What Okta learned governing context for AI agents at scale

At dbt Summit 2026, Okta analytics engineering senior manager Pooja Crahen described how the company shifted its AI-agent efforts from building agents to governing the context they can access and trust. Okta’s early Snowflake Cortex experiments used semantic views created in Snowflake’s UI, but these were disconnected from dbt-governed business logic, difficult to audit, and structured as undifferentiated instruction sets. The team responded by creating a three-layer framework in dbt that combines model and column metadata, markdown context documents, and dbt-managed semantic views, all reviewed through standard pull requests. These resources synchronize with Agents Schema, a warehouse-native metadata standard that lets products such as Claude and Snowflake Cortex use the same governed source of truth through different access methods. Okta also built tooling to convert existing Snowflake semantic views into dbt specifications and is considering whether AI agents should become a first-class dbt node type. The approach is intended to make corrections propagate across all connected agents immediately, while future plans include automated regression testing, broader domain coverage, and measurable agent-accuracy feedback loops.

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.