The Future of Real World Evidence Generation with Agentic AI

Watch the Novel Cohorts podcast with special guests Jeff Morgan and Nishant Aggarwal from Deloitte

Unlike our typical podcast, this episode features a live demo of Deloitte’s Gen AI solution RWE Agent. The walkthrough showed how agentic AI is transforming the RWE lifecycle, helping researchers move beyond rigid, historical self-service systems toward more intuitive, transparent, and efficient workflows.  The session offered a glimpse into what’s possible from natural language cohort building and data model–agnostic workflows to AI-generated suggestions that draw on knowledge bases like ClinicalTrials.gov to interpret trial protocols and recommend inclusion or exclusion criteria based on defined cohorts and outcomes. 

Some key learnings were:

  • Natural Language Cohort Building: Conversational interfaces allow users to define cohorts in natural language, lowering the learning curve compared to historical self-service tools
  • Intelligent Suggestions: AI can propose relevant inclusion/exclusion criteria, prior datasets, and study references based on similar historical analyses drawn from knowledge bases like ClinicalTrials.gov
  • Data Format Agnostic: Agentic AI tools can work directly with source data, removing the need for rigid data model transformations required by earlier self-service systems
  • Speed and Efficiency Gains: Data onboarding, query generation, and cohort creation processes that historically took weeks or months can now be completed in hours or days
  • Visual and Statistical Summaries: AI tools can automatically generate descriptive statistics and visualizations of cohort characteristics, supporting validation and deeper insight exploration
  • Transparency of Outputs: Features like viewing generated SQL code or inclusion/exclusion criteria enable researchers to verify and understand AI-driven results
  • Human-in-the-Loop Design: AI-generated outputs can be validated through human review at key checkpoints to ensure accuracy and trust, especially for inclusion/exclusion criteria and cohort definitions
  • Flexible Deployment: Can be deployed within client environments, giving control and ability to extend the solution, unlike rigid third-party SaaS tools
  • Guardrails and Compliance: Role-based permissions and built-in guardrails help prevent misuse or unauthorized analyses, ensuring safety and regulatory compliance

View Transcript