Most artificial intelligence systems designed to improve business forecasting and decision-making suffer from a fundamental limitation: they lack access to the granular, organization-specific data that would make their predictions genuinely useful. A new approach developed at MIT’s Laboratory for Information and Decision Systems (LIDS) directly addresses this gap, offering a method that can process structured enterprise data at scale and in real time, turning raw information into actionable business intelligence.
The Problem with Generic AI for Business Operations
Devavrat Shah, a principal investigator at LIDS, faculty member in MIT’s Department of Electrical Engineering and Computer Science (EECS), and member of the Institute for Data, Systems, and Society (IDSS), has spent years focused on designing methods that can handle second-by-second decision-making using limited computational resources. “In a sense, with a small amount of resource, you have to do a lot of heavy lifting,” he says. His research interest centers on “the ability to develop methods that can extract information from data at scale in as effective a manner as possible.”
Shah, who holds the Andrew (1956) and Erna Viterbi Professorship and has been teaching at MIT since 2005, recognized that while AI models have become remarkably capable at processing text and images, the structured data that powers most business operations — the familiar row-and-column format of spreadsheets and databases — remained underserved by modern AI approaches.
From Graphical Models to Enterprise Data
The technical foundation for this work draws on graphical models, a class of statistical methods already used in applications such as GPS navigation, where sparse satellite data is converted into an accurate position on the Earth’s surface, and in communication systems like digital watches that transmit data at high speed while conserving energy. Shah’s core question was: “How does one design such graphical models for generic, tabular data?”
In 2019, this research led to the founding of Ikigai Labs, a spinout company that built a foundation model specifically for tabular and time series data. The model, based on years of research in Shah’s lab and patented and licensed by MIT to the company, ingests enterprise data from multiple sources continuously and at scale. It learns iteratively by testing its predictions against actual outcomes, refining its understanding of the business processes it models.
How the Ikigai Model Works
Unlike most AI systems trained on text or images, the Ikigai model takes structured tabular data as its primary input. This allows it to model the interdependent relationships that exist across business operations — supply chain logistics, inventory management, pricing, demand forecasting, and customer support — and provide real-time planning at a scale that traditional methods cannot match.
Shah illustrates the system’s capabilities with a consumer electronics company manufacturing headphones and other devices. Each product contains components sourced from different parts of the world. Once sold, the device requires support and maintenance, and the company must develop new versions, market them, and set prices. “The questions you would typically ask would be: If I were to sell these next quarter or next year, how many will be sold in different places, and what would happen to demand if I change the price, or if I introduce promotion?” Shah explains. All of these processes are interdependent, and decisions at every stage have implications over time. “Digitizing these processes and being able to do predictions and constantly optimize is what leads to ultimately better business operations.”
What Is the MIT AI Model for Business Decisions?
The MIT-developed model is a foundation model for tabular and time series data that uses graphical model techniques to learn from structured enterprise data, enabling real-time forecasting, simulation, and decision optimization. It is designed to work with the kind of row-and-column data found in spreadsheets and business databases, distinguishing it from AI models trained on text or images. The model was commercialized through Ikigai Labs, which was recently acquired by Celonis, where Shah now serves as chief scientist alongside his roles at MIT.
Celonis Acquisition and Enterprise Integration
Ikigai was recently acquired by Celonis, an international firm specializing in process digitization and automation for more than 1,400 large companies worldwide. Shah now serves as chief scientist at Celonis in addition to his MIT roles. The acquisition positions the Ikigai technology to integrate directly with Celonis’s existing digital infrastructure, which already provides a fully digitized operational layer for its clients.
“Once the digital layer of these processes exists and this information layer exists,” Shah says, “now, on top of it, we can put the Ikigai stack to enable decision-making at a much larger scale than otherwise.” The combined system reads data from these digitized operations to build detailed models that allow companies to simulate different options, predict optimal strategies, and forecast the outcomes of specific decisions.
A Cost-Effective, Focused Approach to AI
While much of the AI industry pursues general-purpose models with ever-larger parameter counts, the Ikigai approach deliberately targets a narrower domain. “We are very much focused on part of the domain that the rest of the world is not paying attention to,” Shah says, referring to structured and time-domain data. This narrower focus, he argues, “comes with sharper technology, but it’s broad enough that it’s very valuable.” The approach also provides a cost-effective alternative to massive general-purpose models, making sophisticated AI-driven decision-making accessible to a wider range of enterprise use cases.
Shah describes the system as a form of “world model” for enterprise processes. “The recent buzzword that’s become pertinent in the modern AI popular press is a ‘world model,'” he says. “In a sense, this is trying to build the enterprise process world model, so to speak.”
What This Means for Business Decision-Makers
For organizations already running digitized operations, the integration of the Ikigai model through Celonis offers a path to predictive and prescriptive analytics that is grounded in the company’s own data and processes. The technology is particularly relevant for large enterprises in manufacturing, pharmaceuticals, and consumer goods — sectors where supply chain complexity, demand variability, and interdependent decision-making create a clear need for models that can simulate and optimize across multiple time horizons. Decision-makers should evaluate whether their current systems provide the digital layer needed to support this kind of AIaaaa-driven planning, and consider how structured data from their own operations could feed into models purpose-built for tabular and time series analysis rather than repurposed from text and image domains.
