USC researcher: Generative AI is compressing research timelines and reshaping the scientist’s role
Artificial intelligence is beginning to move from lab assistant to the center of the scientific method, compressing work that once took months into days, according to Mayank Kejriwal, a research associate professor at the University of Southern California and CEO of Grail.
Speaking with host Keith Shaw on the program Today in Tech, Kejriwal said advances in generative AI are enabling systems to help generate hypotheses, design experiments, and interpret results—tasks traditionally handled by human researchers.
He said the shift is already changing research workflows in sectors from enterprise IT to scientific laboratories, offering speed and scale while introducing new challenges around validation, oversight, and trust. Kejriwal traces his view to hands-on experience.
He has worked in AI for nearly 15 years, beginning as an undergraduate at the National Center for Supercomputing Applications in Illinois, earning a PhD from the University of Texas at Austin, and joining USC in 2016. At USC, he runs the AI and Complex Systems lab, focusing on applied, empirical work on complex problems, including combating human trafficking.
About a year ago, he and his students began using tools such as Cursor and Codex. The result, he said, was a rapid acceleration: projects that once took six months dropped to six weeks, then six days. Convinced that “AI for science is real” and could make progress exponential, he said that momentum led to the creation of Grail.
The turning point came more recently. After newer models emerged, Kejriwal and his team asked whether they could build an AI “scientist” to take on more of the research cycle. Testing the approach himself, he used the system to pursue ideas he had shelved for years.
“I completed one of those ideas—ran experiments and wrote large portions of it—in less than a day,” he said, adding that with another day or two, it could be ready for submission. That speed-up, Kejriwal noted, also raises philosophical and practical questions about the role of humans in discovery—whether researchers shift from execution to judgment and direction, and how organizations ensure results are reliable.
As AI tools grow more capable, he said, business and technology leaders will have to rethink research workflows and guardrails. The core question, as framed in the conversation: Is AI simply improving productivity, or beginning to redefine how discovery itself happens?