Moving Into The Future With Artificial Intelligence

Why AI belongs on the digitalisation agenda of every company

Tech // Nov 26, 2020
Csm Tallence Website Header KI 2 blau 54326d5605

Artificial Intelligence (AI), a subject area which has become an integral part of the technology landscape and which companies are increasingly considering in their digitalisation initiatives. In fact, however, this is not reflected in the spread of AI-supported solutions in Germany. According to their own statements, only 6% of the companies rely on this type of technology [1].

The reasons for this are varied, but it becomes clear that companies are unable to identify possible use cases for AI technologies or assess the potential for their business activities. Other companies simply lack the know-how to implement intelligent solutions or a suitable data basis.
We at Tallence see an enormous potential in the use of Artificial Intelligence in everyday business life to relieve people in the performance of their work and to make more qualified decisions through the efficient evaluation of large amounts of data. In this way, AI can be used to make more informed strategic and operational decisions, resources can be used more efficiently, and customers can be better served and addressed more individually. In the long term, this can make a significant contribution to reducing costs and increasing sales. With this article we want to give an overview of the field of Artificial Intelligence and provide use cases to illustrate the added value and the various fields of application. In our opinion, companies should actively approach the field of Artificial Intelligence in order to be optimally prepared for future challenges and to generate long-term competitive advantages.

What is behind the term "Artificial Intelligence"?

Generally, Artificial Intelligence pursues the goal of solving problems and tasks that cannot be solved with conventional, rule-based algorithms or can only be solved with considerable effort. The field of Artificial Intelligence has been actively researched since the 1950s. However, it has experienced a real boom, especially in the last decades. Apart from important progress in research, this can be attributed in particular to the sharp increase in the computing power of computers and the data which is available today in large quantities and in a wide variety of forms. The enormous improvements in these areas open up completely new possibilities and fields of application for Artificial Intelligence today, which would have been impossible years before.
Today, AI algorithms are successfully used in the medical field, e.g. for cancer diagnosis or in the intelligent maintenance of aircraft, keyword: Predictive Maintenance (will be explained in more detail later). Both areas require a high degree of precision and quality, which must be achieved by AI before value can be achieved through its use.

Machine learning solutions, which form a sub-discipline of Artificial Intelligence, are largely responsible for applications in such demanding and various other areas. They enable machines to learn new content based on data within a specific task field and to transfer this knowledge to other situations. The process of machine learning can be compared very well with the human learning process. Human beings learn from the experiences they make in the course of their lives. They use the newly gained knowledge to apply it to other life situations. In Artificial Intelligence, data replace human experience. In contrast to conventional programmes, a machine learning solution is not ready for use without prior training. Such training can take various forms. Different solutions have different requirements on the training data set. For supervised learning, the training data set must contain examples with known results. In unsupervised learning, the programme can independently identify the structure of the training data according to certain criteria. In reinforcement learning, the programme is continuously improved during real use by evaluating the results of its actions. It applies equally to all the above-mentioned constructs that the training determines the quality and accuracy of the machine learning solution. The better the data basis, the better the expected results. It is often true that more and more varied data improve the training effect. There are, however, procedures that achieve a high degree of precision with only a small amount of medium-quality data. These approaches build on a model that has already been pre-trained with a basic stock of data. Chatbot solutions are a good example of this. Often basic information such as sentence structures or vocabulary is already conveyed to the chatbot through training. So for chatbots, a few sample data sets are often sufficient to capture certain contexts within the field of application.
One sub-discipline of machine learning is particularly interesting and associated with great opportunities for the future: deep learning. This sub-discipline describes Artificial Intelligence, which makes use of deep, multi-layered neural networks. These algorithms are - on a simplified basis - modelled based on the processes of the human brain. They benefit greatly from large amounts of data. Thanks to advances in computing power and data availability, these algorithms have made significant progress in recent years. When trained in this way, they deliver results of a quality that was previously unattained. Deep learning solutions have been particularly successful in the analysis of poorly structured data such as speech, text, images and video. In this way, deep learning algorithms are already providing support today for many complex problems such as the diagnosis of tumours by analysing MRI images.

