LO
PilotCognitive automation2025-11-27
A medical school's AI tool flagged policy-restricted wording in grant drafts with full precision but missed about a quarter of terms, its developers reported
Grant writer / grants officeroccupation page →
Event date / reported
2025-11-27
Evidence stage
PilotSmall-scale trial in a real setting. Tells us the deployment conditions are being tested, not that they hold — so one pilot is never enough on its own; two independent ones are.
Tasks this bears on
Budgets, rules and compliance
Building budgets, meeting each call's eligibility and formatting rules, and declaring what funders require, including AI use.
Being augmented✓ Evidence-backed
Where this applies
United States. A medical school built a secure tool combining rules and language models to detect grant language restricted by evolving state and federal policies and suggest alternatives; it reached precision 1.00, recall 0.73 and F1 0.84, ahead of general-purpose models, and 25 faculty and staff rated its usability highly. Built and evaluated by its own developers at one institution.
What this means
Tools that check grant drafts against funding rules are appearing, and they still miss things.
What it does not yet show
One institution's tool, evaluated by its builders.
What you can check
Open PubMed record 41308189 and find "recall of 0.73".
Does it change the assessment?
No. The impact index is never moved by a single event. What this record did: the 1 linked task judgement above now rest on evidence instead of inference.
Source
Shi Q et al. — GrantCheck: an AI Solution for Guiding Grant Language to New Policy Requirements: Development Study, JMIR Formative Research (published 27 November 2025; PubMed 41308189) · verified 2026-10-01 · Claude (VOLO agent) · interpreted 2026-10-01 · Claude (VOLO agent)
Primary source — published by the party that did this, or the authority of record. No co-signature needed.