BSH Home Appliances Group

1. As the Global Head of Data Science, AI & Intelligent Automation, what are some of your key roles and responsibilities that you take while serving at BSH Home Appliances Group?

Before coming to my responsibilities at BSH, let me first explain what we do at BSH as a whole, what data we collect and how we work with it. With 62.400 employees and a turnover of 15.6 billion euros, BSH is one of the leading home appliance manufacturers worldwide and Europe’s No 1. With eleven Brands like Bosch, Siemens, Neff, and Gaggenau we produce over 45 million kitchen and household appliances every year worldwide, such as cooktops, ovens, fridges and freezers, kitchen machines, blenders, laundry machines and driers. Our wide range of products, brands and services cater to a broad spectrum of needs, from entry-level appliances for price sensitive consumers, to extra-large fridges and laundry machines for big families or premium wine coolers for the luxury segment. This allows us to improve quality of life at home based on the individual needs of our consumers. And – as of last year – all large home appliances come with connectivity. These connected appliances, along with machines of our production process, generate the majority of the data we use to improve our operations, products and services.

Back to my specific role: As Head of Data Science, AI & Intelligent Automation I am responsible to build up capabilities in these disciplines at BSH globally. This is achieved by defining the skills required today as well as in the future, by hiring new experts accordingly and by creating training paths to upskill the existing workforce and providing mentorship along their career.

My other main responsibility is to develop the BSH AI Strategy collaboratively between departments, as well as to ensure the creation of domain-specific AI Playbooks that detail the cross-departmental AI development and MLOps models. Finally, my task also is to identify gaps in the existing BSH AI infrastructure and tool chain, and trigger the necessary investments to close them.

Since AI is still an innovative field with many unanswered questions, I am also operationally involved in pioneering AI projects, which, e.g., bring AI to our home appliances. I firmly believe that strategy, process and infrastructure development need to be informed by and happen alongside actual projects, where the true demand becomes visible.

2. Tell us about the center of excellence for Big Data and AI that you have built for BSH Home Appliances Group? How has it helped your organization in propelling ahead of the competition?

Several years ago, BSH decided to address the challenges and opportunities of the digital transformation via a new department dedicated to digital transition. Inside it, the Center of Excellence for Big Data and AI was created. It stood on two main pillars:

Firstly, we were caretakers of BSH's IoT data asset, created from the data generated on connected home appliances. We set up cloud-based big data management, including privacy engineering, and ensured that the data could be utilized by various users and use cases at BSH. Having already answered the questions around technology, processes, governance and (partly) privacy, the Center of Excellence for Big Data and AI proved and still proves to be a major accelerator for subsequent data-driven and ML use cases.

Secondly, our data scientists worked with various BSH departments to identify and prototype data science use cases. Although it took a while until first use cases moved from proof-of-concept to full roadmap projects, these activities greatly contributed to BSH's ability to do machine learning. Starting from scratch, after two years we had introduced a structured approach to analyzing and prioritizing ML use cases, a professional work mode for data scientists, an adequate data science tool chain, thriving communities of practice where experts connected and had clarified many questions around data availability and data access at BSH. Also, our counterparts in business departments had been part of this learning process which increased their understanding of and, overall, BSH's organizational capability to do AI.

The journey towards AI is an exciting, yet extensive endeavor. Approach it with your eyes open. You will only learn to build AI if you do build AI

By now, the Center of Excellence for Big Data and AI has outgrown its exploratory phase and grown into multiple long-term data / AI products, as well as into a team of data scientists that cater for the continued need to experiment and prototype data science use cases.

3. Tell us about the challenges that you have encountered for not only setting up the right IT infrastructure and processes but also fostering the right mindset among colleagues?

The first challenge revolved around technology: When BSH started the journey towards becoming a data-driven organization, there was a well-established business intelligence landscape, implemented on a large vendor's monolithic on-premise suite. Moving from there towards a state-of-the-art cloud-based data lake did not only involve new technology, but also required new skills that were not easily available at BSH. In addition, it raised new, unanswered questions around IT governance and security. All of this demanded of data experts to acquire new skills, and to change their mindset of addressing new challenges by purchasing new components towards building solutions themselves.

The second challenge was about data: Data available for analytical purposes used to be strictly separated from operational databases and, at the same time, was heavily pre-processed. However, the newly joined data scientists required access to low granular, original data, which was mainly available in source systems not prepared for analytical access. There were also no guidelines or regulations for who may provide data access to whom based on which criteria. We needed to shift people's mindsets towards data democratization, towards the mindset that not giving access to data requires reasoning, but denying it does.

The third challenge involved business cases: While we soon had several highly talented experts bringing personal AI capabilities, the organizational capability to do AI equally depends on other roles' understanding of the opportunities, challenges and characteristics of AI. Product managers need to be able to "think" an AI product, which includes putting more emphasis on continuous improvement after launch instead of only focusing on the initial development. Project managers need to understand the inherent uncertainty of AI projects and the impact this has on planning, as well as the "sequentiality" of data collection, data sampling, data labeling and its impact on time-to-market. Middle and top management need to understand that, while a neural network can be trained quickly, it is only a small part of an AI solution and significantly higher effort is required for putting everything else in place. Frequently, expectations of non-experts towards AI are at the same time unrealistically high as well as unnecessarily low.

For us at BSH it is clear that learning to do AI involves more than integrating some new technologies and hiring a couple of experts. It is a transformation process on multiple levels, which requires dedicated training programs and buy-in from top management. This transformation extends the digital transformation of a company and, in some ways, is its end game.

4. What would be your piece of advice for your fellow peers and leaders?

Own your data. If you agree that data is an asset, then it should be treated accordingly. This means to invest in it continuously, to grow your data assets proactively and with foresight instead of only reacting to the next project. Not only the data infrastructure, but also the data itself is an enabler for your future business.

As an AI leader, don't focus on the machine learning aspect of an AI project. It's usually not where the real challenges are.

Take everything the world learned over decades of professional business intelligence, data warehousing, software development, DevOps etc. Then transfer it to the special circumstances of AI projects. All of it is still valid but needs to be considered from a new perspective.

As a manager, be aware that doing data and AI properly is expensive. Prioritize strictly, and then provide sufficient funds for whatever projects you decide to do.

The journey towards AI is an exciting, yet extensive endeavor. Approach it with your eyes open. You will only learn to build AI if you do build AI.