Enhancing Corporate Forecasting with AI: The Vision of Devavrat Shah
Systems leveraging artificial intelligence (AI) for enhanced forecasting, planning, and decision-making have significantly evolved. However, many lack the intricacies of specific company data, limiting their effectiveness. Enter Devavrat Shah, a principal investigator at MIT’s Laboratory for Information and Decision Systems (LIDS) and a faculty member in the Department of Electrical Engineering and Computer Science (EECS). Shah, who also contributes to the Institute for Data, Systems, and Society (IDSS), is pioneering methods for second-by-second decision-making with minimal computational resources.
“In a sense, you have to do a lot of hard work with a small amount of resources,” Shah explains. His research focuses on extracting information from data at scale in the most efficient way possible, thus advancing the capabilities of AI in a resource-constrained environment.
The Andrew (1956) and Erna Viterbi Professor, Shah has been inspiring minds at MIT since 2005. His expertise and commitment to innovation also led to the co-founding of Ikigai Labs in 2019. This venture, based on years of research in Shah’s lab, developed a foundational model for tabular time series data, patented by MIT and licensed to Ikigai.
Revolutionizing Data with Graphical Models
The system Shah devised is an extension of graphical models akin to those used by GPS devices, which convert sparse satellite data into precise location information, or by efficient communication systems. “My interest was: How do you design such graphical models for generic, tabular data?” he elaborates.
Unlike most AI models trained on text and images, Shah’s system utilizes tabular data—structured data in spreadsheets’ familiar row-and-column format. This approach enables real-time planning on an unprecedented scale, offering powerful solutions to large enterprises like consumer goods and pharmaceutical companies.
Practical Applications and Real-World Impact
Consider a consumer electronics company manufacturing diverse products, each comprising numerous components sourced globally. Post-sale, products require support and maintenance, and new versions must be developed and marketed. Shah’s system addresses questions like predicting sales volume across regions, understanding demand fluctuations due to pricing changes, and forecasting the impact of promotions.
By digitizing these processes, Shah emphasizes, companies can predict and optimize operations continuously, resulting in superior business outcomes. Ikigai’s recent acquisition by Celonis positions Shah as a senior scientist, where his model aims to integrate seamlessly with corporate data and processes, enhancing real-world analytics for strategic decision-making.
Celonis, renowned for digitizing and automating operations for over 1,400 large companies globally, now provides a platform for Ikigai Software. This integration facilitates detailed modeling and simulation of various options, enabling optimal strategy prediction.
“Once the digital layer of these processes and that information layer are in place,” Shah notes, “we can now layer the Ikigai stack on top to enable decision-making at a much larger scale than otherwise.”
Shaping the Future of AI with Structured Data
In a world where AI development is diverse, Shah’s focused approach on structured or time-domain data offers a cost-effective AI model. “Narrower focus comes with sharper technology,” he asserts, “but it’s broad enough that it’s very valuable.”
Shah identifies the modern AI buzzword, “world model,” as akin to creating a world model of corporate processes. His approach underscores the potential of structured data to revolutionize AI’s role in corporate decision-making, marking a transformative step in the digital era.
For more details on this innovative approach, visit the original source Here.
“`
