Workflows First in Pharma AI
Levi Lian, CEO of Raycaster, on why vertical AI is workflows first & chat last
The biggest money and time sink in pharma sits in turning a promising molecule into an approved product, not in finding the molecule in the first place. After discovery, teams spend 8 to 10 years moving protocols, batch records, quality narratives, trial documents, and regulatory filings across CROs, CDMOs, sites, and regulators. That is why the winning AI wedge is document workflow, where delays come from handoffs, rework, and inconsistencies rather than lack of scientific ideas.
-
In practice, one change can ripple through many linked artifacts. A protocol change can force edits to consent forms, statistical analysis plans, case report forms, and trial master file documents. On the manufacturing side, tech transfer packs feed batch records, deviations, stability reports, and Module 3 filing sections.
-
Existing leaders like Veeva and IQVIA mainly act as systems of record. They store regulated content and preserve audit history, but the work of spotting gaps, reconciling inconsistencies, and drafting the next document still falls to people. Raycaster is trying to add an intelligence and action layer on top of that installed base.
-
This is different from Benchling’s original wedge, which digitized upstream lab work and experiment tracking. Raycaster is aimed further downstream, where approval odds are about 10%, costs reach the low billions, and a missed quarter can derail financing because protocol amendments and review delays push timelines out.
The next phase of life sciences AI is likely to move from helping scientists search and summarize into running the document machinery that gets drugs through trials, manufacturing, and regulatory review. If these systems can reduce amendment rates, shorten vendor onboarding, and keep filings internally consistent, they become part of the operating spine of biotech, not just another copiloting tool.