Building an AI-native data & insights operating system at Webflow

Webflow’s Data & Insights team developed an AI-native operating system to support trustworthy self-service analytics, arguing that reliable agent outputs require more than valid SQL: they depend on governed data, explicit business definitions, reusable workflows, architectural controls, continuous evaluation, and human judgment. The company rebuilt parts of its warehouse around standardized dimensional models, master entities, governed data marts, searchable documentation, and semantic views that translate terms such as ARR and net dollar retention into approved, machine-readable definitions. It also encoded expert procedures into reusable skills and agents for data platform operations, analytics engineering, and data science, with review checkpoints, data freshness checks, permissions, and provenance labels. Because agents can bypass traditional dashboard-based governance, Webflow introduced a control plane that routes queries toward certified sources and semantic definitions, supported by pre-release testing and post-release monitoring that certifies or warns on responses. The approach reportedly increased pull-request throughput by more than three times, shortened some analyses from about two weeks to hours, and improved internal agent-answer accuracy from 46.2% to 96.2% while reducing tool calls by 3.7 times. Webflow plans to expand governed access while improving its ability to detect plausible but incorrect answers and preserve human review for consequential decisions.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.