I study professionals, using technology as a lens. New technologies, especially AI, are often studied for efficacy and effectiveness. However, I believe new technologies are also useful for what they reveal about professionals rather than for what they do directly.
Drawing on organizational theory, medical sociology, and science and technology studies, I treat moments of technological change as opportunities to study how expertise, status, identity, and expert decision-making actually work.
Automating a peripheral task revealed that expertise is built through the work surrounding it, not just the core judgment itself.
My job market paper asks what happens to expertise when the peripheral work supporting it disappears. Over 18 months observing radiation oncologists adopt AI-assisted contouring, I found that this supporting work is not just overhead—it is part of how expertise works. When AI took over organ contouring (recognized as non-core by physicians and leaders), two things were lost: the mental space that low-stakes tasks created for insight, and the hands-on anatomical practice that built skill. These opportunities were integral to problem refinement and solution tailoring. When AI worked well, these moments disappeared from the scaffold of expertise enactment. (When Supplemental Matters — preparing for ASQ)
Automating even routine scheduling revealed how much patient knowledge nurses carry for the whole team through the context they pick up along the way.
A related project goes a step further: even when AI only handles something as routine as scheduling, nurses lose the patients' social, personal, and clinical context they used to pick up in the process. That loss doesn't stay with one nurse—it cascades into the whole team's understanding of the patient. (Working Paper — in progress)
Adding AI to a clinical team revealed that authority isn't a fixed pie—it can be extended to fill gaps without taking anything from those above.
So far these projects are about what gets lost when work is taken away. A randomized controlled trial in a clinical setting lets me ask the opposite question: what happens when something is added? Here I found that authority isn't a fixed pie. By building physician authority into its recommendations, AI gives non-physician staff what they need to act with full authority—filling in the gaps—without taking anything away from the physicians above them. Expertise loosens from rank, and the hierarchy expands rather than simply flattening. (The Impact of AI on Clinical Decision Making in Healthcare Teams — under review)
Reorganizing how records are presented revealed how clinicians build a patient's "data double," and how it normally encodes authorship and organizational time.
Introducing AI to reorganize how patient records are presented also revealed something about professionals: how they build a patient's "data double." Normally that double carries the authority of whoever wrote each note and follows the organization's sense of time, marked by events like hospitalizations. But the AI presentation shifted both—clinicians began attending to the content of a note rather than who wrote it, and the timeline reorganized itself around the patient's accumulating symptoms rather than organizational milestones. (From Data to Patient — submitted to JAMIA)
Anticipating AI's arrival revealed the medical profession's hidden status architecture: physicians respond not to the technology itself but according to where they sit in the hierarchy.
A second strand uses AI adoption as a lens on the status architecture of the medical profession. In an interview study, I find that physicians don't respond uniformly to AI's anticipated arrival—their reactions track their position in the hierarchy. US medical graduates protect authority through traditional gatekeeping; international graduates treat AI as a chance to legitimize themselves; women orient toward local, patient-centered standing while men focus on profession-wide markers. AI doesn't threaten everyone equally—it surfaces and reorganizes competitions over standing that were already there. (Same Threat, Different Futures — preparing to submit to Organization Science, with Ambar La Forgia and Eliza Brown)
The ethics framing around clinical AI revealed a quiet shift in professional identity: doctors moving from moral agents who exercise judgment to clerks who carry out an algorithm's recommendations.
Finally, I turn to AI and the moral side of professional life—accountability, judgment, and ethics. In an opinion piece, I argue that the way clinical AI is framed as an ethics problem quietly erodes physicians' moral agency, shifting doctors from moral agents who exercise judgment to clerks who execute an algorithm's recommendations—with consequences for professional identity and patient trust that have drawn too little attention. (From Moral Agent to System Clerks — submitted to JAMA Perspectives, with Julia Adler-Milstein, Margaret Eby, and Julian Hong)
This research has been supported by the Fisher Center at UC Berkeley, the Berkeley Center for Workplace Culture and Innovation, and the Institute for Research on Labor and Employment (IRLE) at UC Berkeley. Papers from this agenda have been accepted for presentation at the ASA 2025 Annual Meeting (CITAMS-OOW Panel on Artificial Intelligence in the Workplace), the Organization Theory in Healthcare Association Conference 2025 (invited podium), the NBER Digital Economics and AI Tutorial, and the MIT Rising Scholar Conference.
You can also review the ASQ blog where Leticia Smith and I interviewed Nishani Bourmault and Michel Anteby.