Advanced Analytics and Artificial Intelligence

Author: Niels Pothmann, Head of AI at Arvato Systems

Data has long been considered to be the new gold. But merely collecting and storing vast amounts of data is not enough. To profit from their data stock, companies must consolidate and evaluate their information inventory to derive and systematize the right decisions from the knowledge gained. In this context, one speaks of advanced analytics and Artificial Intelligence (AI).

Business Intelligence (BI) is considered the basis or forerunner of Advanced Analytics. BI is a method with which processes can be systematically analyzed. With the help of BI solutions, users can collect data from different sources, consolidate it in a central location, reliably evaluate it, and visualize it precisely. In this way, they create a solid foundation for making strategic and operational decisions more effectively. Many companies use self-service BI to create their reports and perform individual analyses. The advantage - this does not require any in-depth technological knowledge. The main difference to Advanced Analytics is that BI tools only provide analyses related to the current or past situations. With Advanced Analytics tools, on the other hand, analysis models or simulations can be built. Based on the current situation's available data, companies can then set up data-based forecasts in the next step. With the help of data science and predictive analytics methods, they can ultimately answer this critical question: How are certain company key figures, such as demand, likely to develop in the near future? On this basis, optimized planning can then be drawn up.

Making processes more efficient

A correspondingly large stock of relevant information is required to make data-based decisions - which many companies either do not (yet) have or cannot use profitably. In many places, not all of the required data is available in digital form. That's why companies need to make the necessary data digitally available and provide centralized access, for example, through a modern data infrastructure and cloud-based services. By using powerful modern technologies from the fields of Artificial Intelligence (AI) and Machine Learning (ML), it is then possible to identify complex dependencies and patterns in a provided database. Based on Big Data, such technologies can provide even very granular insights into processes using large amounts of data. Once the patterns have been identified, processes can be mapped automatically or (partially) autonomously by AI systems in subsequent steps. In the media sector, for example, editors use AI to make their research work more efficient. Natural Language Processing (NLP) is essential for this. That is a technology with which natural language can be processed automatically. By resorting to NLP, an AI-based system can extract relevant information from texts and recognize contexts. Journalists can then use the information obtained to conduct in-depth research.

Must have: interdisciplinary project teams

In addition to a consolidated data stock, there is another indispensable prerequisite for the successful use of data potential: interdisciplinary project teams with technically versed experts and technically competent staff. Experts from the respective specialist department specify the content and practical orientation of the solution to be developed. On the technical side, a Data Scientist or Machine Learning Engineer acts as a bridge between technical competence and functional orientation. He must understand the task at hand and transfer it into automated procedures. It is also essential to have a data engineer or data architect who consolidates the required data and provides high-performance access options.

Machine Learning Engineers are becoming increasingly important.

Machine Learning Engineers and Data Scientists play a central role in integrating many different parts and competencies across departments. In addition to professional experts, there will be more and more so-called Citizen Data Scientists, i.e., users with a basic understanding of mathematics and statistics and a keen interest in analytics. With the help of the right tools, they will take over-analytical tasks for operational process support themselves. Such role profiles are being created because access to data science and AI is becoming increasingly easy. Besides, intelligent data processing processes are already a component of many industry solutions.

Using new technologies profitably

Therefore, technologies such as Artificial Intelligence and Machine Learning can lead to solutions that provide valuable support for companies. One example is predictive maintenance applications, which are increasingly being used in maintaining industrial machinery and technical equipment. Thanks to AI, such applications monitor hundreds of sensors at short cyclic intervals, sometimes in the range of a few milliseconds. An AI can detect unusual behavior, so-called anomalies, and wear and tear signs at an early stage from a large amount of information. The system works with pre-trained and self-adaptive models to make swift decisions despite the high complexity and data load. Based on these models, maintenance windows can be planned at an early stage, downtimes can be reduced, and at the same time, the operating time and replacement cycles of technical components can be optimized.

Building up own competencies is very complex

Against this background, many companies ask themselves: Is it worthwhile to hire your own data experts? There is no overall answer to this question. In the environment of data science and AI, different role profiles are increasingly emerging that relate to individual aspects of a solution. It takes a great deal of effort to build up and continuously expand these skills yourself. Whether the effort is worth, it depends on the structure and setup of the company. As an alternative to a comprehensive competence build-up, it is recommended to fall back on specialized service providers. The right service provider should respond to each company's requirements by first analyzing the current situation. If there are already first approaches, he transfers them into practical application. That also includes networking data sources between the specialist departments and contributing specialist know-how for Machine Learning and Artificial Intelligence. For companies that start from scratch, a service provider provides AI-based microservices for individual departments' unique requirements. Regardless of which task the service provider takes on, it must be oriented to the firms existing data structure and processes. And since AI solutions are continually evolving, it is always advisable to focus on long-term collaboration.

If you have questions, please feel free to contact Niels Pothmann, Head of AI at Arvato Systems.