The use of AI has not yet really arrived in the German manufacturing industry. According to a PwC study based on a survey of 500 decision-makers in German companies, only 4% use AI, 2% are in the process of implementing corresponding technologies and 17% have plans to do so. Around half of those surveyed do not consider the topic to be relevant.

If AI is in use or conceivable, it is primarily in the direction of data analysis for decision-making processes (70%) or for the automation of existing business processes (63%). Companies are primarily concerned with supporting and relieving employees - according to 71% of the companies surveyed. Autonomous systems play a rather subordinate role.
According to another study by PwC, the fundamental question for companies is: what is the best way to introduce AI applications in the company? A key aspect here: How can an AI culture be established in the company and how can employees' trust in the technology be built up accordingly?

The basics

"Against this backdrop, there are various conceivable ways of gradually integrating intelligent systems into everyday industrial life," says Dirk Löhmann, Regional Vice President Continental Europe, Epicor.
"In our experience, two steps are important for the manufacturing industry in the course of ERP projects in order to promote the organic development of AI and successfully manage this change: firstly, a central software for company management - in our case the ERP system - is required, which is characterized by a platform-oriented approach and a service-oriented software architecture. This is the prerequisite for flexibly integrating AI services as cloud applications, such as those available via Microsoft Azure, as required.
Secondly, companies need to overcome their fear of artificial intelligence, which is usually associated with autonomous robots. In reality, AI functionalities can be - and are already being - integrated very intuitively into everyday processes and decisions in a form that is perceived as convenient IT technology rather than artificial intelligence. Voice control or trend and pattern recognition at the touch of a button are typical examples of this."

Virtual agent as a first step

Epicor's virtual agent, EVA for short, is an example of such an approach. The virtual agent was developed using Microsoft Azure AI services and can be used in all Epicor ERP environments - whether in the cloud or in your own data center - regardless of size. This means that small and medium-sized companies can also benefit from AI.

Dirk Löhmann, Epicor

© Epicor

Used for mobile applications, the agent appears on the screen as a virtual assistant that users can access via text and, in particular, voice input. Using Natural Language Processing (NLP) - one of the many forms of AI technologies - EVA provides targeted information to make better decisions faster. In addition, the virtual agent uses AI capabilities to proactively alert and take specific actions based on a combination of events, market statistics and historical data. It is designed to use AI in an industry-oriented way in day-to-day work and to pave the way for the expanded use of AI and cognitive technologies.

From voice control to mixed reality

Subsequent versions of EVA are intended to promote this development - for example with other AI-supported Microsoft Azure technologies such as Hololens 2 with mixed-reality applications. Hololens 2 is geared towards industrial augmented reality applications. In conjunction with the virtual, voice-controlled agent EVA, both hands remain free when requesting information, and the view is always focused on the work area.

Other future AI applications could include

- Machine learning-based warehouse optimization in production and manufacturing;

- Forecasts of system failures or predictive maintenance for production systems in combination with machine learning;

- Machine learning real-time analyses of buyer behaviour for optimized production and distribution;

- IoT- and machine learning-based predictive quality assurance in production using sensor data, possibly in combination with image processing

- Machine learning-based cost calculation in production.

For example, the virtual agent could continuously analyze sensor data in production. As soon as anomalies occur in robot production, the AI application sends alarm messages to the production manager's mobile device, who then clicks on the 'Maintenance planning' button and EVA schedules a work order for this. Next, the virtual agent suggests on which other production line the current production order can be carried out and takes appropriate action after confirmation.

Flexible AI options

Using AI services, such as those available via Microsoft Azure, as an integrated part of industry-specific ERP systems opens up ways to implement intelligent functions tailored to individual business requirements. For example, Sistema - a manufacturer of plastic storage containers - took the path of creating a digitized 'closed-loop' in real time between business systems and production (MES). This integrated ERP and MES platform is the basis for further developments towards predictive analytics, IoT networking via sensors on the production machines and business intelligence across the entire organization.

An approach that Bitkom also pursues in its position paper for future-oriented ERP systems. According to this, the architecture of ERP systems is expected to be completely redesigned in the direction of cloud computing and, subsequently, platform computing. A new dynamic will develop from the functions available in the form of cloud services. This will give companies the freedom to combine the necessary specific functionalities and innovations for their differentiation and competitive advantages, at lower costs and with less implementation effort.