Enterprise Knowledge Graphs should be a Trending Technology for CIOs
New Business Models based on Platforms and Ecosystems
Personalization of the offering was one of the key topics discussed, for instance, by the CIO of Henkel, Michael Nilles. He used the term “hyper personalization” to address Digitally Native Vertical Brands (DNVB). An example would be a personalized shampoo which could be created based on individual hair structure analysis performed at your hairdresser. Of course, the product would be shipped directly to your home because retail does not make a lot of sense in this individualized business model.
In order to make that possible, you would need a large-scale digital platform infrastructure that allows for the collection of customer data and its analysis, on which individualized production and logistics would eventually be based upon. The platform would also act as an ecosystem, a community in which the vendor and its clients are connected online, services could be provided, etc. In case of car manufacturers, lot size 1 is the proper term to address the individualization goal. To be able to deliver personalized flashing (software) over the air is a strong technological requirement which presupposes end-to-end availability of product data over the whole product lifecycle, including the driving vehicles themselves and the sensor data they produce. Cars are seen as mobile devices with sensors connecting to the physical world, much like phones, offering a multitude of services that are subject to new mobile business and revenue streams. This is the reason why cars need operating systems (OS) on which the new services may be established. At the same time, an OS protects the mechatronic systems from the additional technical complexity induced by the services.
Hence, data- and software-driven really became similarly key for automotive OEMs. Volkswagen CIO Beate Hofer explained that the company is on its way to become a software developer, delivering mobility services based on a digital production platform that will integrate the entire supply chain. Of course, new economy competition has helped even propel this process in case of automotive OEMs. The same trend applies to smart buildings and smart cities (have a look at the work of Fraunhofer IAO [1]).
Traditional Business Models
Data-driven business is at the focus of attention these days. However, if we look at the analysis given by KUKA CIO Quirin Görtz, a significant portion of the business still is traditional, 70% in case of Kuka. To derive their IT strategy, Kuka analyzed its corporate capabilities needed to perform business (They did not take whole processes as the basis because they argued processes would get changed too often, thus the choice is on more fine granular business capabilities). They found that business capabilities could be classified along three dimensions:
- Areas of Innovation 5-10%
- Areas of Differentiation 20%
- Commodity 70-75%
Based on this analysis, the company had developed a pace-layered capability model from which one could derive its IT investment scheme. Knorr Bremse CIO Michael Hilzinger reported a similar approach to obtain their investment strategy. Traditional business models are usually based on grown infrastructures, on IT landscapes resembling brown fields rather than green fields, and mostly functional organization structures organized in silos. To loosely quote Thomas Mannmeusel, CIO at Webasto, “objectives are assigned to functions, but we work in processes”. This means that there are challenges in capturing the overall process, as silos hinder efficient provision of data to the business roles along the process. Thus, Webasto performs process mining to improve processes but is restricted to SAP systems because the data across the overall process lack semantic connectivity. This does not necessarily mean it should be a goal to overcome silos since these applications typically manage deep vertical know-how. It only indicates a lack of connectivity and business context across the process.
From Traditional Business to New Business Models
Ergo’s CIO Mario Krause envisages an objective often discussed in the financial industries under the rubric of “Know Your Customer” (KYC). He also reported silos as an obstacle to the transition from legacy systems to the "hybrid customer", which enables synchronized cross-selling across different distribution channels such as freelancers, direct and online sales. To advance harmonization, the homework must be done. And this generally is a requirement in all brown field industries. You cannot ignore your core business unless you invest in a completely independent new business, like Merck (James Kugler, CDO at Merck) and Palantir have done. [2]. They combine Merck’s market know-how and customer base and sell software licenses together.
