Bill O’Kane, Vice President and MDM strategist at Profisee talks about the role played by advanced AI and ML technologies for a sustainable master data management
Some of the most promising themes within digital transformation revolve around new possibilities with Artificial Intelligence (AI), Machine Learning (ML), and the rise of the Internet of Things (IoT). The organizations that are best at encompassing these emerging technologies and the surrounding business models and services will position themselves at the front of the pack and will ultimately survive into the future.
However, that journey is fraught with potholes, detours, and blind alleys. One thing is clear, if organizations don’t have a firm grip on data quality, they will run into several obstacles that will seriously delay success.
If the AI processes that run on data is not unique, accurate, consistent, and timely, they will not produce reliable results and will lead to unwanted and extremely negative business outcomes.Worse, it can severely impact a company’s reputation all of which create reluctance on the part of the business to embark on new AI initiatives. Typical obstacles include:
- Making different decisions for two customer master data records that actually describe the same real-world party.
- Recommending a product to a customer who has previously returned a similar product.
- Accepting a purchase from a vendor whose company family has previously been rejected.
Rationalizing Data for ML, AI & IoT
Machine learning is the discipline used to ignite AI. As organizations look to support the machine with training data for ML, it may be tempting to cleanse the training data as it will not be part of continuous, future AI processing. The risk of this approach is that each training dataset would be cleansed little differently. Therefore, when more AI-supported business processes start to interlink, they will have a different way of thinking. The answer is to derive the training datasets from already-rationalized production data that all share the same master data foundation.
Likewise, Artificial Intelligence-supported business processes must also be based on master data that is unique, accurate, consistent, and timely to achieve positive business outcomes when applying AI. Taking the time to do so means the results will be reliable, the processes repeatable over time and the concepts will be reusable in other scenarios.
Lastly, the Internet of Things has enormous potential, as the use of interconnected smart devices continue to proliferate. Even more significantly, manufacturers will use smart machines that are connectable within the Industrial Internet of Things (IIoT) environment, sometimes referred to as Industry 4.0. In traditional data management, the product is classified by the model, i.e., a model of a refrigerator or drilling machine. When each produced instance of that product model becomes intelligent, organizations will realize requirements and opportunities for handling each instance — i.e., each occurrence of a thing or asset — as a master data entity, similar to opportunities that exist with customer master data today.
In the traditional master data set-up, the business would have to encompass relationships between customers and products like product models a customer has bought and which locations a customer (and other party roles) or product is connected to. With IoT, the number of relationships will increase exponentially. For example, a “thing” will have relations to many party roles, e.g., manufacturer, operator, maintainer, and owner. A thing will change locations, and an organization will need to be more precise about that location. A product will be comprised of many things with varying configurations produced as a given model with its basic specifications. In other words, things will begin to act like people.
While there is a plethora of technologies that claim to enable a digital transformation and help drive insights from AI, ML and IoT, one, in particular, has proven itself as a cornerstone to success; Master Data Management (MDM).
How MDM Provides Sustainable Data Quality and Encapsulates Complexity
The lifeblood of AI, ML and IoT is data, which must be circulated automatically. However, this is only possible when an organization’s data is of the highest quality. Master data that is used throughout many business processes, and for various scenarios, must meet a range of data quality dimensions to be sure it can underpin these automated processes.
If data quality is not continuously maintained, it can quickly decay at a given time, eroding any accuracy in the data and the insights the business is hoping to attain. Obviously, unmaintained, poor quality data will not be suitable in AI-supported business processes or for operating and analyzing the data IoT environments and analyzing the data. What’s most compromised in this challenge is the master data that describes the core entities involved in these business processes and environments. By leveraging MDM technology, companies can quickly and accurately onboard reusable master data that describes core entities and controls the lifecycle of that data.
As humans, we have a natural-born capability to understand the complexity of the core entities involved in business processes and data gathering, but machines must have a structured way of getting that understanding. MDM provide AI with an encapsulated description of the related core entities involved in business processes and how they are involved in connecting smart devices in IoT environments.
Enterprises must avoid the risk of starting too fast and producing unwanted business outcomes. This is because data quality cannot meet the need for fast circulation of unique, accurate, consistent, and timely data, which will remove obstacles that hinder long-term data-management success, including missing relationships among customers, products, assets, and locations — as well as critical data elements related to these entities. A true multidomain MDM platform can mitigate several risks around data protection and data privacy that can potentially block the way forward.
When MDM is applied correctly, the data only needs to be cleansed once to prevent data quality issues from reoccurring. Once done, the organization can then manage the complexity of the many data entities and their relationships throughout the deployment of AI-supported business processes. It can also be used to operate the IoT landscape and analyze its big data sources as well as rationalize the use of training datasets in ML.
To avoid delays and getting lost in utilizing AI, ML and IoT, companies must sustainably operate with high-quality data and provide shared, digestible digital representations of all the master data entities involved. Only then can they mitigate the risks ahead, remove the many obstacles, and ensure a successful and timely execution of a digitalization initiative.
For more such updates and perspectives around Digital Innovation, IoT, Data Infrastructure, AI & Cybsercurity, go to AI-Techpark.com.
