There is no consistent definition of digitization. Every industry and every department perceive it differently. How would you define digitization with respect to your position as Managing Director at Heldenkombinat GmbH?
Digitization is like innovation: Totally meaningless in itself without a clear goal or vision. Most companies limit their digital vision to automation – but this is just the first step. Our digital vision is broadened to autonomization: The purpose of making products or processes digital is not (only) economies of scale or speed. The real value is ‘true mastering of a domain‘. So if we, for example, build a recommendation engine that adapts to customer behavior over time autonomously, our clients’ vision is not to substitute humans through software. Instead, the vision is to enable new creativity based on new domain knowledge that is derived through the interaction of our software with customers and to share these new insights with the human employees involved in order to further enhance or to create innovations.
At Heldenkombinat, you focus on AI and machine learning. Where do you see the most important challenges for implementing these mechanisms for the business purposes of large companies at present?
I see three main challenges. First, AI knowledge is insufficiently decentralized in most large companies. At present, only few people truly understand the real benefits of AI and what is needed to develop and implement AI into the core processes of a matrix organization. Second, there are so many service providers and start-ups out there claiming to use AI, but in most cases these offerings are either not advanced and/or too standard and/or too costly. You have to talk to and kiss a lot of frogs to find the prince also in the AI world. If you play it safe and focus on the big service providers, you are not able to establish competitive advantages in your industry. Third, AI research is moving very fast, and it is also an organizational challenge to keep pace on the one hand and to deselect on the other hand the ideas and approaches that are not business ready yet.
In which fields do you already see some clear business readiness of selected AI applications, and which fields do still rather have research status?
In general, most current advancements in AI (deep learning) are based on labeled data and hence we see a lot of business applications there. But human labeling is expensive, error prone and time consuming. That is one reason why research is focusing on unsupervised and also deep reinforcement learning to facilitate (semi-)automatic labeling or learning from unlabeled data and pure interaction. Capsule networks fall into that thinking, too. They try to better learn from small data and to enable transfer learning across data sets. Generative models are also a just emerging family of algorithms that have a great potential to be disruptive. In a nutshell, such algorithms will start to generate content rather than just analyzing existing data.
How important are business model innovations for capturing the value of AI and machine learning in large established companies?
It is essential to deeply link business model innovations with AI for the reasons I explained earlier. With deep learning one can extract features and patterns of features that are relevant to optimize specific outputs. Beforehand, human experts were needed to define the relevant features. If you transfer that thinking to business models, AI can help to understand valuable patterns between the respective elements, and it can propose (and later execute) changes autonomously. Then you will get valuable insights how to further innovate business models. This is an interesting journey that companies should start early.
In what way do you rely on AI and machine learning in your internal activities at Heldenkombinat?
We have internally developed a reinforcement learning agent to come up with new strategies for how to invest our own free capital and cash flow. And we are using a disruptive 3rd party AI enabled chatbot (from Passage AI) for some internal and external communications.
What is your personal goal for the next year or so regarding digitization and innovation?
We want to continue our wonderful journey. ;-)
About Olaf Erichsen
He is Managing Director and Co-Founder of Heldenkombinat GmbH based in Hamburg, Germany. In the last 18 months, Olaf has evaluated approximately 200 AI providers (global service providers and start-ups), thus providing an excellent overview of what is promised in this area and what is actually held. His company not only advises large companies in the fields of AI and machine learning but Heldenkombinat develops state-of-the-art AI and deep learning applications with its own developers.