Systems that use artificial intelligence to enhance forecasting, planning, and decision-making in companies have proliferated in recent years, but in many cases they lack detailed and specific information about the organization 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), focused on how to design methods that can handle second-by-second decision-making using limited computational resources.
“In a sense, with a small amount of resources, you have to do a lot of heavy lifting,” he says. As a researcher, “my interest is in being able to develop methods that can extract information from data at a large scale and in as efficient a way as possible.”
Professor Andrew (1956) and Erna Viterbi have been teaching at MIT since 2005.
In 2019, he also co-founded a subsidiary called Ikigai Labs. Ikigai built a basic model for tabulated time series data based on years of research in Shah’s lab, which was patented and licensed by MIT to the company. This model can take input from enterprise data from a variety of sources, continuously and at scale, so that it learns as it goes by testing its predictions against real 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 the location on the Earth’s surface, or by a communications system such as that found in a digital watch that communicates at high speed and in an energy-efficient manner.
“My interest was: How does one design such graphical models for general tabular data?” He says.
While most AI models are taught using text and images, this system takes tabular data as its input — data structured like the familiar type of row-and-column format used in spreadsheets. It therefore provides this type of planning in real time, on a much larger scale.
The idea of Ikigai 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 uses this system.
“Let’s say you make headphones and all sorts of different things. And each of the products you make has a lot of little parts that come from different parts of the world. And once you sell the device, you have to support it and maintain it. You have to come up with new versions of the product, you have to market them, you have to price them… So the questions you would typically ask are: If I were to sell these devices in the next quarter or next year, how many would they sell in different places, and what would happen to the order if I changed the price, or if I offered an upgrade?”
He adds that all of these processes are interconnected, and at each stage of the processes decisions must be made that have implications over time. “At some level, the digitization of these processes and the ability to make predictions and continually improve is what ultimately leads to better business processes,” he says.
Ikigai was recently acquired by Celonis International, where Shah now serves as 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 data and business processes to provide real-world analytics that can help make predictions, plans and decisions.
Shah adds that Celonis specializes in digitizing and automating processes for more than 1,400 large companies around the world. Now that these systems are fully digital, they provide a platform for Ikigai software to take the next step, reading data from these digital systems in order to provide detailed models to allow simulation of different options, predict optimal strategies, and predict the outcomes of a given set of decisions.
“Once we have the digital layer of these processes and this information layer, we can now, on top of that, put the Ikigai package to enable decision making at a much broader scale than anything else,” says Shah.
While many companies are working on different aspects of AI, “we’re very focused on a part of the field that the rest of the world doesn’t care about,” which is structured or time-domain data. By starting with this data, he says, it provides a very cost-effective version of artificial intelligence.
“A narrower focus comes with more precise technology, but it is broad enough to be of great value,” he says.
“The recent buzzword that has become relevant in the modern AI popular press is ‘global model.’ In a sense, this is trying to build a global model of corporate operations, so to speak,” Shah adds.
(Tags for translation) MIT IDSS
