German Mittelstand meets Artificial Intelligence – Basics for Entrepreneurs

Thursday, 11. April 2019

A guest article by Peter Renner
 

Intelligent algorithms are revolutionizing our everyday lives – both personally and professionally. They learn with the help of data and are able to depict complex issues much better than humans in a short period of time, thanks to their structure. Companies dealing with this technology at an early stage will gain fundamental advantages over their competitors.

Algorithms and Artificial Intelligence (AI)


An algorithm is a kind of step-by-step guide that can be used to solve a particular problem or accomplish a particular task. Algorithms are usually written by hand and follow rigid rules, which means that they can only adapt to new events to a limited extent. If the requirements change, the rules have to be rewritten.

However, it is impossible to fully map all rules in a complex and dynamic environment. This is where intelligent algorithms are developed that learn by themselves and can process much more complex issues than humans could ever reproduce through individual rules. Talking about AI, we think of algorithms that do not follow rigid structures and rules, but learn them by themselves.

Artificial neuronal Networks


Artificial neuronal NetworksThe human brain consists of countless neurons that absorb information from the environment and process it in so-called neuronal networks. In response to this information, they for example send commands to our muscles. Humans learn by repeating these processes. For example, the more often a golfer performs his swing, the more intuitively he hits the ball and like this, the neural network is trained to do a golf swing.

An AI is basically an algorithm that tries to imitate the human brain with the help of artificial neural networks. The algorithm is trained by providing it with information in the form of data, which it can then use to form deductions. This training process is called Machine Learning. The more data is available to the algorithm in this training process, the more reliable it will function. Once the training process is completed, we speak of an Artificial Intelligence.

Machine Learning


An AI starts its development as an untrained model, meaning an untrained algorithm. At the beginning, this model is fed with training data sets. With the help of feedback loops, the self-learning algorithm checks its results during training and adjusts itself accordingly. The trained model is then used as an AI. With every additional data input, the AI can improve its results. However, the algorithm behind it does not change anymore.

An AI, however, only finds solutions on the basis of the data input provided and does not develop any creativity or awareness of its own. One therefore speaks of a weak AI. It is specialized in one field and only serves the purpose for which it was programmed. We already find it all over our everyday life, for example in chatbots, speech assistants, navigation systems and many other applications.

AI in Practice


AI in PracticeAI can therefore be used wherever there is data, which it can learn from. Not only does it find its way into our personal, but also into our professional lives. It makes company processes more efficient, supports and optimizes market development, creates completely new business models and much more.

Especially in the areas of marketing, sales and customer care, AI solutions can offer quick and tangible benefits. The energy group RWE, for example, uses AI for the recording and processing of its customer correspondence. The AI recognizes patterns in texts (e. g. letters or e-mails) and can thus analyze the customer's concerns precisely and control further processing. This way, 80 percent of incoming service requests are automatically transferred to the systems by the AI1.

However, the use of AI solutions will also be indispensable in many other areas in the future. In retail and logistics, for example, AI already optimizes pricing, purchasing planning, shipping, logistics, sales forecasts, supply chain and many more. In production, AI can be used to predict failure probabilities or optimize machine utilization, among other things. In purchasing, widely standardized processes can be implemented more efficiently. Solutions, for example, include the integration of voice control and digital assistants.

AI also provides support in management and administration. Through extensive analyzing of digitized meta- and company data, AI can for example provide fact-based decision support for the company’s management. Similar to the diagnostics in medicine, AI can derive recommendations for investments. Particularly in administration, tasks often follow fixed rules and can easily be taken over by AI – e. g. by automatically entering and processing receipts and invoice information or by reconciling account movements with incoming and outgoing invoices.

Implementation of AI solutions


However, many companies are reluctant to adopt AI solutions because they don't understand them and certainly don't know how to implement them. It is therefore recommended to choose a pragmatic way to begin with AI. Start with small projects and focus on a specific application case. It is almost impossible to fully equip a company with AI solutions overnight. When selecting the first application case, always ask yourself what concrete benefit the AI should bring, where the benefit is going to be most visible and where it can be introduced most easily.

Meanwhile, there are catchy software solutions and partners for almost all AI applications. Use these to gain first experience with AI applications and to build up an understanding of AI and its potential. Afterwards, AI competencies can slowly also be built up internally.

Especially for small and medium enterprises (SMEs), working with AI solutions is still in its early stages. Thus, customers and employees will probably still forgive minor misadventures during implementation today. In the mid and long term, you can secure fundamental advantages over your competitors. The credo therefore is: get started!

Source


1Bitkom: Künstliche Intelligenz Gipfelpapier (2017)

Credit (from top):

© istock.com | monsitj
© istock.com | ktsimage
© istock.com | Pinkypills
© WeissmanGruppe



About the Author

Peter RennerPeter Renner is a consultant to Weissman & Cie. GmbH & Co. KG, a business consultancy specialising in family businesses. In addition to strategy development and implementation, his main areas of activity are digitalisation and innovation in medium-sized businesses.

Please find further information on www.weissman.de.