June 2026 Summaries

At the 2026 Databricks Data + AI Summit, the newly merged Fivetran and dbt Labs showcased their commitment to building an Open Data Infrastructure for the evolving agentic era, emphasizing the need for reliable, governed, and contextual data to support AI-driven processes. Key presentations included insights from George Fraser and Tristan Handy on how AI is reshaping data infrastructures, and Mars Dauer's discussion of the "BI why" problem, highlighting the importance of underlying business logic for accurate data interpretation. The event also featured Jon McManus's case study on how Inova Health rapidly modernized its data strategy using Fivetran, dbt, and Databricks, achieving substantial cost reductions and enhanced data governance. Additionally, exciting new developments were unveiled, such as the dbt Wizard, an AI agent designed for analytics engineering, and the first alpha release of dbt Core v2.0, which promises significant improvements in performance and integration capabilities, further cementing dbt's role in driving innovation within the data community.

Dustin Dorsey, Senior Director of Data Engineering at phData, discusses the often-overlooked issue of semantic debt in organizations, which arises when business meanings are not consistently encoded into data structures, leading to different teams having varying interpretations of the same metrics. This results in inefficiencies, as teams spend significant time reconciling conflicting data interpretations, a problem exacerbated by AI systems that cannot distinguish between these inconsistencies. Unlike technical debt, which involves code shortcuts, semantic debt involves the absence of a shared understanding within data, requiring ongoing efforts to make data-encoded meaning a systemic property rather than relying on institutional knowledge. This is critical as AI scales, and organizations must proactively address semantic debt to maintain trust in AI outputs and ensure reliable decision-making. Dorsey emphasizes the importance of treating semantic work as a strategic investment to build resilient data environments, highlighting tools like dbt and the role of phData in facilitating sustainable semantic consistency.

The Model Context Protocol (MCP) has gained significant traction, with the dbt MCP server emerging as a widely adopted tool within the ecosystem, especially after its transition to the Linux Foundation. The server is utilized in various innovative ways, including conversational analytics through interfaces like Claude or ChatGPT, which enhance accuracy by using governed metric definitions, and generating first-draft documentation from column-level lineage and test coverage to improve model documentation. However, using it for real-time queries against large, unpartitioned tables can unexpectedly increase costs and timeouts, highlighting the need for pre-materialization or query guards. Advanced use cases involve integrating the dbt MCP server with CI/CD for faster pull request reviews and orchestrating model runs based on upstream data quality signals to ensure data integrity. Tools like dbt Wizard CLI are recommended for analytics engineers to efficiently scale data infrastructure by automating and streamlining processes.

AI agents often fail not due to weak models but because of inadequate data infrastructure that was not designed for autonomous decision-making, especially as they transition from simple chatbots to more complex systems executing multi-step business processes. Traditional AI infrastructures, which are suitable for batch model training and human analytics, become bottlenecks in agentic workflows due to issues with data freshness, latency, and semantic consistency. Autonomous agents require an adaptive data infrastructure that integrates real-time data ingestion with centralized semantic layers and standardized interfaces to ensure accuracy and reliability in executing tasks. Key challenges in implementing such infrastructure include the need for sub-minute data freshness, diverse access patterns, and semantic grounding to prevent operational failures caused by context vacuums, lineage blind spots, metric drift, and integration overhead. To address these, a five-layer agent data architecture is proposed, consisting of an ingestion layer for real-time data capture, a processing layer for rapid data retrieval, a semantic layer for machine readability, a validation layer for automated quality checks, and an interface layer for standardized access control. The dbt platform is highlighted as a tool that can help build this infrastructure by providing a governed and semantically consistent data foundation, ensuring reliable agent operation in production environments.

In the evolving landscape of data work, the 2026 State of Analytics Engineering report highlights a growing emphasis on data trust, with 83% of data teams prioritizing it, while simultaneously adopting AI-assisted coding, which is used by 72% of teams. However, only 24% of teams have integrated AI-assisted observability, leading to a trust-speed paradox where rapid AI-driven data production outpaces governance systems, risking data quality and trust. Gartner's prediction that 60% of AI projects will be abandoned due to inadequately prepared data underscores this issue. The solution lies in establishing robust governance frameworks, akin to the dbt workflow, which include preemptive testing, model contracts, and a semantic layer to ensure data consistency and reliability. This approach emphasizes the importance of infrastructure readiness before integrating AI, offering both acceleration and trustworthiness in data work. By adopting such practices, data leaders can bridge the gap between AI adoption speed and governance readiness, ensuring sustainable and trustworthy AI-driven data systems.

