Permissions: Users must have the Study Configure privilege to enable Data Advisor.
Users must have access to Data Central and study View privileges.

Overview of Data Advisor

Data Advisor is a multi-agent, AI-assisted analysis capability embedded within Data Central that helps reviewers understand data, identify insights, and take action without leaving their workflow. 

Data Advisor automatically analyzes study data, highlights trends and anomalies, and provides contextual guidance based on the current visualization or listing, reviewer role, review objectives, study metadata, and protocol information. It supports faster time-to-insight, reduced manual analysis, and more consistent review workflows.

Key Capabilities

• Instant understanding: Provides contextual summaries, insights, and recommendations.
• Guided exploration: Ask questions in natural language using text or voice input to explore data further.
• Context-aware intelligence: Uses study metadata, review objectives, study protocol information, and broader study data.
• Interactive analysis: Generates temporary tables and charts for exploratory review.
• In-workflow actions: Create domain-, record-, or field-level issues directly from insights.

Identifies

• Data quality issues: Missing values, inconsistencies, and duplicates.
• Outliers and unusual measurements: Values or patterns outside expected ranges.
• Imbalances across sites, subjects, or cohorts: Variations that may require further review.
• Trends: Patterns that may indicate safety or operational concerns.
• Deviations from review objectives: Findings that may conflict with expected review criteria or study expectations.

Note: Data Advisor is not enabled by default and must be enabled in Study Settings by an eClinical Solutions representative.

Important: Data Advisor is a decision-support tool. All outputs require user review and validation. Data Advisor does not replace user judgment and does not take actions automatically.

AI Processing and Grounding

Data Advisor uses a multi-agent system (MAS) AI architecture that combines large language models (LLMs), retrieval-augmented generation (RAG), and tool-based data access to generate contextual outputs for clinical data review.

• Retrieves relevant study data and contextual information at runtime. Contextual information includes review objectives, protocol information, and relevant elluminate Help Center content.
• Separates data retrieval from interpretation.
• Uses structured retrieval and tool-based computation to support analysis.

Data retrieval and computations remain deterministic. Natural language responses may vary slightly in wording, but core findings remain consistent when based on the same data, inputs, and user role.

Validation and Limitations

All insights can be validated directly within the platform through the underlying study data and contextual sources used during analysis.

Data Advisor underwent validation through Subject Matter Expert review, structured evaluation criteria, and testing across multiple studies and datasets.

Data Advisor may not identify every relevant pattern or insight, and may surface findings or interpretations that require additional review and validation.

Safeguards

• Restriction to relevant clinical use cases.

• Prevention of unsupported or unsafe requests.

• Validation of outputs for structure and completeness.

• Enforces controlled tool access and platform protections.

Data Scope and Security

• The model operates entirely within secure eClinical AWS infrastructure and the elluminate platform environment. Study data, prompts, and responses remain within this controlled environment.
• Enforces Role-Based Access Control (RBAC), study-level isolation, data blinding requirements, and controlled API access to ensure only necessary, context-specific data is accessible to the user.
• Does not send data to external systems or use study data for model training.

Outputs and Insights

The Data Advisor panel delivers contextual outputs aligned to the current visualization or listing, reviewer role, and assigned review objectives.

• Summary: Explanation of the data in the current view.
• Review Objectives: Contextual objectives aligned to role.
• Insights: Patterns and anomalies.
• Recommendations: Suggested next steps and actions.

Interactive analysis supports natural language queries, text or voice input, multi-language interaction, iterative follow-up questions, and cross-domain exploration across supported listings and visualizations.

Users can also highlight and reuse response content, queue multiple prompts, and generate temporary tables and charts for exploratory analysis within the chat window. 

Access Data Advisor

Use Data Advisor to review insights, investigate findings, create issues, access related review objective configuration, and view the Help Center documentation directly within the workflow.