Data Driven Business Platforms Use AI but Lack Business Context
What all these business transformations have in common is the commodification of data. Be it to improve customer services by analyzing vehicle braking behavior gained through technical sensor data of physical devices, by optimizing production efficiency through analysis of machine data, or by improving sales through evaluation of customer transactions in case of insurances and banks. To provide the needed business insights, data scientists make heavy use of analytics platforms. However, often with limited success because the homework mentioned above has not been made sufficiently or even at all, and thus, business context is lacking. In many cases, large and expensive projects (often in the order of several hundred million euros) have been undertaken in different industries in the past to consolidate the data of entire processes in one integrated monolithic solution. However, they were often prone to fail, in countless cases because the underlying business requirements changed faster than the projects could be implemented. This was not only because top-down waterfall methods were used instead of agile methods, but also because of the complexity of the projects themselves. As a negative side effect, well-established vertical solutions were often eliminated, and business got lost.
Business Context Represents the Knowledge About Processes
The actual aim of such projects was to establish data connectivity across processes. From a holistic point of view, connectivity of business objects represents the business context in which specific business operations along the process need to be embedded. What does that mean? As for the hybrid customer: knowing everything about him, his transactions and his connections to other business partners and companies may not be so easy to achieve in a sales process that, in the case of large banks or insurance companies, involves a multitude of globally distributed distribution systems.
Without this connectivity, however, AI-based money laundering applications could have limited success because they lack the necessary data across the entire customer journey. I figure the challenge will be the same when it comes to aligning the customer journey with the End-2-End company journey, which is a business objective reported by the Beiersdorf CIO Annette Hamann. In the case of Bosch, whose CIO Vijay Ratnaparkhe spoke about problems related to the improvement of components produced as part of a collaboration between different companies, it is crucial to be able to trace the entire value chain back to the production house where that part was made. Simliar to the financial use case, process data are maintained in different systems by disconnected people, even across company borders and thus connectivity and business context is lacking.
How to Establish Business Context by Means of Enterprise Knowledge Graphs?
Business context is about data and the semantic interpretation of the data within the context of a business process. We must distinguish the reasoning about the data from the data themselves. AI and Machine Learning algorithms can only be as good as the training data they are fed with. Thus, if transactions are subject to investigation, e.g., w.r.t. fraud detection or money laundering, it makes sense to feed ML technology with the network of entities that establishes the semantic context of the anticipated crime scenario. This might happen by linking all potential and available data sources relevant, e.g., the data of the involved CRM systems (scalability concerning automated graph generation is thus a strong requirement). In case of the traceability challenge concerning the product value chain of a discrete manufacturer, the network of entities relevant might be data connectivity across the whole product lifecycle.
.png)
Here, it is very likely insufficient to only analyze the sensor data of the products that are stored in a data lake. Both examples illustrate that business context gets established by linking data to provide business meaning. This is best captured by a technological methodology called Enterprise Knowledge Graphs. The latter represent the knowledge of the enterprise by means of connected business objects across processes and even across company boundaries. They can be seen as abstractions from the business processes and reside decoupled from the authoring systems (CRM, ERP, PDM, Data Lakes, etc.) from which they are computed and updated in an agile and automated fashion. Hence, they allow light weight integration of data across processes representing an Enterprise Long-Term Memory. In this way, a “future sensing organism”, as described by Siemens CIO Hanna Hennig, could combine AI/ML with a company-wide brain to perform meaningful reasoning. Data scientists could combine such semantic structures with AI/ML methods to improve their insights. Enterprise Knowledge Graphs can be used to connect the dots in case of money laundering or to materialize the hybrid customer but also to provide traceability functions as demanded for product traceability, predictive change impact or maintenance tasks. In other words, they can be applied in different industries and business scenarios to provide business context – wherever the use of complex connectivity is necessary to drive business insights. A major advantage is that they can be calculated automatically from existing company data. They act as cross-process, light weight data connectivity devices and can be implemented without changing given processes. The return on investment is thus available in weeks. Enterprise Knowledge Graphs are used by all major tech platforms such as Amazon, Google, Facebook, etc. to connect business objects and provide business context about people or products and other contents. I think, it is about time they get used in industrial settings as well.
