Most people may already have come into contact with machine learning (ML). Appropriate algorithms, for example, use solutions for speech and image recognition. Voice assistants, chat bots and recommendation engines from e-commerce platforms would therefore not work without machine learning. ML also plays an important role in the analysis of processes and their optimization. That is why researchers and, of course, the providers of corresponding solutions classify machine learning as future technology.
This finding has obviously also reached users in Germany, as the “Machine Learning 2020” study by IDG Research Services in cooperation with Lufthansa Industry Solutions and A1 Digital shows. Around 73 percent of companies have started ML projects. Compared to the study from 2019, that’s almost 20 percent more. Larger companies with 1,000 or more employees in particular have taken on a pioneering role: Around 40 percent of them are already using a whole range of machine learning solutions. This is not particularly surprising. Because large companies usually have the money, keyword IT budget, and the necessary human resources to start such projects. However, SMEs have also recognized the value of machine learning. Almost 30 percent of the companies use the technology in several areas, and almost as many use the first machine learning solutions.
From a user perspective, it is gratifying that machine learning projects do not need a long lead time to bring countable benefits, according to the study by IDG Research. Around 71 percent of the companies registered positive effects after three months at the latest, and 22 percent of those questioned immediately after using ML-Tools. There are two main criteria for the success of a project:
• an increase in employee productivity (52 percent) as well
• reducing costs (49 percent).
Factors such as higher sales and better quality of products and services (around 31 percent) follow at a clear distance. Strategic goals, such as a higher degree of innovation or the development of new types of offers, rank lower.
To the study “Machine Learning 2020”
One possible explanation for these results is that companies initially focus on collecting the “low-hanging fruits”, ie achieving quick results. This is understandable. Excessive expectations on the part of management and specialist departments often lead to disappointment quickly spreading when a new technology is not supposed to bring the hoped-for benefits.
The providers of artificial intelligence (AI) and machine learning solutions as well as consulting companies therefore recommend that “expectation management” be introduced. It is intended to ensure that all project participants, from the IT department to the specialist departments and the management, are, as it were, on the carpet. In other words, the parties involved define realistic goals that are to be achieved using machine learning. The same applies to the implementation in practice: a policy of small steps is better than trying to immediately achieve resounding success with overly ambitious projects.
However, this does not mean that companies should limit themselves to using machine learning to optimize existing processes. According to the study, this is the most important goal of ML projects for 56 percent of those surveyed. But almost half (45 percent) also see machine learning as a tool for developing new business models. Around 44 percent want to use the technology to expand the range of their products and services. Digital offers are likely to play a central role in this.
However, there is reason to think that this strategic view of machine learning is particularly pronounced in large companies with 1,000 employees and more. SMEs and smaller companies have a lot of catching up to do on this point. An example: Only a good 29 percent of medium-sized companies see ML as a means of developing new offers, whereas 55 percent of large companies. This discrepancy is problematic for two reasons: the digitization pressure that companies of all sizes face and the fact that small and medium-sized businesses are still the backbone of the German economy. A rethink is urgently required for medium-sized companies.
Speaking of the strategic importance of machine learning: One result of the study indicates that managing directors and other business decision-makers increasingly assume central responsibility for such projects. They have already taken over management in 37 percent of the companies; Around 45 percent of the companies have IT and technology managers, such as the Chief Information Officer (CIO), the Chief Digital Officer (CDO) or the Chief Technology Officer (CTO). However, compared to the study from 2019, the proportion of “technicians” fell by five percent. This is an indication that managers recognize the importance of machine learning, for example as a component of digitization.
The departments, however, play only a subordinate role in ML projects. Only in 15 percent of the companies do they have the central responsibility for such projects. It is all the more important that companies ensure a comprehensive exchange of information and experience across departments. Because the requirements of the specialist departments for a machine learning solution should be taken into account. After all, it is their employees who ultimately work with such tools.
To the study “Machine Learning 2020”
But users don’t just have to consider the communication channels or the quality of a machine learning model. The biggest problem is the lack of general know-how (39 percent) related to machine learning. But too strict data protection requirements (35 percent) and the insufficient quality of the input data (22 percent) are cited as stumbling blocks. There are also “soft factors” such as an inappropriate corporate culture (22 percent).
It is remarkable – and worrying – that the departments come to very different assessments as to the greatest inhibitions. For example, managing directors see data protection and compliance requirements as a problem (51 percent), while IT specialists see a lack of knowledge in the areas of programming (28 percent) and statistics (26 percent). The departments in turn put the lack of know-how (39 percent) and the poor quality of the databases (32 percent) into the field. These discrepancies indicate that the individual areas only look at machine learning from the perspective of their department. An understanding of the overall context is obviously underdeveloped.
Such deficits can be remedied with the help of project teams in which members of the various departments and departments work together. Such interdisciplinary working groups can help to avoid the formation of information silos. If necessary, external consultants and IT service providers provide assistance by assuming the role of a neutral specialist body.
But IT service providers are not only in demand as mediators. Rather, they bring in the expertise and machine learning specialists that users often do not have. That should be the reason why two thirds of the companies in the ML area work together with external service providers. The majority of them even use two to five ML and AI specialists.
However, this does not mean that users give up sovereignty over such projects. Slightly more than half remain in charge and access external help when necessary. Only 15 percent have completely outsourced machine learning to external companies. This result of the study is also understandable. Because for the majority of the companies surveyed, it is important to develop their own know-how in a promising field such as ML.
The bottom line is that the study by IDG Research Services paints a picture that gives hope. German companies have apparently recognized the importance of machine learning and neighboring disciplines such as artificial intelligence and deep learning. This proves the number of companies that are already using ML, and with success. However, these positive signals should not hide the fact that there is still a lot to be done. Interestingly, this does not only apply to areas such as the development of know-how and the optimization of databases that machine learning needs.
An important point concerns communication and cooperation between the departments and management levels. Traditional hierarchies and information silos are clearly proving to be inhibitions when it comes to implementing machine learning projects. Companies shouldn’t just take a look at the technical challenges associated with using machine learning. It is also important to put the corporate culture and internal communication structures to the test.
To the study “Machine Learning 2020”
Editor: COMPUTERWOCHE, CIO, TecChannel and ChannelPartner
Gold partner: Lufthansa Indutry Solutions
Silver partner: A1 digital
Population: Top (IT) managers of companies in the D-A-CH region: strategic (IT) decision-makers in the C-level area and in the specialist areas (LoBs), IT decision-makers and IT specialists from the IT area
Generation of participants: Sampling in the IT decision-maker database of IDG Business Media; personal email invitations to the survey
Overall 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