In the last decade, the topics of big data, machine learning (ML) and artificial intelligence (AI) have gone from being the domain of technical specialists to a strategic imperative across all dimensions of most business enterprises. The chemical and materials industries are transforming the way materials innovation is done – from early discovery to product development, through scale-up into manufacturing, to customer engagement and support.
The need of the hour is to harness the power of empirical data and fundamental modeling to predict product performance properties and innovate a new and more efficient era for manufacturing.
A Realized Example: Predictive Intelligence
My company, Dow, is celebrating its 125th anniversary this year. 125 years of existence equals 125 years of data. With the advent of AI and ML, we’ve put this big data to use in the creation of numerous predictive capabilities from Paint Vision in the Coating business to the Predictive Intelligence (PI) capability in our Polyurethanes business.
These capabilities place our vast material science expertise and immense data archive at the fingertips of our customers. Diving into the PI capability as an example, it completely transforms product formulation process by predicting formulation properties and simulating customer processes. For a single new formulation, there might be over 15 variables with 1000 different options each, leading to over a million data points to consider. The R&D process in the lab for a new formulation would take scientists months just in the discovery phase, but using the power of digitalization, our predictive modelling capability can take over for this discovery phase and drastically cut down the time to market by many months.
The Realization Journey
The vision and promise of predictive modeling is easy to embrace when you see the immense benefits of capabilities like PI in action. The actual journey that so many companies are on is complex and fraught with pitfalls and frustrations. As a passionate advocate for the digital transformation of innovation, I have personally been on a journey with several teams at Dow and we have learned lessons along the way which I would like to share for those interested in joining the digital revolution.
The need of the hour is to harness the power of empirical data and fundamental modeling to predict product performance properties and innovate a new and more efficient era for manufacturing
The starting orientation of technical professionals (whether R&D or IT or manufacturing) is usually to start with the domain expertise aspects of the problem. But this approach can be shortsighted, putting the cart before the horse. I believe the problem needs to be approached holistically across multiple interrelated elements.
• Business Outcomes: Digitalizing an innovation workflow of any scale and complexity is a MAJOR investment. Definition of what the system will enable through measurable value creation is fundamental to both the justification and design of that system. For example, with PI we are enabling a radical increase in speed of response to our customers so that our project capacity is increased, and our customers go to market faster.
• Data Collection and Management: At risk of stating the obvious, its ALL about the data and this is where the major component of investment come in. Typically, large datasets are required for predictive modeling of complex systems. Data acquisition needs to automated in labs that are often distributed around the world and sometimes using different equipment. And finally, the data needs to be curated to make it ‘modeling ready’ and minimize rework.
• Futureproofing: The digitalization journey often starts with ardent technical professionals launching grassroots efforts that gain momentum and grow into big ideas. When the digital tools are intended to be key capabilities for the business, the organizational capabilities to sustain and extend them must be put in place. Without that, they will die a death of accidental attrition as key early contributors move on.
• Collaborative culture: Functional siloes are massive decelerators of success. Successful predictive capabilities (meaning those that fulfill their business value) are a multidisciplinary team effort – probably more so than most initiatives that most of us will engage in during our careers. The diverse skills of teams compromising experts from IT, AI, chemistry and materials, marketing, communications, and sales all need to be aligned in service to the objectives of the program, requiring the establishment of an intensely collaborative culture that enables collective, enterprise-based work.
• Change Management. I believe that this last topic is often an overlooked element that determines the ultimate success of the digital investment. Because the journey is long and complex and because the culture change is massive, deliberate change management is critical. In addition to all of the technical dimensions, we are ultimately asking people to change the way that they work. The ultimate measure of success is adoption – both by employees and customers. People need to be supported to understand how to navigate and embrace new ways of working based on the impact on THEM as individuals and work groups.
Reflecting on the journey of Predictive Intelligence at Dow – we have learned and applied all of these lessons. I expect as we extend our scope to support our customers’ manufacturing operations, we will keep learning more. The most important lesson I have learnt in my digitalization journey is that it is never over, we need to be flexible and agile and keep looking for new solutions and new ways to create the future of our industry
