November 2026 Summaries
5 posts from dbt
Filter
Month:
Year:
Post Summaries
Back to Blog
No summary generated yet.
Nov 16, 2026
1,701 words in the original blog post.
AI projects often take two to four years to deliver returns, compared with seven to twelve months for typical technology investments, while agentic systems can drive substantial and difficult-to-justify token spending. The post argues that Model Context Protocol implementations can inflate costs by loading numerous tool definitions for every call, repeatedly transmitting raw data, and re-retrieving the same context across sessions. It proposes a warehouse-first approach in which data from sources such as call transcripts is processed, classified, embedded, and stored as a governed reusable context layer, allowing agents to query a single warehouse instead of many MCP servers and receive only relevant excerpts rather than full payloads. Drawing on internal results and Anthropic examples, it says this method can reduce token use and improve response quality while enabling organizations to reuse processed context across teams. To make the financial case to leadership, it recommends measuring baseline token, tool, connection, and API costs, prioritizing expensive sources already overlapping with warehouse data, using dbt to prepare unstructured information, and comparing costs before and after implementation.
Nov 11, 2026
1,475 words in the original blog post.
Fivetran’s founders argue that AI agents, defined as intelligence combined with tools and data, will reshape knowledge work more profoundly than previous technological shifts by enabling systems to answer questions and perform actions with extensive organizational context. They contend that agents are particularly effective in domains with accessible context and verifiable outcomes, such as software engineering and data analysis, but require access to both structured business data and unstructured information to address complex problems accurately. The letter presents existing data infrastructure as a foundation for agent adoption while emphasizing the need for emerging capabilities such as evaluations, persistent memory, and shared business context. It anticipates a transition from AI augmenting individual workers to autonomous agents managing end-to-end processes, supported by “agent engineers” who redesign workflows, connect systems and data, evaluate performance, and guide organizational change. The founders also advocate open data infrastructure, standards, and portability to prevent vendor dependence, while maintaining that human qualities including empathy, goal-setting, and innovation will remain essential as organizations adopt increasingly autonomous AI systems.
Nov 11, 2026
1,378 words in the original blog post.
AI maturity depends less on adopting advanced models or agents than on whether an organization’s data, governance, ownership, and business context can reliably support increasingly autonomous workflows. While 41% of surveyed organizations use agentic AI in production, only 15% report having a fully prepared data foundation, highlighting a gap between deployment and readiness. The progression from trusted data to trusted AI requires standardized, tested, documented data; scalable governance and accountability; machine-readable semantic definitions, metadata, and lineage; and interoperable infrastructure that preserves context and controls across platforms. As these capabilities mature, organizations can allow AI agents to act across more complex and consequential processes while retaining traceability, oversight, and clear responsibility. The central argument is that autonomy should be an outcome of a strong, adaptable foundation rather than an initial objective, enabling organizations to adopt evolving AI technologies without repeatedly rebuilding trust and control.
Nov 10, 2026
1,176 words in the original blog post.
AI readiness depends less on simply adopting AI than on having a scalable data and organizational foundation to support production systems and increasingly autonomous agents. The Fivetran and dbt Labs maturity model evaluates readiness across six interconnected dimensions: data quality, governance, ownership, cost and operational efficiency, organizational context, and data architecture and interoperability. Reported research indicates that although many organizations are investing heavily in agentic AI, relatively few believe their existing data foundations are fully prepared, with data quality and lineage among the most common obstacles. Organizations are encouraged to assess each dimension separately because maturity is often uneven across teams and systems, then bring together stakeholders from data, technology, governance, and AI to identify differing perspectives and prioritize improvements based on planned use cases. Readiness should be treated as an ongoing progression, requiring regular reassessment as AI applications expand in scale, complexity, and autonomy.
Nov 05, 2026
1,098 words in the original blog post.