Students in a university setting utilizing digital interfaces for academic assessments and oral examinations.
NEW YORK · May 7, 2026
The New York University Stern School of Business has successfully piloted an AI-driven oral examination system, marking a significant shift in how higher education validates student knowledge. As generative AI makes the traditional written essay increasingly obsolete as an assessment tool, NYU’s pilot offers a scalable solution to the growing crisis of academic integrity. By utilizing specialized voice models to conduct real-time "viva voce" exams, the university is moving away from static homework toward interactive, unscriptable defense of work.
The Economics of Accountability
The pilot involved a 36-student undergraduate course where an AI system reviewed each student's submitted project before conducting a tailored 25-minute oral exam. The AI questioned specific claims and reasoning within the student’s own work, making it impossible to rely on pre-generated text. The most disruptive aspect is the cost: the full cohort was assessed for just $15, or approximately 42 cents per student. Compared to the $750 average cost of hiring human Teaching Assistants to conduct similar assessments, the AI model provides a level of rigor previously impossible for large-scale undergraduate classes.
Cross-Model Verification
Technologically, the system leverages ElevenLabs to provide a natural-sounding voice interface, reducing the "uncanny valley" effect during high-stakes testing. To ensure grading fairness, NYU employed a multi-model consensus system. Three separate AI models: Claude, Gemini, and ChatGPT: independently reviewed the transcripts of the oral defenses to generate scores. These models then reviewed each other's evaluations to reach a final grade. While 70% of students felt the exam was more stressful than a written test, a majority agreed it was a more accurate measure of their actual understanding of the course material.
Expanding Beyond Engineering
Following the success at the Stern School, NYU is preparing to expand AI-powered oral exams into the humanities and social sciences for the 2026-2027 academic year. Unlike engineering or business, where answers are often quantitative, the humanities expansion will utilize AI "avatars" programmed with greater emotional intelligence to handle nuanced debates. This transition suggests that the ancient Oxford-style oral exam is not only returning but becoming the new standard for digital-age pedagogy, ensuring that a degree represents human comprehension rather than a well-worded prompt.