Artificial Intelligence is generally divided into three classes that relate it to human intelligence. The AI solutions and application areas already mentioned can be assigned to the area of "Weak AI". Solutions in this area perform specified tasks in a defined area. They are adapted for these specific applications and trained with the necessary data. "Strong AI", however, acts on an equal level with human intelligence, even for difficult tasks. Its field of application is not limited to individual areas. We refer to "Super AI" when Artificial Intelligence is superior to humans in many or even all aspects. Even if, according to current expert opinion, it will still take some time before the latter two forms of intelligence are realised, the potentials and risks are already being discussed controversially (reference: Open Letter on Artificial Intelligence, Steven Hawking).

Grafische Darstellung von Man und Frau am arbeiten

Valuable applications of AI in everyday business.

The vast number of different applications of "Weak AI" already offers great potential for companies across all sectors. Individual tasks or operational processes can be intelligently supported and automated in order to increase efficiency and build up long-term competitive advantages. The diversity of the application areas of machine learning shall be illustrated by the following examples:

  • Forecast of sales volumes: Time series analyses are used to evaluate historical data to make predictions about the future. Machine learning models (e.g. recurrent neural networks) extend classical time series analyses. They are trained using historical sales figures and other variables such as sales regions or product characteristics and can then make predictions about future sales. These results offer a broader basis and, combined with the experience of experts, can lead to more qualified strategic and operational decisions.
  • Automated processing of support requests: The customer support of a company can be greatly relieved and improved by using machine learning-based Natural Language Processing (NLP) and Understanding (NLU). These algorithms analyse and process natural language up to an understanding of the content. In this way, for example, a chatbot can answer customer requests on defined topics quickly and automatically or, if necessary, refer them to appropriate customer service staff. In addition, customer e-mails can be evaluated and categorised with regard to e.g. topic or emotional situation. In this way, service staff already have relevant information available for processing the request.
  • Personalisation and automated tagging of web offers: Recommender systems are used by almost all web platforms such as Amazon. They can be used to evaluate user behaviour across a web offer and, based on this information, to recommend personalised products or services to users. Machine learning methods can also support the generation of metadata by extracting information from images and texts from the web offer. This data enrichment can be used to identify further relationships between the individual web objects and between them and the users. In a concrete example, an online retailer for shoes can automatically tag his products in this way. A typical scenario here would be a categorisation by colour or style. This not only relieves the human being of a time-consuming, manual activity, but also enables a much more granular categorisation.
  • Intelligent maintenance in the form of predictive maintenance: The aim of this field is the proactive maintenance of machines, equipment and plants. By analysing historical data, it is attempted to identify patterns which allow conclusions to be drawn as to when system failures occur, and which components require maintenance. In this way downtimes can be minimised and overall efficiency and quality can be increased. A very good example of the advantages of using predictive maintenance is the maintenance of lifts. In this area, high reliability is particularly important. Unplanned downtimes cause costs and cause dissatisfaction, because nobody wants to get stuck in the lift. With the help of sensors installed in the lift, important operating and maintenance data can be collected and evaluated. This enables early detection of problems and thus prevents imminent failures.

The examples mentioned make clear that Artificial Intelligence can relieve people in their work and, moreover, in certain areas, can even help them to make more qualified decisions. It can relieve employees of repetitive, manual tasks and thus give them more time for creative and complex activities. In addition, AI enables the analysis of large multidimensional and multivariate data sets which, due to their complexity, are difficult for people to consolidate and evaluate. Another factor is the acceleration of work and decision-making processes. This extends to automatic real-time evaluation and processing of incoming data. In this way, an enormous improvement in quality and efficiency can be achieved. Tallence therefore does not see AI as a substitute for skilled labour, but rather as a driver of innovation and a meaningful addition to the human workforce as we know it today. In our view, it is the balanced combination of intelligent assistants and skilled workers that enables companies to generate maximum value from Artificial Intelligence. At Tallence, we are convinced that innovation can improve long-term competitiveness and thus lay the foundation for future growth.

In past projects, we at Tallence have already successfully implemented AI solutions for our customers. Among other things, we have developed a chatbot that uses information from the client's existing interfaces to answer users' questions in an intelligent way. In another project, we integrated an e-mail classification solution into an existing e-mail infrastructure.
In our next blog post, you will learn how we approach projects in the field of Artificial Intelligence and how we support you as a competent partner from the idea to an intelligent solution.

Article by: Dr. Bastian Eggert, Timo Fuhrmann and Dr. Kai Matzutt