This article is currently being updated. Please come back later for updates.
Important: AI capabilities are not enabled by default, and availability varies by elluminate release, contracted products, URL settings, study settings, and assigned privileges. Contact an eClinical Solutions representative for details about enabling AI capabilities.
Types of AI in elluminate
elluminate uses two types of AI:
Generative AI: Large language models (LLMs) interpret natural language and generate content, such as summaries, SQL queries, risk statements, and suggested Review Objectives. Generative AI capabilities ground each response in the study data, metadata, and documents available to the user.
Machine learning and statistical models: Purpose-built models developed by eClinical Solutions analyze study data to detect anomalies, inconsistencies, and classification issues. These models flag records for review and provide confidence indicators or similarity scores rather than generated text.
AI capabilities in elluminate are purpose-built for clinical workflows, assist expert work rather than replace human judgment, and are governed across development, use, and audit. For details, refer to the AI Governance, Privacy, and Data Security in elluminate article.
Extracts study details from the Study Protocol and suggests Review Objectives from study documents and the Global Library. Users accept, edit, or reject each suggestion before it is applied.
Generates domains, variables, and Source Field, Method Name, and Mapping Arguments recommendations for SDTM specifications. Users accept, reject, or update each recommendation.
Summarizes the current visualization or listing, highlights insights relevant to the reviewer role and Review Objectives, recommends next steps, and answers follow-up questions in natural language. Users confirm all issues before the issues are created.
Converts natural language questions about SDTM data into SQL queries and returns the results with a plain-language explanation of the query logic. elluminate Assist cannot modify or delete data.
Flags anomalous records and subjects, such as atypical adverse event durations or lab value shifts, with confidence indicators. Reviewers assess each flagged record and provide feedback on model predictions.
Generates a structured 'if / due to / then' Risk Statement from a Risk Description. AI-generated content displays a disclaimer until a user edits the content.