Systems that use artificial intelligence to enhance forecasting, planning, and decision making in businesses have been growing rapidly in recent years, but in many cases, they lack detailed, specific information about the organization, limiting the usefulness of those tools.
Devvrat Shah, a principal investigator at MIT’s Laboratory for Information and Decision Systems (LIDS), faculty member of the Department of Electrical Engineering and Computer Science (EECS), and member of the Institute for Data, Systems, and Society (IDSS), is focused on designing methods that can handle second-by-second decision making using limited computational resources.
“In a way, with little resources, you have to do a lot of the heavy lifting,” he says. As a researcher, “I’m interested in the ability to develop methods that can extract information from massive amounts of data 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 spinoff company called Ikigai Labs. Ikigai created a foundation model for tabular, time series data based on years of research in Shah’s laboratory, which was patented and licensed to the company by MIT. The model can take input from enterprise data from a variety of sources continuously and at scale, so it can learn by testing its predictions against actual outcomes.
Shah explains that this system is an extension of the type of graphical models that are used, for example, by GPS devices to convert small amounts of data received from satellites into accurate models of the situation on the Earth’s surface, or by communication systems such as digital watches that communicate at high speeds in an energy-efficient way.
“I was interested: How can one design such graphical models for simple, tabular data?” He says.
While most AI models have been taught using text and images, this system takes tabular data as its input – structured data such as the familiar type of row-and-column format used in spreadsheets. And then it provides real-time planning on a very large scale.
The idea of Ikigai was to provide forecasting and decision-making technology for large businesses such as consumer goods manufacturers and pharmaceutical companies.
Shah gives an example of how a consumer electronics company might use this system.
“Let’s say you’re making headphones and all kinds of different things. And each product you make has a lot of little pieces that come from different parts of the world. And once the device is sold, it needs support and maintenance. And you have to come up with new versions of the product, you have to bring them to market, you have to price them… So the questions you’ll typically ask would be: If I want to sell these next quarter or next year, how many in different locations Will be sold, and what if I change the price, or if I start a promotion and ask?
He says that all these processes are dependent on each other and at every stage of the processes decisions have to be taken which have an impact over time. “At some level,” he says, “digitalizing these processes and being able to make predictions and continuously adapt ultimately leads to better business operations.”
Ikigai was recently acquired by international firm Celonis, where Shah is now chief scientist in addition to his roles at MIT. Ultimately, he hopes the model he developed for Ikigai will help Celonis provide tools that can integrate with a company’s own data and business processes to provide real-world analytics that can help with forecasting, planning, and decision making.
Shah says Celonis has gained expertise in digitalizing and automating the operations of more than 1,400 large companies around the world. Now that these systems have become fully digital, they provide a platform for Ikigai’s software to take the next step, to read data from these digitized systems to provide detailed models to allow the simulation of different options, predict optimal strategies, and predict the outcomes of given decisions.
“Once the digital layer of these processes is in place and this information layer is in place, now, on top of that, we can put an ikigai stack to enable decision making on a much larger scale than we otherwise would,” says Shah.
While a lot of companies are working on different aspects of AI, “we’re very focused on a part of the domain that the rest of the world isn’t paying attention to,” which is the area of structured or time-domain data. He says that by starting with this kind of data, it provides a very cost-effective version of AI.
“A narrow focus comes with fast technology,” he says, “but it’s so broad that it’s very valuable.”
“A recent buzzword in the modern AI popular press is ‘world model,’” says Shah. In a way, this enterprise process is an attempt to model the world, so to speak.
