Workflows First in Vertical AI

This claim means public data is no longer a moat, because the core extraction work is being absorbed into general purpose AI research tools. Once OpenAI and Anthropic can browse the web, read PDFs, and return cited tables on demand, a startup built mainly on organizing public documents gets pulled toward commodity pricing. The durable layer shifts to internal workflows, approvals, and organization specific context that generic agents cannot see or safely act on.

• Raycaster reached this view by testing public document research in life sciences legal and R&D, then shifting toward internal document chains like tech transfer packs, batch records, specs, methods, and Module 3 sections. Those artifacts are tied to reviews, redlines, and downstream filing work, which is much harder for a horizontal model to copy.
  1 sacra
• Finance shows both sides of the pattern. Transcript search over earnings calls is valuable, but products like AlphaSense defend themselves by bundling premium content, structured data, and analyst workflows in one place, not by raw access to public filings alone. The product becomes sticky when it helps build comps, models, and monitoring inside daily work.
  5 sacra 6 sacra
• The next winning vertical AI companies look more like workflow systems than chat boxes. In law, finance, and medicine, the strongest players pair model output with firm or clinician specific guardrails, licensed or proprietary data, and integrations into the place where work already happens. That is what turns AI from a feature into operating software.
  1 sacra 7 sacra 8 sacra

The market is moving toward a split where horizontals own broad reasoning and retrieval, while vertical products win by owning approvals, telemetry, and high stakes workflows. As general models keep improving, value will concentrate in software that captures edits, routes decisions, and turns company specific work patterns into a living system of record and action.