“There is an urgent need to develop more efficient AI algorithms that require fewer resources and data in order to justify the use of technology when in times of climate crises."

These are processes that nobody or even hardly anybody wishes to follow. For instance, this also includes faxing or scanning documents, which are admittedly now a thing of the past. So why would people not outsource unpleasant document processing to an AI device? This is was the idea behind the Berlin start-up Workist who then developed the WorkKI for sales representatives. Alexander Müller, who is one of the three founders, tells us how the AI concept and the company was formed, what wishes he has for the future of artificial intelligence, and he even has a hot tip for the cold winter spell.

Hello Alexander! Your headquarters with Workist is in Berlin, so why does the establishment of an AI start-up business work so well there?

Berlin is clearly, and is without a shadow of doubt the founding capital of Germany. There is a short distance ecosystem here and you can meet up with three venture capital firms in just one afternoon. This would simply not be possible anywhere else in Germany. There is plenty of talent and money under one roof here which also means that a number of AI start-up businesses can emerge and develop very successfully. A number of fabulous concepts then develop, which are then quickly implemented and meet VCs who are willing to finance founding business owners with serious plans.

You still have two co-founders, Tim Wegner und Dr. Fabian Brosig. How did you three join forces and how did the Workist concept come to fruition?

I met my two co-founding partners when working at my previous employer, Minodes. Fabian and myself were responsible for the technology division at the start-up business that Tim founded. We initially went our own separate ways once Minodes was sold to Telefonica. However, it became clear to all of three of us during 2019 that we would like to start something together once again and we started working on Workist.

Our starting point was: We wanted to use state of the art AI technology to make work easier for people. We started travelling throughout the country and started speaking to potential clients. We experienced some situations where our clients' employees were then typing forty-page orders from a fax, a PDF, or from paper into the SAP system. It seemed so far out of date that we knew we had to make changes. It was obvious that we had actually come across a problem which remained unsolved once we had looked into this along with other potential clients. Virtually, the entire B2B finance back office staff is kept running by companies who are sending PDF files or Excel spreadsheets back and forth and where data is being typed from one IT system to another.

You apply modern and in-depth learning methods for document processing at Workist. Was there one initial moment when you decided to start a company focused on this domain? And how does your AI system work?

It quickly became apparent that we wanted to use our skill set here to solve some major issues based on mine and Fabian’s experience in the field of in-depth knowledge and being able to learn about machines. We started developing a product once we had identified the issue. We worked very closely alongside clients to develop a solution that enables clients' employees to automatically read orders and other documents and to be able to transfer them to an ICT system.

Our product can generally be described as a document intelligence solution which specialises in extracting data points from documents. Our AI system can for instance easily extract data (such as item numbers, delivery dates, order numbers and much more) from order documents.

We are combining in depth learning methods such as Graph Convolutional Neural Networks, Computer Vision and Transformer models in order to make this possible. The entire document intelligence domain is developing at a rapid pace and we are constantly refining our algorithms and are trying out new things on a daily basis.

At Workist, you currently have a product in your portfolio called WorKI.. Are you planning to develop any other solutions in the near future?

WorKI is the name of our AI. As things stand, we have mainly focused on sales from manufacturers. We are also however working on solutions for purchasing and accounting due to the high demand from other industry sectors. This means that there will soon also be “WorKI for Procurement” and “WorKI for Accounting” platforms.

Keyword future: What other developments in the field of artificial intelligence and machine learning do you expect to see over the next few years? Where will we be in 2030?

My wish is that AI will be used by us human beings in 2030 so that they can carry out activities that do not exploit our human strengths. This for instance means that human beings will use AI as a tool and will not allow social media platforms to act as their slaves.
Plenty of research still needs to be carried out in order to implement the broadest potential AI applications. I do genuinely believe that there is an urgent need to develop more efficient AI algorithms which require both fewer resources and less data. This is the only way that people can justify using technology in times of a climate crisis when our industry is not under pressure to justify itself.

And last of all, we are looking forward to your personal tip: Where in Berlin can people feel most at ease during the winter spell?

History, along with AI is my “guilty pleasure” and there is more than enough of that here in Berlin. People often go to the museum to keep warm during the cold winter period. I would specifically recommend the German Historical Museum, which is free to enter. I've spent a number of afternoons there and I always learn something new!