A glimpse into the future with the AI Incubator


Interview with Dávid Molnár, project manager within the DTSE AI Incubator. According to him, by using artificial intelligence, we can deliver better services to our customers and shape a creative work environment for our employees.

How would you describe the AI Incubator?

The DTSE AI Incubator is not a typical project, but a program with a start-up spirit. After the business owners identify a process or a task that Artificial Intelligence might improve, we then go through the “Double Loop Journey” to better understand its current situation and then to come up with an AI-Powered solution. And just like that, we have a Use Case. 

During the prototype phase (4-month project accelerator) the magic begins and we create a functional or simplified test model in a close cooperation between the business side and the Data Scientists. In the validation phase, we verify if the model meets the criteria of the Proof of Concept. If the validation is successful, there are two ways forward: operationalize or scale.

What did you already accomplish?

The journey started in July 2019 and since then we have gone a long way to build a solid structure and strategy. I joined the team last October to support the AI Incubator and act as Project Manager. I believe it is important that the team is stable and composed of people enthusiastic about AI topics and complementary in terms of skills and responsibilities. This is an important cornerstone to build on, especially when we are sitting in various locations (Cologne, Brno, Bratislava).

Another achievement was the first DTSE CZ Hackathon in Brno, supported by LovelyData, Engeto Academy, the Data Science Community from Prague and the independent platform Machine Learning Meetups. This represents the beginning of our AI Incubator, as we hired Stefan, one winner of the Hackathon, who is now part of the AI Incubator team as one of the data scientists.

What is the current status of the AI Incubator?

We have successfully validated following use cases: 

  • Smart FI Automation: goal of the project is to increase the automation of the existing posting process for incoming FI vendor invoices. The AI solution predicts the G/L account and the tax code by taking into account additional information obtained by reading the invoice (PDF) with an OCR solution.
  • Customer Feedback Evaluation: with the help of Machine Learning, we process natural (human) language automatically. We extract topics of concern and highlight priorities in the feedback offered by customers. Thanks to sentiment analysis and the filtering of concrete suggestions product owners know exactly where to start with improving our services.
  • Minerva: the target of the use case Knowledge Research on Minerva is to make unstructured content (documents such as meeting minutes, design documents, guides & tutorial, PowerPoint presentations and Excel files, etc.) available for further data analysis, to provide a knowledge base for users sharing same content and to provide apps to search and access the knowledge.
  • Delphi Forecast: based on historical data, we predict the potential future KPI values in the Profit & Loss Statements using Machine Learning and Time Series. The aim is to highlight deviations and provide reasoning in monthly flash reports, as well as to provide forecasts based on the reporting figures delivered by customers.

These use cases are ready for further enhancement and offered to additional customers who want to improve their processes.

Additionally, we are also working on further use cases and we have also started AI in Action Calls, in which we present to our colleagues our running projects and offer a “behind the scenes” look.

How do you bring the AI topic close to all employees?

One additional and important element of the AI Incubator is the learning part. We want to educate our colleagues and enable them to understand what artificial intelligence is. That’s why we have developed two training programs. The AI Influencer program (altogether 30 colleagues have enrolled) for people curious about AI, data, communication. Some influencers also had the chance to contribute in the marketing of AI Incubator or to take on some role in one of our projects.

The second program is the AI Tech Trainee for colleagues who would like to go deeper in the AI world and learn about Machine Learning, Python, Natural Language Processing, etc. 

What are the next steps? 

Foremost, we want to accomplish the 4 ongoing use cases we are working on and decide to which next project to allocate our resources. Looking ahead, we plan for the next Autumn the second Hackathon in Brno. Although it might seem far away, we have already started working on that. We also want to offer the next & last step in the learning journey: AI Specialists, for those who want to steer their careers into the data science direction. There we plan to offer coaching, mentoring and on-the-project training. 

Thanks to these activities we hope to increase our presence and awareness of AI in all DTSE locations and keep on integrating our AI influencers & trainees into the activities of AI Incubator, based on their interests.