| Thomas Bögel
Digitization strategy and AI as pillars for standardization and quality improvement
What is digitization and artificial intelligence?
Caused amongst others by the current situation, many companies see themselves confronted with the topic of "digitization" in order to implement work processes in a digital form and thus independent of a specific location. Due to the current health situation and its subsequent safety policies, many companies were forced to switch to digital working abruptly in order to sustain business continuity. More often than not, the resulting and rather chaotic implementation led to short-term and short-sighted solutions instead of permanent resilience through digital processes.
Digitization is a broad term and ranges from the support of currently manual processes by IT components to fully automated processes. The field of artificial intelligence (AI) plays an important role in this context. Sensational reports regularly suggest that AI can be used ad hoc. However, when AI is actually applied in a real-world business context, it is often noticeable that AI doesn’t yield sufficient quality due to inept processes, data and algorithms. The guiding principle thus shouldn’t be the desire to introduce artificial intelligence but instead the desire for sustainable process optimization and a digital strategy – subsequently, AI will find its appropriate place.
In this blog post, we would like to use the example of the digital transformation of the HWI group to provide an overview of why a holistic digitization strategy should be implemented for solid and sustainable digitization as well as its advantages – especially for our clients.
Digitization strategy - seeing the forest for the trees
Often times, companies choose a "fire hose approach" and selectively introduce individual digitization components without looking at the big picture. In the end, this approach can lead to additional costs and even a loss of efficiency.
In order for digitization to be able to play out its advantages in the long term, however, it must be precisely aligned with the corporate strategy, rolled out iteratively and support technical processes from the very beginning. A major challenge here is the wealth of possibilities and options available today, so that questions such as the following arise: how and at what point in the company can digitization be implemented and to what extent does automation make sense? Suitable methods help to bring clarity and priorities into the cornucopia of options. Overall, a holistic approach is the road to success. Sound quality needs to be combined with speed and agility to achieve results quickly and at the same time be successful in the long term.
The digitization strategy of HWI is based on the following four building blocks:
- Structured process digitization and automation
- Knowledge and data management
- Garage model for AI minimal viable prototypes
- Digitization platforms for subsequent use and data linking
The topics are flanked by data protection, regulatory requirements and issues such as trust and values in automation.
In the following, the individual building blocks are briefly presented to demonstrate that digitization as a whole is much more than the introduction of new "tools". It must be considered strategically in its entirety and embedded into the corporate strategy.
Structured process digitization and automation
The foundation of any digitization approach should be an inventory of the existing processes and IT components within the company. In a so-called "enterprise architecture", actors, processes and systems are brought together and thus deficits in basic digitization become apparent.
Using a suitable methodical approach, potential for digitization and AI is identified and prioritized in workshops. Design Thinking and agile working methods proved their worth in this context. After identifying suitable processes, a target architecture is derived - always with the involvement of the subject matter experts - which represents a technical solution to the business expert requirements. In agile implementation sprints, technical prototypes are iteratively shaped to meet the business requirements. The integration of new digitization components after successful implementation should also be planned in a structured way and architecture decisions should be weighed up: for instance, the gradual use of cloud infrastructure for deploying new components (public cloud, hybrid and private cloud) facilitates planning and scaling options and at the same time requires thinking about considerately opening interfaces and allowing for external connections.
Knowledge and data management
Well thought-out knowledge and data management should be part of any digitization strategy. Thinking about data, it is important to choose the right model and the right degree of structure for each use case: from classic data warehouse architectures to lose data collections (so-called data lakes), many options are available. Any form of data silo creation and data storage without flexible interfaces should be avoided: even if no potential for artificial intelligence or process automation is currently discernible, data architectures must be flexible and able to grow iteratively. This is of utmost importance because data that is difficult to access is one of the biggest challenges for the implementation of future, data-driven use cases. The knowledge available in data is fed from various sources, processes and teams and is always dynamic. As a result, any fixed structure will rarely be of lasting value.
