Companies want to use smart tools, algorithms and bots to make better decisions and find faster ways of working. Entire departments are currently being set up for this purpose. However, other parameters apply to German SMEs. Here, tangible solutions that enable rapid efficiency gains are usually still expected. "However, if you have the courage and foresight to embark on an iterative process, you can only win when using AI," says Gregor Hüls, Solution Manager at Modula.

As the McKinsey Global Institute recently calculated, the field of artificial intelligence (AI) is expected to generate a higher growth spurt in the medium term than the steam engine did in its day. Even the value creation potential in the field of industrial robots and the former spread of information and communication technologies cannot be compared with this. And yes, AI also represents a great opportunity for SMEs. However, they need to take a close look at their own business model and business processes. The experts at Modula therefore advise taking this path early on and show that a lot can be achieved even with small steps.

The right partner at your side

In addition to modern ERP and MES solutions, Modula also provides the right AI technology platform with AI products that enable SMEs to master the challenges of digital change.

According to Modula, the recipe for success in the field of AI is based on platform technologies that are suitable for SMEs and practicable solutions that are suitable for SMEs. "Whether in administration or production - AI technology enables completely new business process approaches in SMEs and creates new competitive conditions," says Hüls. "However, expectations of artificial intelligence are now so high that there will also be disappointments. But there will also be many solid applications with great added value. It is important to share your vision with the right partner."

Knowledge management offers great opportunities

"We can only answer our customers' question as to whether AI is intelligent in the classic human sense with a 'no'," emphasizes Hüls. "However, an AI solution can store and apply corporate knowledge. The processing of enormous amounts of data in a very short time using AI knowledge models is the real added value of this technology. For example, we see great potential in the automation of master data management through the use of cognitive business robots (CBR). We see further potential in the area of digital secretarial services using CBR. Here, companies can iteratively build up a knowledge database that optimizes established routine tasks from day-to-day business to the extent that employees have more time for their actual, value-adding activities."

AI in business processes

But where exactly is the knowledge that is needed to build so-called knowledge models? "90 percent of it is in the heads of the employees involved and unformatted in daily correspondence - this treasure must be salvaged in line with the company's objectives," answers Gregor Hüls. "If you cleverly link emails, ERP and DMS with each other, this results in a knowledge base per se that can be used in multiple ways for AI purposes."

Although some companies already use these links, for example by linking their incoming mail with various automation processes (e.g. invoice receipt workflow), the knowledge transported remains unused. For example, although the information on a scanned incoming invoice can be read using OCR, the knowledge itself is not available in a controllable form, as would be the case with AI.

Hüls therefore advises medium-sized companies in particular to take a close look at AI. "Before deciding on a solution, the requirements should be precisely defined. If AI is to work, the on-site preparation must be thorough. For software to behave intelligently, it needs to be fed with real data such as email content. The higher the amount of training data supplied, the more successfully the AI can work."

The system is made better by people

The amount of attention an AI project requires, especially at the beginning, can be illustrated using the example of handwriting: Handwritten information, such as trade fair contacts, usually has to be manually entered into the system by the secretaries and assigned to another function. "For the AI application, we have to extract the model and turn it into an identity," explains Gregor Hüls. The model can be, for example, "offer", "new customer contact", "copy to person XY" or, in the case of complaints, "forward to various recipients".

The problem in this case is often illegibility. "Not even we humans can decipher everything one hundred percent with the help of our brain power." So the "machine" first has to learn what is right and what is wrong and where it should organize the information. "The system is improved by humans," explains Gregor Hüls. The system remembers the assignment made by the employee, for example in the case of an illegible address.

The human therefore teaches the system something every day through constant interaction. "As an ERP integration, for example, we offer a self-learning cognitive business robot that acts as a digital assistant and enables a highly automated AI secretariat. When entering numbers, the automatic hit rate is already 90 percent. For words and grammar, this is of course somewhat lower, but our results improve with every customer. In the end, 90 percent of emails can be processed automatically. The robot recognizes the subject of an e-mail, assigns it to the recipients or even pre-formulates a reply. As a result, massive cost savings can be made within the written processing procedures in the first year."

AI as software-as-a-service

The computing power required for AI can hardly be afforded by a medium-sized company. This is why Modula offers the AI secretariat as a hosted service from the data center. The tool can then be integrated into ERP systems such as oxaion via a BPM process, for example. "Knowledge cannot be summarized in specifications. It takes several iterations to build trust in this new technology," says Gregor Hüls. "We also invite interested parties to our workshops to reduce initial reservations."