Enterprise data infrastructure is crucial for managing large-scale, data-driven applications, especially in the era of Generative AI (GenAI). This infrastructure involves a comprehensive framework for data ingestion, storage, transformation, and governance, ensuring data quality, compliance, and accessibility for authorized personnel. Companies often face challenges in achieving efficient data infrastructure due to fragmented data, heterogeneous systems, and the high cost and complexity of integration. However, modern solutions like dbt offer a vendor-agnostic data control plane that facilitates seamless data integration, transformation, and governance across diverse systems. By optimizing data pipelines through automation and standardization, organizations can enhance data quality and scalability, thereby improving the return on investment and supporting advanced AI applications. With tools like dbt, companies can focus on deriving insights and driving data-driven decisions without the burden of building infrastructure from scratch.

Daniel Poppy's guide emphasizes the pitfalls of a "lift-and-shift" approach when migrating to the Data Build Tool (dbt), where teams replicate their existing data workflows without re-evaluating or optimizing them, leading to continued inefficiencies and technical debt. He outlines common symptoms of poorly executed migrations, such as direct mapping of stored procedures to dbt models, lack of staging models, and inadequate testing and documentation, which result in fragile architectures that fail to take full advantage of dbt's capabilities. The guide advocates for a thoughtful migration strategy that involves triaging existing models to determine which should be eliminated, rewritten, or translated with care, and building new components based on business requirements. It stresses the importance of quality over mere completion, especially as AI systems increasingly depend on robust data infrastructure, and provides a checklist to ensure healthy migration practices that prioritize model clarity, testing, and documentation.

Context engineering is emerging as a critical skill within analytics engineering, particularly for users of the Data Build Tool (dbt), as it involves structuring information to ensure AI systems can accurately interpret and reason about data. This practice is distinct from prompt engineering and retrieval-augmented generation, focusing instead on providing AI with well-organized, machine-readable business knowledge to prevent errors and unreliable outputs. Analytics engineers, who have long been encoding this knowledge in dbt projects, are well-positioned to excel in context engineering by making implicit business logic explicit through mechanisms like model descriptions, MetricFlow definitions, schema contracts, and column-level lineage. These structures enable AI systems to access consistent and defensible data interpretations, thereby enhancing their accuracy and reliability. As organizations recognize the value of this skill set, the role of analytics engineers evolves, expanding their impact from assisting human analysts to empowering AI agents with trustworthy data insights.

In the AI-driven era, the gap between AI-generated data and its readiness for practical use is significant, with only 7% of enterprises fully prepared for AI integration according to a 2026 study by Cloudera and Harvard Business Review Analytic Services. Despite AI's ability to build data structures quickly, without robust foundations such as tests, contracts, and a semantic layer, these structures can be unreliable and prone to errors. The recent merger of Fivetran and dbt Labs aims to address these challenges by providing an open and portable data infrastructure that supports AI agents effectively. Key tools like dbt State, dbt Wizard, and dbt Core v2.0 are designed to enhance trust, reliability, and efficiency in agentic workflows by ensuring schema constraints, validating data accuracy, and optimizing compute resources. The guide emphasizes the importance of establishing a trust foundation with model contracts and data tests, utilizing a semantic layer for reliable query responses, deploying dbt Wizard for efficient agent operations, and leveraging dbt State for cost-effective pipeline management. These measures are crucial for building a scalable, trustworthy, and cost-efficient data infrastructure capable of supporting AI agents effectively.

In the evolving landscape of agentic, self-healing data pipelines, the transformation layer, particularly as implemented in dbt (Data Build Tool), is critical for ensuring correctness in autonomous systems. While AI agents can efficiently manage data pipelines, they lack intrinsic understanding of specific business rules and data semantics, which must be explicitly encoded in models, tests, contracts, and semantic layer definitions provided by dbt. This governed transformation layer ensures that AI agents operate within trusted boundaries and current context, preventing silent failures and ensuring data accuracy. Governance, therefore, precedes and enables AI autonomy by defining what constitutes correct data transformations, making analytics engineers pivotal in creating reliable, autonomous workflows. They shift from coding to defining semantic layers, ensuring their work serves as a foundation for trustworthy data systems.

In 2026, the role of an analytics engineer has evolved significantly from its 2023 focus on model development and SQL coding to emphasize system design, governance ownership, and AI context provision. With AI now automating repetitive coding tasks, analytics engineers are increasingly responsible for making architecture decisions, structuring semantic layers, and ensuring reliable AI-driven data interpretations. Governance has become a key deliverable, with engineers producing tests, contracts, and semantic definitions that underpin trustworthy data systems. Business judgment, semantic precision, and system thinking have grown in importance as AI takes over individual model implementation. These skills enable analytics engineers to define data meanings, govern usage, and build the context AI relies on, thus enhancing their value in a data ecosystem increasingly influenced by AI capabilities. Tools like dbt support this transition by storing semantic context and providing AI agents grounded in project lineage and definitions, further amplifying the analytics engineer's role in ensuring data reliability and scalability.

