In recent years, systems utilizing artificial intelligence to enhance forecasting, planning, and decision-making in businesses have proliferated. However, many of these systems lack detailed information about the organization itself, limiting their utility. Devavrat Shah, a principal investigator at MIT’s Laboratory for Information and Decision Systems (LIDS), has focused on designing methods for second-by-second decision-making with limited computational resources. He states, "With a small amount of resource, you have to do a lot of heavy lifting." His research aims to develop methods that effectively extract information from data at scale.
Shah has been teaching at MIT since 2005 and co-founded Ikigai Labs in 2019. Ikigai developed a foundational model for tabular time series data based on research from Shah's lab, which has been patented and licensed by MIT. This model continuously learns from enterprise data across various sources by testing its predictions against real outcomes.
Shah describes this system as an extension of graphical models used by GPS devices to convert sparse satellite data into accurate location models, or by digital watches for high-speed, energy-efficient communication. His focus was on designing graphical models for generic tabular data. Unlike most AI models trained on text and images, this system uses structured tabular data, enabling real-time planning on a larger scale. The goal of Ikigai is to provide forecasting and decision-making technology for large enterprises, such as consumer goods manufacturers and pharmaceutical companies.
For instance, a consumer electronics company might utilize this system to predict product sales and assess the impact of pricing changes or promotions. Shah emphasizes that digitizing these processes and optimizing predictions leads to better business operations. Recently, Ikigai was acquired by Celonis, where Shah now serves as chief scientist.
He hopes that the model he developed will help Celonis integrate with companies' data and processes to deliver real-world analysis for forecasts and decisions. Celonis specializes in digitizing operations for over 1,400 large companies worldwide, providing a platform for Ikigai’s software to analyze data from these systems, enabling simulation of options, predicting optimal strategies, and forecasting outcomes.
Shah points out that while many companies are exploring various aspects of AI, they focus on structured or time-domain data, which offers a cost-effective AI solution. He concludes that a narrower focus leads to sharper technology while still providing broad value. The recent buzzword in AI discussions is the "world model," which Shah interprets as an attempt to build an enterprise process world model.
Blogger's Review: Shah's research highlights the potential of AI in complex business decision-making, particularly how structured data can optimize processes. Ikigai's integration of multi-source data showcases AI's ability to enhance real-world forecasting and planning, making it a noteworthy development in the industry.