How Are Assessment Offices Using Generative AI?

p strongGenerative AI has entered assessment work, but use varies sharply by institutional size and still requires clear rules about evidence and validation./strong/p p Primary Research Group surveyed assessment leaders at 63 colleges and universities. Respondents at the largest institutions reported using ChatGPT for a mean of nearly 2.5 hours during the preceding week. The report also examines Gemini and other tools alongside standardized testing, rubrics, curriculum maps and competency-based approaches./p p Assessment offices may use AI to summarize comments, draft survey items, classify open-ended responses, generate examples or help write reports. Those tasks do not carry equal risk. Student work, demographic data and accreditation evidence may be confidential, and automated summaries can erase minority views or create patterns not supported by the underlying record./p p A useful policy should define approved systems, prohibited data, human review and documentation. Staff should retain the original evidence, check generated claims against it and disclose material AI assistance in formal reports. Repeated tasks can be tested on known datasets before the tool enters a live workflow./p p AI can reduce administrative effort, but it cannot decide which outcomes matter or whether an institution's interpretation is educationally sound. The value comes from faster preparation that leaves more time for human analysis and action./p p  /p