Dustin Dorsey, Senior Director of Data Engineering at phData, asserts that the inconsistency and unreliability in AI results often stem from data model issues rather than faults in the AI itself. While AI proofs of concept (POCs) tend to be successful due to controlled environments with pre-curated, well-understood datasets, real-world applications falter as they encounter a broader, more ambiguous data landscape. Dorsey argues that the problem lies not in the quality of data but in its design, as centralized data storage often lacks centralized meaning, leading to inconsistent interpretations by AI systems. He emphasizes the importance of dimensional modeling as a solution, which structures data around business processes and clarifies relationships and definitions, thereby reducing ambiguity and improving AI reliability. This approach is supported by tools like dbt, which helps encode business meaning into the data transformation layer, ensuring that AI systems operate on a solid foundation aligned with business processes.

AI adoption is facing challenges not due to a lack of ambition or model capability, but because the data foundation is not adequately prepared. A significant portion of IT leaders express concerns about data readiness and governance, with many organizations struggling to ensure their data is trusted, governed, and contextual. The traditional data stack, initially designed for human analysts, fails to accommodate the autonomous and continuous actions of AI agents, necessitating a shift towards a governed context layer for AI to function reliably. The Data Build Tool (dbt) offers a solution by providing a structured context layer that defines data semantics, allowing AI systems to operate from a single source of truth. This approach reduces redundant queries and improves the reliability of AI outputs. At ACV Auctions, the implementation of rigorous data governance and a semantic layer has transformed self-serve analytics and improved workflow efficiency. Governance, rather than hindering speed, facilitates AI adoption by reducing risks, costs, and ensuring scalability. As the AI landscape continues to evolve, the importance of a robust, interoperable data foundation becomes increasingly critical for organizations looking to implement AI effectively.

dbt Labs has been awarded the 2026 Data Integration Snowflake Product Partner of the Year for the fourth consecutive year, recognizing its commitment to delivering exceptional outcomes for joint customers through its collaboration with Snowflake. At the Snowflake Summit 2026, dbt Labs was also acknowledged for its role in the adoption and transformative results achieved via Snowflake’s coding agent and control plane for builders. The company has become a crucial player in enabling AI and analytics use cases, with over 75% of Snowflake customers utilizing dbt tools such as the dbt Semantic Layer and dbt Fusion engine. This partnership has helped joint customers like WHOOP streamline data processes and build AI/ML models, enhancing organizational trust in data and optimizing ROI. Additionally, dbt Labs' collaboration with Fivetran further strengthens its ability to provide a reliable data infrastructure that supports analytics, operations, and AI across various systems.

Fivetran and dbt Labs have merged to create a unified company focused on building open data infrastructure tailored for the age of agentic AI, which requires data pipelines that can handle the massive, autonomous data consumption of AI agents. The merger aims to address the current bottleneck in data infrastructure, which has shifted from technical limitations to issues of trust, governance, and context, especially as AI agents become the primary consumers of enterprise data. Fivetran ensures data is complete and fresh, while dbt provides governed and trustworthy context, crucial for agents operating at machine speed without human context. Together, they offer a flexible and portable foundation that can adapt to changing AI landscapes and support scalable production AI demands. The merger also underscores a commitment to open source, with the announcement of dbt Core v2.0, featuring a Rust-based rewrite for improved performance, furthering the mission to empower data practitioners in the rapidly evolving AI ecosystem.

Fivetran and dbt Labs have completed their merger, creating a robust data infrastructure designed to facilitate trusted AI agents. This merger, announced in 2025 and finalized in 2026, integrates Fivetran's data management capabilities with dbt Labs' expertise in structured data standards, fostering an open data infrastructure that supports autonomous AI agents. The combined entity, led by George Fraser as CEO and Tristan Handy as President, serves over 100,000 data teams globally, including major brands like OpenAI and Zendesk. The merger aims to redefine data management for AI by ensuring data is reliable, fresh, and governed across systems, thus enabling AI agents to function autonomously and efficiently. The partnership has already led to innovations such as the release of dbt Core v2.0, dbt State, and dbt Wizard, which enhance data pipeline efficiency and support AI-driven workflows. These solutions are built on open standards, allowing organizations to maintain control over their data architecture while preparing for future advancements in AI.

At the Snowflake Summit, the significant announcement was the merger of Fivetran and dbt Labs, aiming to create a unified foundation for data infrastructure that emphasizes open, portable, and reliable data management. This merger combines Fivetran’s expertise in reliable data ingestion with dbt’s prowess in data transformation, solidifying their role as central components in modern data stacks. Additionally, dbt announced the release of dbt Core v2.0, featuring the new Fusion engine that improves performance and scalability by rewriting the original engine in Rust, and introducing dbt State, which optimizes data pipeline efficiency by rebuilding only changed data, thus reducing compute costs. Another highlight was the introduction of dbt Wizard, an AI agent specifically designed for analytics engineering that ensures efficient and accurate code generation while maintaining data lineage and governance standards. These developments reflect a strategic focus on enhancing data management practices, reducing costs, and increasing the reliability and efficiency of data processes, with further insights to be shared in upcoming virtual events.