1. Open a supported visualization or listing in Data Central.

   • Data Central Visualizations (excluding system table charts such as the Review Plan Dashboard and those in the elluminate folder)
   • Data Listings (excluding Subjects, Issues, Queries, and Records Review Summary panels)

2. Apply filters to the visualization or listing, as needed.

3. Click the Data Advisor icon to open the side panel.
   

4. Review the generated outputs in the following sections:

   • Summary

   • Review Objectives

   • Insights

   • Recommendations

5. If review objectives are available, click a link to open the associated review objective in the left navigation.

6. Ask follow-up questions in natural language to further investigate findings or explore related data.

Note: Data Advisor does not create queries automatically. It can suggest when an issue may be required and generate draft issue text. Users must review and confirm all issue creation.

Explore Data

Use Explore to continue investigating study data across supported listings and visualizations.

• Ask follow-up questions in natural language using text or voice input.
• Investigate findings through iterative analysis and cross-domain exploration.
• Review updated summaries, insights, recommendations, and contextual outputs based on follow-up questions.
• Highlight text or values in the side panel and click Add to prompt to include the selected content in a follow-up question or analysis request.
• Reopen previously viewed supported listings or visualizations during the same session to restore the associated Data Advisor context and outputs.
• Generate temporary tables and charts, as needed, to support exploratory analysis.

Data Advisor maintains contextual information for the active visualization or listing during the current session. Contextual history is cleared when the screen is refreshed or the session times out or ends.

Sample Explore Questions

Note: Sample questions are examples only. Available responses depend on the current listing or visualization, study data, and user permissions.

General Data Exploration

• Summarize the key findings in this visualization.
• What trends or anomalies should be reviewed?
• Are there any subjects that appear outside expected ranges?
• Which sites show the highest number of issues or outliers?
• Are there patterns across cohorts that require further review?

Cross-Domain Analysis

• Show subjects with elevated liver function tests and related adverse events.
• Are there subjects with abnormal vital signs and corresponding medication changes?
• Compare adverse events across treatment arms.
• Identify subjects with missed visits and related protocol deviations.

Generate Tables or Charts

• Create a pie chart of adverse events by severity.
• Create a listing of subjects with Grade 3 or higher adverse events.
• Show a table of subjects with duplicate records in this domain.
• Generate a chart comparing enrollment by cohort.

Create Issues

• Create an issue for subjects with missing lab results.
• Create a domain-level issue for this trend.
• Create a field-level issue for the missing value identified in this record.
• Create a record-level issue for this subject.

Follow-Up Investigation

• Why is this site an outlier compared to other sites?
• Show the underlying records contributing to this trend.
• Are these findings aligned with the review objectives?
• What additional data should be reviewed for these subjects?

Create an Issue

Create domain-, record-, or field-level issues directly from Data Advisor insights and recommendations.

1. Review the generated insights and recommendations.

2. Click an issue link in the Recommendations section or the Create Domain Issue icon.
   

The Create Issue window opens.

3. Review the prepopulated issue details (if available), including suggested text and relevant context.
4. Update the fields in the Create Issue window.
5. Click Save.

Use Tools

Click the Tools icon.

Select one of the following options:

• Upload Study Protocol: Open the study protocol folder in eDrive to upload or review protocol documentation used for contextual analysis.
• Configure Review Plan: Open Data Central Configuration on the Review Plan tab to review or update review objectives and study review settings.
• elluminate Help Center: Open Help Center documentation in a new browser tab.

Regenerate Data Advisor Output

Regenerate Data Advisor outputs based on the current visualization, listing, filters, and conversation context.

1. Click the Regenerate icon when:
   • Filters or context have changed.
   • Additional study data becomes available.
   • Responses need to be refreshed.
2. Review the updated summaries, insights, recommendations, and generated outputs.

Provide Model Feedback

Rate Data Advisor responses using:

• 👍 Thumbs up
• 👎 Thumbs down

Click the Close icon to close the Data Advisor panel.