Singapore researchers reported improved user ratings after redesigning a chatbot that helps academic medical faculty navigate promotion and tenure rules.
The evaluation was a small pilot, not proof that AI gives more accurate career advice.
What researchers studied
The SingHealth Duke–NUS Academic Medical Center has about 3,000 clinical faculty members and handles roughly 400 appointments and promotions each year.
Relevant guidance is spread across around 30 documents, creating a practical information-search problem.
Researchers developed AskADD, a chatbot intended to answer questions about institutional promotion and tenure requirements.
Thirty-five faculty and staff recruited by email tested an initial prototype and subsequently evaluated a revised version based on feedback.
The first version used Microsoft low-code tools; the next used a custom generative AI system on an institutional platform.
The team analysed structured user-survey responses descriptively, focusing on perceived usability, responsiveness, empathy, accessibility and whether questions were resolved.
What they found
The proportion of participants classified as promoters increased from 23% to 51% between the two evaluation phases.
Respondents also reported improvements in usability, empathy and query resolution.
Participants particularly valued answers that were relevant to their own circumstances, required little rephrasing and linked directly to authoritative policy documents.
They wanted promotion guidance and human-resources information to be easier to find together.
Around half remained passively satisfied, rather than enthusiastic.
Some wanted more personalised assistance, including analysis of gaps between their CVs and promotion requirements.
These are user perceptions, not independently measured improvements in promotion outcomes.
What it means
The pilot illustrates how an organisation can use staff feedback to refine an internal AI information service.
Providing links back to source policies may make answers easier to check and support appropriate human oversight.
For universities and employers, a useful design lesson is to prioritise trustworthy retrieval and clear references rather than conversational polish alone.
Institutions would still need controls for outdated rules, confidentiality and misleading answers before relying on such systems for consequential decisions.
Important limitation
Only 35 volunteers from one academic medical centre took part, and the analysis relied on self-reported assessments.
The two versions were evaluated sequentially without a randomised comparison, so the higher promoter proportion cannot confidently be attributed to the redesign alone.
The authors describe the work as preliminary proof of concept and call for validated measurement instruments and objective performance testing.
The findings should not be generalised to other institutions or taken as evidence that chatbot advice improves promotion decisions.
Source Information
Study: Developing and evaluating an AI-assisted chatbot for promotion and tenure guidance in an academic medical center
Authors: May May Yeo; Deborah D. Navedo; Patrick J. Casey; Bernard C. Y. Tan
Journal / Source: Frontiers in Artificial Intelligence
Published: 5 November 2026
