With great context comes great autonomy

At dbt Summit 2026, dbt Labs’ Ben Moser and Fivetran’s Kevin Kim argued that trustworthy AI agents depend primarily on well-governed business context rather than agent sophistication alone, since ambiguous questions about metrics, definitions, permissions, and authoritative data sources can otherwise produce unreliable results. Moser described context as a spectrum ranging from structured semantic metadata, documentation, tests, and lineage already available in dbt projects to harder-to-model institutional knowledge contained in support tickets, Slack conversations, and sales calls. He cautioned against attempting to provide all information at once, emphasizing instead the need to progressively model and govern relevant context to prevent inaccuracies from scaling rapidly. Kim presented Fivetran Context Layer as a system that connects knowledge, operational, and analytics sources, builds semantic search indexes, infers relationships among concepts, and writes the resulting information into open warehouse tables called Agents Schema. These tables include definitions of schema contents and company-specific terminology, can be queried by AI clients through MCP, and support trace features that identify low-confidence answers or missing context for ongoing curation.

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.