Of course, all the flexibility and agility of the data model does not excuse the data architect from the need to clarify important formal points, such as regulatory requirements and data protection.
Garage model for AI minimal viable prototypes
The potential of artificial intelligence is enormous and possible points applications are omnipresent. Systems based on AI are able to search and extract keywords from unstructured documents (e.g. from process instructions or product information sheets) and store derived knowledge in a structured form in order to make it easily accessible to subject matter experts and thus reduce the effort for reading long documents. Using AI chatbots, systems can even understand natural language input and thus respond to information needs in natural language or guide business users through complex processes. Since about 80% of accumulated data in companies is unstructured, automatically processing this data in particular holds great potential.
The usability and quality of AI models are strongly dependent on the respective use case. Many factors play an important role and cannot all be fully considered in advance: for instance, missing, incorrect or insufficient data are the biggest hurdles for AI, so discussing data availability and dealing with issues around data will be omnipresent. In addition, AI will never work error-free. The question "how many errors can we tolerate?" and aspects of the impact and handling of AI errors can usually not be answered ad hoc but require discussions between technical delivery team and subject matter experts at eye level.
According to the credo "fail early, fail fast" from agile development, it is advisable to start with manageable, self-contained projects in which small AI use cases for so-called "minimal viable prototypes" (MVPs) are built and tested in direct interaction between the subject matter experts and the technical team. The goal is to be able to estimate the effort required for a concrete implementation after a short time in order to decide whether the AI experiment should be pursued further. Rather than focusing on the perfect solution, rough feasibility checks and prototypes directly from the workbench (thus the “garage” metaphor) provide a good overall overview of upcoming efforts and challenges.
Here, too, a holistic approach is indispensable: in addition to the actual algorithms, the model quality and the required data quantity or structure, questions regarding changes in the way of working, integration into existing processes and implications for other process steps need to be discussed and clarified. Since the end-to-end implementation of AI use cases often induces considerable effort, working free from any bias and regarding failure as the foundation of success are essential during the MVP phase.
Digitization platforms for subsequent use and data linking
The first three building blocks already provide great potential for optimization and ensure quality, standards and measurability in processes and data. Additional benefits can be generated by linking data beyond the boundaries of your own organization. As soon as data and digital processes can overcome company boundaries through platforms, there is an immense potential for knowledge pooling, especially for medium-sized companies. Thus, it is possible to gain access to more extensive data and knowledge, which has so far only been available in large companies. Synergies also arise for all participants with regard to AI: Where a few data points within one company might not be sufficient for the use of artificial intelligence, the aggregation of data from different partners allows for the implementation of a strong AI model.
As a partner of the KIKS Ecosystem, HWI is a winner of the AI innovation competition of the BMWi. KIKS intends to link and analyze medical data across different data providers electronically in the future. The underlying platform is based on a digital ecosystem to be developed together with our partners, which will ensure compliance with legal and ethical regulations through state-of-the-art architecture and security technologies.
The goal is a digital ecosystem from which patients, healthcare providers, pharmaceutical companies and medical technology manufacturers will benefit equally.
HWIs Way into Digitization
As an innovative partner of the pharmaceutical, biotech and medtech industries, HWI exploits the potential offered by digitization and artificial intelligence, especially in the highly regulated GxP environment. We are positioning ourselves with digitally supported services whose high regulatory standards will also benefit our clients.
One pillar of our digitization strategy is the active participation in the KIKS ecosystem. Within the scope of the project, approaches for process automation in the field of pharmacovigilance and vigilance of medical devices are developed using agile methods.
For us, digitization is not only a means to an end, but part of our corporate strategy and always interwoven with the specific focus of our customers: in the end, the success and quality of each AI system depends on the expert training by our colleagues with their broad know-how.
Phone: +49 7272 7767-2564
Do you have any questions, suggestions or criticism about this article?
You are welcome to contact me.
Monday - Friday