is no getting around this technology. Anyone who now fails to gain experience with ML and to develop application scenarios will find it difficult to survive in the market. This applies to most of the industries.
Therefore, users should not be put off by the hurdles that have to be overcome when introducing a new technology. But there are always such start-up problems when new approaches are established. Just think of virtualization and cloud computing. It is better to see machine learning and AI as an opportunity to secure your own economic success and to open up new business areas.
To the study “Machine Learning 2020”
Editor: COMPUTERWOCHE, CIO, TecChannel and ChannelPartner
Gold partner: Lufthansa Industry Solutions
Silver partner: A1 digital
Population: Top (IT) responsible persons for companies in the D-A-CH region: strategic (IT) decision-makers in the C-level area and in the departments (LoBs), IT decision-makers and IT specialists from the IT area
Participant generation: Sampling in the IT decision maker database of IDG Business Media; personal e-mail invitations to the survey
Total sample: 406 completed and qualified interviews
Investigation period: February 17th to 25th, 2020
Method: Online survey (CAWI)
Questionnaire development: IDG Research Services in coordination with the study partners
Execution: IDG Research Services