IRB-assist: AI-powered decision support platform that enhances the speed, consistency, and accuracy of institutional review board reviews - institutional review boards (irbs) are vital to protecting…

The National Center for Advancing Translational Sciences (NCATS) at the National Institutes of Health awarded IRB Assist, LLC $314,141 on September 1, 2026, under the NCATS program (CFDA 93.350) to develop and deploy IRB-ASSIST, an AI-powered decision support platform designed to enhance institutional review board efficiency and compliance.

IRB-ASSIST provides secure document upload and real-time analysis of study protocols grounded in federal research regulations (OHRP, FDA, HIPAA). The system's AI engine identifies regulatory, administrative, and ethical issues across protocol documents and flags each issue with explanations and suggested resolutions for reviewer acceptance, rejection, or editing. Additional features include automated classification of study type (exempt, expedited, or full board review) and generation of structured review summaries. The platform operates in the loop with IRB professionals rather than replacing their judgment, addressing documented resource constraints in institutional review boards—57 percent of IRB professionals reported increased workloads without added personnel as of 2022, with nearly half of IRB offices operating at staffing levels of one to five full-time employees while reviewing more than 500 new studies annually.

The award runs from September 1, 2026, through August 31, 2027, with place of performance in Freeport, Florida. This is a Project Grant (R43 mechanism), classified as assistance funding under NCATS's translational science mission to reduce bottlenecks in the research pipeline.

IRB-assist: AI-powered decision support platform that enhances the speed, consistency, and accuracy of institutional review board reviews - institutional review boards (irbs) are vital to protecting human research participants, but as research volume and study complexity continue to increase, irbs are being asked to do more with fewer resources. From 2011 to 2020, protocol submissions surged while staffing stagnated. By 2022, 57% of IRB professionals reported increased workloads without added personnel. Nearly half of IRB offices operate with only 1 - 5 full-time staff, and over a quarter reviewed >500 new studies annually. These pressures lead to frequent revisions, prolonged review cycle, and costly delays. Each preventable revision costs an estimated $200-$600 and adds 3-6 weeks to startup, contributing to over $100 million in annual administrative waste. In time-sensitive conditions, these delays can cost patients their only window for treatment. Despite advances in artificial intelligence (AI), IRB platforms cannot interpret protocol content or flag regulatory and ethical issues, leaving a critical GAP in existing market tools. IRB-assist fills this growing and significant market need by working in the loop with IRB professionals to deliver intelligent, regulation-aware analysis that improves efficiency, consistency, and participant protection for human research studies.grounded in federal research regulations (ohrp, FDA, hipaa), the system enables secure upload of study documents for real-time analysis. The AI engine identifies regulatory, administrative, and ethical issues across documents. The intuitive user interface flags each issue and delivers a brief explanation with a suggested resolution. Reviewers can accept/reject/edit each flag. IRB-assist shortens review cycles by catching errors early and streamlining evaluations. Features include auto-classification of study type (exempt, expedited, or full board) and generation of structured review summaries. By reducing administrative burden and enhancing review accuracy, IRB-assist protects participants and accelerates study launch. While the commercialized platform will incorporate all features above, phase i targets prototype development, AI validation, & pilot usability study. First, we will develop a working prototype to process IRB documentation and report details to reviewers. This includes a backend data architecture and front-end user workflows. Next, we will develop and validate a AI/NLP engine that can process uploaded protocols, extract key regulatory elements, and generate reviewer-facing outputs. We will test the prototype on a diverse set of DE- identified IRB protocols. Outputs (e.g., missing consent elements, risk classification) will be compared to expert- coded annotations derived from IRB reviewer feedback and adjudicated consensus review. Finally, the prototype and AI engine will be integrated at the system level and deployed in a pilot usability study with IRB reviewers.