Why SAP has built an “AI factory”

Dusseldorf Invoices or reminders, delivery notes or terms and conditions – before the employees at Villeroy & Boch’s head office can process customer concerns, they have to sort them out. Colleague computers have been doing this annoying task for some time now: A system from software manufacturer SAP automatically opens all e-mails and classifies them using keywords. This works in 90 percent of the cases.

Villeroy & Boch is a showcase example for SAP that people like to show at their own events, most recently at the Tech Ed developer conference. The software manufacturer sells products with which companies organize business processes from personnel selection to material procurement to controlling – and artificial intelligence is supposed to take on more and more tasks.

There are already a number of application scenarios – from chatbots that users can give commands to virtual accountants who match incoming payments with invoices. After experiments with the technology, however, it is important to broaden the knowledge, says Feiyu Xu, who has headed the division as “Global Head of Artificial Intelligence” for a year and a half.

An interdisciplinary team is now working on improving the various products in the SAP portfolio with algorithms – internally there is talk of an “AI Factory”, a kind of factory for algorithms. “We are now producing intelligent processes and applications,” says Xu.

Top jobs of the day

Find the best jobs now and
be notified by email.

The software manufacturer is reacting to a trend that is affecting the entire IT industry. Many companies lack the know-how to build systems with artificial intelligence themselves. There is therefore a “development towards standard software”, says Peter Buxmann, professor for business informatics at the TU Darmstadt. Oracle and Salesforce, Amazon Web Services, Microsoft and Google also offer preconfigured algorithms. It’s a billion dollar market.

Customers can access functions such as image recognition, text analysis and translation via the cloud. “This significantly reduces the programming effort for the customer,” says the scientist, who focuses on the transfer of technology into practice. However, it will not work completely without your own knowledge in the future either.

Second wave of digitization

The trend is particularly important for SAP. The German software manufacturer already shaped the first wave of automation with its programs for controlling business processes, says Xu. Now it can also play a decisive role in the second wave, which is about making processes intelligent. “SAP has an important role because the entire ecosystem benefits from it.”

That was an important motivation for Xu to switch to the German software manufacturer. The scientist has been working at the interface between research and application for many years – at the German Research Center for Artificial Intelligence (DFKI), in her own start-up and most recently at Lenovo.

At SAP, the scientist has set up an organization in which employees from different areas come together – the company does not quantify the size. On the one hand, it should ensure that artificial intelligence finds its way into all parts of the portfolio and, on the other hand, that possible problems are taken into account at the same time. “This is about business-critical applications,” emphasizes the manager.

On the one hand, the team consists of specialists in artificial intelligence who have mastered the technology. In addition, experts from business areas such as human resources and production are represented, who know the customer’s point of view and can thus assess the needs of the market. There are also lawyers and IT security people who take care of data protection, for example. There has also been an ethics committee since 2018 that can intervene in critical cases.

Another important initiative by Xu: Since March, the software manufacturer has been asking existing and new customers to provide data from their systems – voluntarily and anonymously – in order to train the algorithms. It is the subject matter without which Artificial Intelligence cannot work.

SAP headquarters in Walldorf

The software manufacturer claims to have improved several hundred business processes with artificial intelligence.

(Photo: Bert Bostelmann / image portfolio for Handelsblatt)

“The customers help to continuously improve our products,” says Xu. For many companies, it can be tricky to disclose such information. “A majority,” says the SAP manager, but agrees.

Supply chain guards and chatbots

So far there have been “a few hundred” application scenarios from different areas. For example, SAP customers can match incoming payments with outstanding invoices, which saves tedious work. You can use algorithms to identify impending delays in the supply chain. And they can program chatbots that answer questions. Customers include Hewlett Packard Enterprise, Dulux, Mitsui and Döhler.

However, Xu’s vision goes further: the software for corporate planning, called “Enterprise Resource Planning” (ERP) in technical jargon, will one day assist management. “The ERP system should become an agent who, for example, communicates with the CEO or CFO and makes suggestions.” The advantage of the algorithms: You can react to events in real time, 24 hours a day and multilingual.

However, Xu rejects a self-regulating company, analogous to autonomous cars. “People have to be able to intervene at any time.” Because: The decisive factor with algorithms is the data with which they have been trained – “garbage in, garbage out” is a common saying.

The technology also uses past observations to predict the future. She cannot therefore take into account unexpected events such as the pandemic.

The introduction of software with machine intelligence is not, however, trivial, despite the preparatory work by SAP. For example, Villeroy & Boch had to train the algorithms with its own data so that they could reliably identify the documents. With other products, the IT departments should be able to introduce the functions without experts such as data scientists.

Either way: companies should build up their own know-how, emphasizes Christian Hestermann, analyst at the market researcher Gartner. IT must be able to assess how good the prefabricated processes and solutions are. Control mechanisms are also needed. “AI is a young technology, there are always teething problems.”

More: In order to allay fears about artificial intelligence, the decision-making processes of the AI ​​software must be as transparent as possible, demands the AI ​​boss at SAP, Feiyu Xu.

source site-13