Revolt BI at dbt Coalesce 2025 in Las Vegas

Openness, AI, and the Future of Data

This year, we joined thousands of data professionals in Las Vegas for dbt Coalesce 2025 — four days filled with inspiration, product launches, and ideas shaping the next chapter of analytics and AI.

Key Announcements

This year’s product lineup showcased dbt Labs’ focus on integration, intelligence, and efficiency.

• dbt Insights (GA) introduced a new analytics layer that provides real-time visibility into project performance.
• dbt MCP Server (GA) opened direct integration with the Model Context Protocol (MCP) ecosystem — connecting dbt with the growing universe of AI and metadata tools.
• In beta, Analyst and Discovery Agents are embedding AI into documentation, data discovery, and observability — helping teams better understand and manage their data assets.
• In preview, dbt Fusion emerged as one of the most significant developments. It unifies modeling, testing, and orchestration into a single interface, now powering dbt Cloud Studio (IDE) — giving developers the full VS Code experience directly in the browser.
• Fusion also powers State-Aware Orchestration (SAO), which intelligently skips unnecessary runs by detecting unchanged data and optimizes dbt data test’s performance. During testing, dbt Labs reported 9% compute savings immediately after enabling Fusion and SAO, with further optimization targets reaching 30–60%.
• Another standout announcement was dbt Compare Mode in Visual Studio Code extension, enabling developers to preview data changes before materializing objects in the database — a simple but revolutionary step toward more transparent and efficient development workflows. Something that’s coming hopefully very soon.

Beyond any single company, dbt Coalesce highlighted a growing community-led movement toward an open semantic interchange.

Organizations such as ThoughtSpot, dbt Labs, and Snowflake are jointly shaping a new, interoperable, vendor-neutral foundation for AI and analytics.

The idea is simple but transformative: when the world speaks the same data language, everyone benefits.

Openness is no longer just a philosophy — it’s becoming the practical standard for collaboration, scalability, and trust.

The Power of Openness

From the opening keynote, dbt Labs made its mission unmistakably clear: the goal is not to lock users into a single platform but to build a framework based on openness, choice, and interoperability.

“Open means open.”

dbt positions itself as the open alternative to today’s walled gardens — the large, closed ecosystems dominating the data and AI landscape. This commitment to flexibility spans across compute engines, LLMs, data catalogs, and BI tools, creating a foundation that’s vendor-neutral and future-proof. 

A major highlight was the official merger between dbt Labs and Fivetran. As Tristan Handy described it:

“The combination will form the most complete, most widely-deployed open data infrastructure platform on the market.”

Rather than limiting dbt’s open-source spirit, this partnership promises to make Fivetran even more open — reinforcing a shared vision of collaboration over consolidation.

Deep Dives and Roundtable Discussions

Beyond the keynotes and product reveals, we also took part in several highly engaging sessions and roundtables that went deeper into the technical and strategic side of modern dbt development.

We explored topics such as:

• agentic workflows, 
• how to build a data or dbt mesh, 
• and techniques for designing optimized incremental models. 
• Another standout track focused on automatically generating semantic layers, showing how AI-assisted metadata modeling is becoming a practical reality.

For engineering leaders, sessions on scaling dbt Cloud with Infrastructure-as-Code (IaC) in large organizations offered concrete approaches to governance, automation, and cost control at enterprise scale.

We also joined several community roundtables — one-hour discussion groups where dbt users exchanged real-world experiences and best practices around Fusion, dbt Catalog, and the evolving semantic layer.

These peer-to-peer sessions captured what makes dbt Coalesce so special: a collaborative community openly sharing how to make data work smarter, not harder.

Data Quality as the Foundation of AI

A central theme of the conference was clear: AI is only as good as your data.
Speakers including Ashley Kramer (with experience from NASA, Tableau, GitLab, and OpenAI) reminded the audience that trust, validation, and context are the cornerstones of meaningful AI.

Without trustworthy data, even the most advanced models are just guessing. This is where the semantic layer plays a crucial role — bridging the gap between raw information and real understanding. For machines, it provides context to reason; for humans, it ensures confidence in results.

Tools like dbt Fusion are turning this principle into reality — building context-aware, explainable systems that bring both human and machine understanding together.

What It Means for Revolt BI

For us at Revolt BI, attending dbt Coalesce 2025 was honestly a blast — an eye-opening experience packed with inspiration, ideas, and genuine conversations that we’ll definitely carry into our projects.

It wasn’t just about the keynotes or sessions (though those were excellent). What stood out was also the networking — spending time with technical people who build, maintain, and actually use dbt every day. Talking through real use cases, sharing challenges, and seeing how others approach the same problems we face was interesting to see.

The organization itself deserves credit too — everything from the easy navigation between sessions to the amazing catering made the whole event to the next level. You could really focus on connecting with others.

We left Vegas feeling energized and full of ideas — both from the official talks and the hallway chats that turned into learning moments.