(More Than) One Step Ahead of the Law

One SAP employee’s seemingly off-the-wall idea to apply machine learning in pursuit of legal compliance has already borne proofs of concept for numerous industry use cases.

Bruno Terlizzi de Renzo, Senior Localization Product Manager, SAP Globalization Services
Bruno Terlizzi de Renzo

Bruno Terlizzi de Renzo now knows that no new idea is too crazy, no challenge is off the table when it comes to applying the latest innovation platforms that apply new technologies like machine learning to automate business processes.

The senior localization product manager from SAP Globalization Services was faced with the job of updating SAP Localization Hub, tax service with changes to the Brazilian tax code. Sure, thought Terlizzi de Renzo, somebody has to do it, but instead of updating the service manually, wouldn’t it be easier to give the machine enough examples so that it can learn by itself to describe each taxation rule and formula?

That idea was ambitious enough, thought Terlizzi de Renzo, but there was more to it than that. As legislation changes all the time in Brazil, it was clear that a complete machine learning model would be outdated before it saw the digital light of day. This triggered the second question: “What if we build a second machine that will read through legal changes, detect new tax rules, and generate data to teach the first machine?” For Terlizzi de Renzo and his development team, the feasibility of the idea was fading quickly – but that’s not the end of the story.

1,000 Legal Changes Across 100 Countries

There are around 1000 legal changes delivered across 100 countries per year. Deciding what is important or not can be challenging and missing a legal change can significantly impact business. The most valuable benefit of deploying machine learning to simplify legal change processing is that experts can be freed up from tedious manual work so that they can focus on activities that require their expert skills. And most importantly, the automation of these processes can significantly reduce non-compliance and fines and provide adequate time for the manual confirmation of the system-proposed data.

“Law-to-Action” is Born

Not long after Bruno Terlizzi de Renzo shelved his idea, he discovered that there was another team at SAP working on updating legal texts. The two ideas were merged and the concept was named “Law-to-Code.” So at the end of 2016, Terlizzi de Renzo embarked on his quest for self-learning program for legal-change processing.

In September 2017, it became clear to Terlizzi de Renzo and his team that a great deal of groundwork would be covered if he changed the approach to development by using the text services of SAP Leonardo Machine Learning Foundation.

Terlizzi de Renzo explains: “The amazing APIs delivered by SAP Leonardo Machine Learning Foundation pushed the innovation in a very positive way as the team could then focus more energy and time on the core proof of concept, to develop an application that utilizes machine learning. A combination of these services accelerated the proof of concept development in ways nobody could imagine, making a very small team even more efficient.”

Markus Noga, Head of Machine Learning at SAP
Markus Noga

“With the SAP Leonardo Machine Learning Foundation, machine learning can be developed quickly and without profound data science knowledge,” says Markus Noga, head of Machine Learning at SAP. “The team was able to paste our intelligent and seamlessly integratable foundation services into their application – easy to consume, scalable, and secure.”

Since the beginning of 2018, the proof of concept is now being shared with customers with the help of SAP Leonardo Machine Learning Business Development teams. The use cases covered by this initiative are growing, and currently include:

  • Compliance: As the legal obligations and regulations multiply worldwide, the need to stay compliant has become more critical than ever and highly qualified experts struggle with complex legal and tax regimes. With machine learning, customers can proactively scan and analyze thousands of records for legal requirements using algorithms to determine relevance of new changes so that experts can be alerted instantly with automatic workflows.
  • Pharma and Chemical: The pharmaceutical industry has one of the most regulated systems in the world. The U.S. Food and Drug Administration (FDA) publishes new and changed regulations on a daily basis and it is quite a challenge for pharma companies to keep abreast with these updates and also change recipes within the stipulated time. In this use case, machine learning allows customers to track changes such as the announcement of banned substances and notify the relevant departments in time so that they can take action immediately.
  • Food and Beverages: In the global food industry, the same product is likely to be sold across most countries, however with different packaging that complies with country-specific legal requirements. By plugging a robot to the regulators, a food producer can instantly get informed when new regulations are published, and the initiate compliance requirements in the packaging.
  • New Industries: Emerging industries like autonomous driving and blockchain are disrupting entire markets, but are not immune to governmental regulations. Given the small scale and lack of experience in dealing with regulatory bodies, companies in these industries can use Law-to-Action can help them stay compliant with regulations.
Ferose V.R., Head of SAP Globalization Services
Ferose V R

Ferose V R, head of SAP Globalization Services, is convinced this is just the beginning a long and productive use of machine learning in tracking the unrelenting jungle of legal changes worldwide. “The law-to-action use case is easy to understand and demonstrate: It demystifies machine learning; shows the power of machine leadning and, with appropriate human in the loop, drives unique value to end users,” concludes Ferose.

Who Made “Law-to-Action” Happen?

Law-to-Action Team in SAP Globalization Services: Michael Depner, sponsor; Jan-Klass Heinsohn, sponsor; Gert Eichberger, product manager; Bruno Terlizzi de Renzo, product owner; Vivek Gollapudy, developer/intern; André Dantas, developer; Marco Furlanetto, developer; Micael Ramos, developer; Guilherme Dellagustin, developer/fellow; Tamara Juschkova, developer/fellow.

SAP Leonardo Machine Learning Foundation Team: Markus Noga, Christian Boos, Helena Staader

Vivek Gollapudy, Micael Ramos, Guilherme Dellagustin, Bruno Terlizzi de Renzo, Gert Eichberger, Marco Furlanetto.

SAP customers interested in co-innovation can contact Michael Depner from SAP Globalization Services.

Namita Gupta-Hehl is head of Marketing and Communications for Globalization Services at SAP.