After companies have finally realized that their businesses run on data and CIOs and senior management have heavily invested in building up Master Data Management (MDM) capabilities, the next wave of CIO topics is at hand.
We are right in the middle of digitalization and new buzzwords like smart dust, edge computing, digital twin, artificial intelligence (AI), and machine learning (ML) are filling the market. At the same time, the amounts of data generated are increasing massively and along with it the need to structure these data and make them understandable and interpretable for decision processes, etc.
Companies like Google and Amazon have already been using ML for decades. However, also in the traditional industries and B2B business first projects have been successfully set up and first results are already visible. Optimized complex and energy-intensive production processes based on patterns within a lot of different variables, for example, resulted in a stabilized reduced energy consumption during the production process. Another example is an AI that learned when to initiate an emergency alert based on different parameters like body symptoms (e.g. heart rate, breathing, etc.) as well as the position of a person within an apartment (e.g. lying on the kitchen floor vs. lying on the bed). These are just two examples of how AI/ ML is already in use today.
Is MDM still worth an investment or already out of fashion?
In reference to the AI model from a former blog post, we are at the verge of the 1st stage for AI and machine learning for Master Data Management purposes. Assuming that AI and machine learning are the future and AI algorithms will control and manage master data for us without any/much manual intervention – the equivalent to 3rd stage in above mentioned model –, does it still make sense to invest in Master Data Management today? How can AI and machine learning be used or be of use in Master Data Management anyways?
As mentioned in “Artificial intelligence in master data management – it’s starting now!”, there are many use cases for artificial intelligence and machine learning in MDM. If you want to have more information the PoC concept, please read the article Tackling AI in MDM with a collaborative innovation approach.
Here are just some of the AI-MDM use cases that CAMELOT is currently focusing on:
Machine learning-enabled data determination for migration initiatives
Situation: Migrations often happen when new systems are introduced or as a result of mergers & acquisitions and always entail significant data cleansing and transformation efforts.
Approach: An algorithm will analyze the data of the target structure, build decision trees, and then determine values for the new records to be migrated from the old system(s). The idea is to determine the maximum amount of field values by answering a minimum amount of questions.
Benefit: Reduction of migration cleansing and transformation efforts.
Master data outlier detection
Situation: Master data are spread across various tables with no knowledge of data dependencies (e.g. when maintaining a specific combination of attributes, another set of attributes is always the same).
Approach: Identification of company-specific patterns within the master data and master data structures.
Benefit: Identification of
- company-specific patterns: global, per plant, per sales org.
- classifications whether a record is an outlier or not
- combinations of fields/values that make it an outlier
- value recommendations to ensure material conformity and identification of SKUs that don’t fit to the pattern (outliers).
Ontological model for standard operating procedures (SOP) for validation & support
Situation: Documents are used in all companies, e.g. technical specs, product information sheets, certificates, etc. These documents contain information which often has to be entered into the systems manually.
Approach: The machine learning algorithm analyzes documents and transforms the information (e.g. guidelines, context information, product information, specifications, etc.) from these forms into an ontological model that can be read by MDM applications.
Benefit: Enhanced user support, metadata is populated during document upload without manual effort.
These are just some examples of AI applications for MDM brought to live by our AI in MDM Community. To learn more about this topic, join the AI in MDM Community.
First steps are already being taken
The current initiatives to realize AI/ML for MDM show that we are still in an early phase, but it is no hype in this context anymore as concrete actions are being taken as first steps towards the future of AI in MDM. Some of the community members are already hiring data scientists in their MDM teams, which shows their commitment to the vision.
Thus – to pick up the question asked earlier – does it still make sense to invest in Master Data Management today? Definitely yes! AI and machine learning are the future and their development will eventually lead to a world where AI algorithms will control and manage master data for us without much manual intervention. By investing in MDM and in high-quality data today, you are getting ready for whatever future developments in digitalization will come and each implementation of AI and machine learning will benefit from the investments in data quality you make today.
Also, even companies that follow a different innovation strategy as early followers / late follower and decide against investing in proof of concepts today, should still consider ways to make their information more machine-readable to prepare for the future today. Small changes in the way information is handled are already a meaningful contribution to your company’s successful future.
Click here to discover further application scenarios for AI in MDM.