While AI was once framed primarily as a frontier technology, it is now firmly embedded in our daily workflows and strategic decision-making across industries. 

While traditional computational tools have played a part in structuring data and speeding up analysis, large language models and agentic AI systems have turbocharged research – including at Columbia Business School.

Across the university, CBS faculty are using AI not just as a tool, but as a collaborator, helping to design studies, surface patterns, and write code. To understand what this shift looks like in practice, we asked CBS faculty how AI fits into their research workflows.

This article is part of a series in which CBS faculty reflect on the past year in AI.

CBS: How are you using AI in your research?

Much of my recent work uses AI and machine-learning tools to measure economically important but previously hard-to-observe constructs, such as quality and organizational capital, and to study how employers and employees respond to the rapid diffusion of generative AI. 

In my recent paper, Employer and Employee Responses to Generative AI: Early Evidence, we use a forward-looking, task-level measure of exposure to GenAI to examine how firms adjust hiring and skill demands, and how employees update their perceptions, career choices, and mobility in response to generative AI.

In another paper, Measuring Quality, we applied machine-learning models to millions of employee-written reviews to construct a dynamic, firm-year measure of quality. The AI-based measure predicts future product recalls, brand value, and profitability, demonstrating how text-based models can reveal early signals of firm-level performance.

  • Wei Cai, Assistant Professor of Business in the Accounting Division

I use generative AI tools to overcome “the tyranny of the blank page”—the sometimes-daunting task of initiating a project. Gen AI can be useful in producing a quick survey of the key literature, voices, and arguments that could be addressed in answering a question. On any individual point, these surveys are rarely insightful, but they helpfully sketch out the scope of ideas that ought to be addressed and complemented in a focused line of inquiry.

  • Brett House, Professor of Professional Practice in the Economics Division

I often discuss blue-sky ideas with AI. Not only is the AI very helpful for dissecting what's interesting and finding related references, this process itself is useful because it forces me to clearly express my ideas in a conversation. I often refer back to past conversations with AI to remember how I got to certain ideas.

  • Will Ma, Roderick H. Cushman Associate Professor of Business in the Decision, Risk, and Operations Division

AI now plays a role in every stage of my research: brainstorming ideas, conducting literature reviews, drafting and revising papers, writing and debugging code, and running experiments. It has become hard to imagine doing research without these tools.

  • Tianyi Peng, Assistant Professor of Business in the Decision, Risk, and Operations Division

I use AI for literature reviews and bibliographies, as well as re-writing text and bouncing around ideas. In my personal life, I use AI for parenting advice, recipes, and travel.

  • Laura Veldkamp, Leon G. Cooperman Professor of Finance & Economics in the Finance Division

Researchers, including myself, have long used AI, particularly natural language processing techniques, for structuring unstructured (often textual) data, but these uses were fairly limited. Recent advances in large language models now allow coding constructs from textual data that are ever expansive, for example, in my work, measuring firms' strategic alignment and entrepreneurs’ perceived cost of mistakes . LLMs are also incredibly useful intervention tools in experimental settings. For example, I have employed them to create near and far international shocks for entrepreneurs by training chatbots to be customers from different countries; entrepreneurs incorporating customer feedback into pitches; and tools lowering the cost of producing signals to investors and employers.

  • Nataliya Wright, Assistant Professor of Business in the Management Division