The Future (and Present) of ML and AI in Master Data Management
At the end of May 2017, my colleagues Henrik Baumeier and Christian Fuchs posted in our Digital Excellence Blog about how Artificial Intelligence (AI) will impact master data management (MDM).
Beginning of March I joined the Gartner Data and Analytics Summit in Grapevine, TX. I had very high expectations on the role that AI and Machine Learning (ML) will play at the conference, and I was not disappointed when I saw that the number of vendors that were promoting ML or AI capabilities. They increased from a handful to a few dozen. What was surprising for me, was how careful several Gartner analysists and the key note speaker were when they talked about the current state of AI and ML capabilities. They all agreed on the enormous potential and that AI/ML will change our way of working and living but they also saw this more in a long run future than I was expecting.
My surprise was based on the substantial number of vendors at the event with some capabilities in the AI/ML area but mostly on the advancements we made in the last year. Customer were reaching out to us for support in definition and realization of AI/ML based use-cases. We’re also seeing our customers building up internal teams that are specialized on AI and ML, and how they can be utilized to support data management. Then I consider what the CAMELOT data management experts were able to develop and simplify based on this promising new capability in such a short time.
Data management is changing and the changes are already happening in all areas from governance to technology.
I want to share some thoughts on how data management could be affected by AI. For this I want to tell you a short (silly and simplified) story about a friend, let’s call him Chris B., an employee of a large pharma company working in the marketing and sales department.
Material Master Data Maintenance in 20XX
Portfolio Management and Marketing have agreed Product X needs a new package size for the US market. Chris needs to request a new master data record for the product they want to launch as soon as possible. Until last year, he went to the MDM tool and was requesting the information he got from the existing products, artwork and the portfolio management by entering the information in the MDM system. Chris never liked that job, he does not do it frequently enough to be an expert and although he got some training he never really was sure about the data values and what they mean. On top of that, the system was slow and he did not understand the technical error messages, which popped up frequently. For these reasons, he always called Michael form the master data team to help him. Michael is a real MD expert and helped him to create his requests properly.
But in the beginning of this year things changed. Chris opened his company messenger and was connected to Scott. Scott is a virtual data maintainer – a chatbot. Chris is telling him what kind of product he needs by uploading the artwork and some other basic information he has. Scott is operating instead of Chris in the technical system. Scott will ask Chris if he needs additional information and Chris can ask Scott if he has questions about what specific values mean and how they are related to other values. Scott will also help him by putting up additional explanations and recommendations for values. Scott can do this because, Scott is not only a chatbot he is connected in real-time to the existing data bases. He is comparing similar materials and will alarm Chris in case of a possible duplicate.
Scott will predict values based on similar products and compare the values with the information on the artwork. Scott also is checking the consistency of data based on validation rules. The rule base Scott is validating against is dynamic and developing constantly, because of Scott’s machine learning and rule mining capabilities. While in the past, Chris’ colleagues from the master data governance team worked for years to define, test and implement rules that were limited to the complexity that the human mind could digest and the complexity the IT department was committing to implement in their system. Scott is identifying and developing rules daily. Scott checks with the master data governance team frequently and with his new suggestions, the system is improving with this combination of data analysis, comparison and expert input the data quality.
Data Governance Team’s Response
Chris’ colleague from the data governance team, Tina, has never been so happy. She can now focus on managing and improving the MD processes as well as improving Scotts capabilities by monitoring and helping him to learn. Tina gets recommendations on the lifecycle of the material and suggestions on when to adjust them based on transactional information like stock level and demands. With every cycle and every new request Scott provides better results. Tina was skeptical before but since she understands that she is monitoring Scott’s activities and each “mistake” Scott is making helps to improve his capabilities. Her job has changed, and is now responsible for more value-adding activities and less robotic tasks.
Are we ready for this? Where is the human interaction? Every time Chris called Michael in the past, Michael would tell him a new joke, and Chris loved that. He was always telling the joke in the evening to his kids. Next week Scott will get a new feature and will be able to ask Chris if he wants to hear a new joke. Scott will check a joke website and Scott will also ask Chris if he likes the joke. Since Scott wants to get better he will not tell Marie, Emma and Thomas the other marketing requester the same joke later, if it is not good. Scott will store the responses and will create a database of jokes that is recognized as funny by the employees. The plan for the next releases is that Scott will learn to understand human language better, to make it even easier for the requesters to ask for new materials. And to improve his accuracy in humor he will start to check the faces of the requester as well as finetuning his jokes based on other inputs like age, job profile, and timing (people have different moods during the day and these moods follow a predictable schema like Daniel Pink perfectly described in his keynote at the Gartner DA and his book “When: The Scientific Secrets of Perfect Timing (January 2018)”.
When this will happen?
If you ask me, it is not the maintenance process 20XX, it is the possible processes for 2018. All the components are available. We need to stop talking about AI and ML and start to exploit the benefits and realize solution. In times where Siri and Alexa managing our private communication and entertainment it seems a little bit outdated to let somebody maintain 300 material master data fields on a Web Dynpro UI. Will there be challenges for integration and trust? Yes! Will there be disruption? For sure! Will the companies mastering these challenges and the new technologies have a huge competitive advantage? Absolutely!
CAMELOT’s approach is to master these new capabilities, not just talk about them. We are working internally on solutions for our clients as well as with them in co-innovation projects. We have the objective to go from an idea to a prototype in less than 8 weeks.
The Global AI in MDM Community has already more than 35 leading global companies from all industries. Last week we conducted an AI in MDM design thinking workshop in Europe and we will conduct a second design thinking WS in the USA in April.
More than 30 use cases are currently in our innovation tunnel, dozens of them already realized in a prototype like our AI showcase leveraging IBM Cloud Machine Learning capabilities (Natural Language Processing). Integration to MDG/Fiori App chat bot or an own developed ML tool that predicts the lifecycle of a product based on demands and sets derives the material status accordingly.
Join us on our mission – stop talking about AI – create solutions that improves your business and makes you a leading innovator .