Workflow-First Vertical AI for Pharma Development
Levi Lian, CEO of Raycaster, on why vertical AI is workflows first & chat last
Raycaster is positioning around the part of pharma where AI can become embedded in the company’s actual operating machinery, not just help people search or chat. In drug development, the core work product is the chain of specs, batch records, protocols, quality narratives, and submission modules that move a drug from lab to clinic to approval. Raycaster is built to draft, review, and reconcile those artifacts inside regulated workflows, which makes it closer to Harvey’s legal work product and Hebbia’s document reasoning than to biotech discovery tools.
-
The key difference from discovery AI is where the bottleneck sits. Raycaster is aimed at the 8 to 10 year development phase after discovery, where teams coordinate across CROs, CDMOs, sites, labs, QA, and regulators, and small document inconsistencies can trigger rework, amendments, or review delays.
-
The Harvey comparison fits because both products attack expensive document workflows in regulated fields. Harvey sells into legal research and drafting for firms and in house teams. Raycaster applies the same logic to pharma development documents, where each file is tied to SOPs, approvals, and auditability, and where the document itself often determines whether work can proceed.
-
The Hebbia comparison fits on the control layer. Hebbia became useful by letting non technical teams run multi step analysis over large document sets. Raycaster extends that pattern into action, where the system is not just finding answers in documents, but proposing edits, generating the next artifact, routing reviews, and learning from pass fail history inside customer workflows.
The category is moving from AI over public information to AI inside proprietary workflow systems. As horizontal models absorb generic research and chat, the durable companies in regulated markets will be the ones that own internal context, evaluation loops, and approval paths. In life sciences, that makes document native workflow software a much larger wedge than discovery chatbots.