Artificial intelligence: cognitive systems for SMEs

Products that approach artificial intelligence are now widely spread, as Apple's Siri assistant, Microsoft's Cortana assistant or IBM's Watson program show. Meanwhile, SMEs also have to deal with newer developments because, for example, voice or image recognition programs or digital assistants can also handle volumes of data in their areas.

Digital payload

"It is possible for large companies to buy in the technologies and use them as software packages," says Jana Koehler, a specialist in artificial intelligence in the Department of Computer Science at the Lucerne University of Applied Sciences and Arts. Nevertheless, "Swiss SMEs often lack the money, knowledge and manpower to tailor the programs to their needs." In addition to the licensing costs for the software packages, there are further costs to train the software and integrate it into the company's IT.

The electronic colleague must be able to access data and be integrated into the work steps of the human colleagues.

Cognitive services for SMEs

"Artificial intelligence solutions will change the division of labour between humans and machines," says Jana Koehler. That's why her research team from the Lucerne University of Applied Sciences and Arts and the IT service provider UMB AG in Cham have joined forces in the "Cognitive Services for SMEs" project to play through what changes an implementation of artificial intelligence will mean for companies.

Assistant prototype

With the prototype of an intelligent assistant for an IT service desk, the team analyses process flows and develops solutions for the tasks that cognitive services can take over. Together with UMB, the team from the Lucerne University of Applied Sciences and Arts selects cognitive technologies and implements the intelligent assistant. UMB additionally provides data and knowledge in customer support.

Software companies such as Microsoft, Amazon and IBM offer cognitive technologies directly as services in the cloud on their websites. One service recognizes language, for example, one can translate text, a third searches the archive for texts with similar content, and a fourth makes the computer talk. "The programs need data tailored to them and must first learn how to perform their tasks in the company," says Jana Koehler.

For the research project, Jana Koehler's team divided the work steps of an IT service desk into small steps: The texts of the incoming emails have to be analyzed. What kind of problem is it? Has a similar one already been solved? How urgent is the situation - has a critical software system failed, or has an employee just forgotten his password? What area does the request fall into? Who is the best contact person for the customer? Accordingly, the wizard creates a ticket, analyzes the request, solves simple problems itself and immediately forwards urgent and difficult ones to the expert.

In the beginning, such an assistant needs help from human colleagues, but over time it learns to solve more and more queries on its own by drawing on knowledge and information sources.

In doing so, the assistant must adapt to the customer and correctly receive their requests. This assignment poses challenges.

"Tit for tat"

Technologies already exist that recognize a person's mood. But Jana Koehler also reports on a program that reacted to harsh formulations with equally harsh responses. The computer could only respond in a mirror image - and had to be taken off the market after 24 hours. "The use of cognitive services must be controlled," says Jana Koehler. "Otherwise, the risk is far too great."

Please find further information about the research project of the Lucerne University of Applied Sciences and Arts here