Elevating user experience in master data management with microservices and generative AI

Organizations must consider the right microservices-based architecture to support the capability to integrate generative AI models when thinking about their MDM solution.

Traditional MDM implementations have sub-optimal user experience

MDM platforms are not intuitive. They typically require extensive training before a data analyst can start using them. Often while designing MDM user experiences, the organization’s business process and the way a data analyst would use the product are not considered. This is partly because of the capabilities most MDM products offer.

The user experience is centred around the data model, and the data model design heavily influences the user experience. However, a typical interaction an end user has with the MDM platform is part of a business process. For example, the user might want to update a customer’s sales organization and may have no interest in the hundred other attributes the customer master has. Here lies the impedance mismatch.

Out-of-the-box MDM screen flows are not designed for organization-specific business processes since they focus on a lowest common denominator that would work across organisations. A typical customer master update screen flow would require the end user to navigate through multiple screens, even if they are looking to do a simple update. Consequently, the usage of MDM platforms is not democratized across the data analyst community and is restricted to a small set of trained users, which limits its usefulness.

It stems from monolithic MDM architectures

As part of our everyday lives, we interact with digital systems daily and  are used to intuitive user experiences. Whether we’re shopping online, browsing for a movie to stream or looking up a customer, we’ve become accustomed to user interfaces that don’t require training to use them.

That’s not the case with MDM. MDM solutions are typically sold as a bundle, which includes data integration tools, data models, data quality engines, workflows and many other components, including user interface. Organizations are constrained by what the product offers out of the box and have limited flexibility in designing a user experience that makes sense to them.

Let’s look at what’s being done to get around these limitations.

Step 1: Consuming web services in a modern UX framework

A paradigm shift is happening from monolithic architectures to microservices architectures. Many cloud-based MDM platforms are exposing their features as web services, which allows for API-based integrations and use case-based integrations. This allows for designing user experiences that makes sense for the business. Rather than relying on the out of the box user interfaces provided by the MDM platforms, organizations can use modern UX frameworks and design the screen flows tailored to their business processes.

Open-source frameworks such as React.js and Angular allow developers to build leading-class user experiences that would integrate with MDM platforms using microservices. This enables organizations to get the best of both worlds: creating high-quality master data in an intuitive system that needs no training.

Step 2: Enriching master data creation through generative AI

The next step in this evolution to high-end user experience is the advancement of generative AI technologies. ChatGPT has taken the world by storm and has taken assistive technology to the next level, reducing application development time and bringing powerful capabilities to non-technical users.

Let’s take the example of Material Master, which is a very common master data object for industrial organizations. When Material Master — such as a rotary equipment — is created in the MDM system, it is required to provide a description. This description field is extremely important in determining the quality of data, as this is used while detecting duplicates and searching master data. The quality of the description is typically tied to the person who inputs it and it takes time, and hence ends up being not very descriptive.

With a microservices architecture, organizations can now seamlessly invoke ChatGPT as a web service and ask the AI model to create the description. This AI model is trained on millions of records in online material catalogs, and can often generate an intuitive description, thereby saving time. Modern cloud-based MDM products already have the capability to make this happen using a microservices architecture.

The key enabler in MDM that makes generative AI technology feasible is its built-in data governance safeguards. A leading practice is to have a data steward in the middle to validate the results of AI, which takes substantially less time than a data analyst defining it by themselves.

Step 3: Embracing conversational interactions

Human civilization is built on conversations. With the advancement of generative AI technologies, we are now able to challenge the need for any screens at all and embrace conversations. With these conversational technologies — think Alexa or Siri — users could just say “Create a record for John Smith who lives on Main Street, works for Acme Inc. and is related to Jane Smith.” And the MDM system will create the customer records, the relationships and the hierarchies.

A well-designed microservices-based MDM solution integrated with a natural language processing model in conjunction with generative AI systems could make this happen.

Conclusion

MDM technologies have already proven their value for the vast majority of companies by delivering high-quality, trusted, single sources of truth. The one complaint has often been around the user experience. As organizations are thinking about the next generation of their MDM solution, proper thought should be given to designing the right architecture that would support a microservices-based architecture with the capability to integrate generative AI models so that it opens the door for building user experiences that data analysts would embrace.