Systems that use artificial intelligence to improve forecasting, planning and decision-making in companies have proliferated in recent years, but in many cases they lack detailed, specific information about the company itself, limiting the usefulness of these tools.
Devavrat Shah, a principal investigator at MIT’s Laboratory for Information and Decision Systems (LIDS), a faculty member in the Department of Electrical Engineering and Computer Science (EECS), and a member of the Institute for Data, Systems, and Society (IDSS), has focused on designing methods that enable second-by-second decision making with limited computational resources.
“In a sense, you have to do a lot of hard work with a small amount of resources,” he says. As a researcher, “my interest lies in the ability to develop methods to extract information from data at scale in the most effective way possible.”
The Andrew (1956) and Erna Viterbi Professor has been teaching at MIT since 2005.
In 2019, he also co-founded a spin-off company called Ikigai Labs. Based on years of research in Shah’s lab, Ikigai developed a foundational model for tabular time series data that was patented by MIT and licensed to the company. This model can ingest enterprise data from disparate sources continuously and at scale, allowing it to learn over time by testing its predictions against real-world results.
Shah explains that the system is an extension of the type of graphical models used, for example, by GPS devices to convert a sparse amount of data from satellites into an accurate model of a position on the Earth’s surface, or by communication systems such as that of a digital clock, which communicates at high speed and in an energy-efficient manner.
“My interest was: How do you design such graphical models for generic, tabular data?” he says.
While most AI models have been trained with text and images, this system uses tabular data as input – structured data like the familiar row and column format used in spreadsheets. And then it enables the kind of real-time planning on a much larger scale.
Ikigai’s idea was to provide forecasting and decision-making technology to large companies such as consumer goods manufacturers and pharmaceutical companies.
Shah gives an example of how a consumer electronics company could use this system.
“Let’s say you make headphones and all sorts of other things. And each of the products you make are made up of lots of little parts that come from different parts of the world. And once the device is sold, it needs to be supported and maintained. And you need to develop new versions of the product, you need to market them, you need to price them… So the questions you would normally ask would be: If I were to sell these in the next quarter or next year, how many would be sold in different places and what would happen to them.” Demand if I change the price or introduce a promotion?”
He adds that all of these processes are interdependent and decisions must be made at each stage of the processes that have an impact on time. “At some level,” he says, “digitizing these processes and being able to predict and constantly optimize ultimately leads to better business operations.”
Ikigai was recently acquired by the international company Celonis, where Shah is now a senior scientist in addition to his duties at MIT. Ultimately, he hopes the model he developed for Ikigai will help Celonis provide tools that integrate with a company’s own data and business processes to provide real-world analytics that can help create forecasts, plans and decisions.
Shah adds that Celonis specializes in digitizing and automating operations for more than 1,400 large companies worldwide. Now that these systems are fully digitized, they provide Ikigai Software with a platform to take the next step and read the data from these digitized systems to provide detailed models that can enable the simulation of various options, predict optimal strategies, and predict the outcomes of a specific set of decisions.
“Once the digital layer of these processes and that information layer are in place,” says Shah, “we can now layer the Ikigai stack on top to enable decision-making at a much larger scale than otherwise.”
While so many companies are working on various aspects of AI, “we are very focused on one part of the space that the rest of the world doesn’t pay attention to,” namely the area of structured or time-domain data. By building on such data, he believes it offers a very cost-effective version of AI.
“Narrower focus comes with sharper technology,” he says, “but it’s broad enough that it’s very valuable.”
Shah adds: “The current buzzword that has gained traction in the modern AI popular press is ‘world model’.” In a sense, this is trying to build the world model of corporate processes, so to speak.”
https://news.mit.edu/2026/helping-ai-models-meet-real-world-0714
