The problem
A new study setup requires a dense protocol PDF to be converted into structured information: overview fields, visit schedules, procedures, and a matrix linking each procedure to its relevant visit. The work is repetitive but requires care. Visit windows expressed in weeks and months were also being converted into days manually in a purpose-built Excel sheet for each study.
What we built
We delivered a scoped proof of concept that processes a real protocol from start to finish. Narrative pages use text extraction. Schedule-of-activities tables use visual extraction so visit columns, procedure rows, and footnotes remain available. The workflow produces structured study-setup data, an Excel workbook, and a Word outline for review by the CRIO team.
The workflow does not write to the live CRIO environment. It prepares materials for a qualified person to check before any system update is considered.
What the site team said
Coordinators reviewed the output jointly. The visit-window conversion alone replaces the Excel sheet one coordinator previously created to convert weeks and months into days. With the corrected procedure matrix, the team estimated a saving of approximately 45 minutes per study setup.
Next phase
The proof of concept intentionally ends at review. The proposed next phase is a workflow that creates and updates the study shell in CRIO, retaining a human review step before any data is committed.