dbt Labs

We’ve got 9 in-person dbt Meetups coming up in May 🙌 If you’re looking to learn alongside fellow members of the dbt Community, and have some fun while doing so, join us at one of the sessions below: 🇺🇸 Seattle | May 7th | Organized by Bishal Gupta, Reuben McCreanor, and Taylor Dunlap 🇧🇪 Belgium | May 12th | Organized by Sam Debruyn 🇺🇸 New York | May 13th | Organized by David A. Gelman, Ph.D., Ken Rickabaugh, and Guru Mahendran 🇹🇼 Taipei | May 20th | Organized by Karen Hsieh, Laurence Chen, Allen Wang, and LI KUAN LIAO 🇬🇧 London | May 27th | Organized by Nathan Purvis, Edward Hayter, and James Charnley 🇰🇷 Seoul | May 27th | Organized by Thomas Kim and Joshua Kim 🇨🇿 Prague | May 28th | Organized by Stephan Durry and Michal Mike Vilímek 🇧🇷 Sao Paulo | May 28th | Organized by Angela Alves Ferreira, Bruno Lima, Letícia Suzin Miorelli, and Thales Donizeti 🇨🇦 Toronto | May 29th | Organized by Daria Sukhareva, Moiz Ali, and Eddy Z. We're looking for dbt Community members to co/organize meetups in: Copenhagen, Dublin, Munich, San Francisco, Berlin, Paris, Northern Germany, Sydney, Tel Aviv, Brisbane, Los Angeles, Vancouver, Barcelona, Bogotá, Prague, Wellington, Atlanta, Switzerland, Philadelphia, Boston, Austin, Stockholm, Melbourne, Baltic, Budapest, Portland, Rhein-Ruhr, Washington DC, and Seattle. Fill out this form to get started https://lnkd.in/gciu6TQi If you don’t see a meetup near you, find your local dbt chapter at https://lnkd.in/ekknesFN and join the group to get notified about future events.

Can robots do our jobs? For data people, the signs are mixed. Benn Stancil wrote a thorough piece on exactly this, through the lens of ADE-bench, an open-source benchmark built with dbt Labs and Macro to test AI agents in real dbt environments. His take: writing queries was never really the job. The real work is glue work, mapping this team's data to that team's goal, dealing with ambiguity, and knowing when to say no. Here's what the research found: • The prompt matters as much as the problem. Vague directions trip up agents on simple tasks more than hard ones. • The toughest tasks require agents to query the data itself, not just read the code. Bad values in a table are harder to catch than a syntax error. • Context is the real bottleneck. Better models help less than better documentation, richer metadata, and more business context. • Multiple-choice questions are a legitimate eval strategy for analytical reasoning when there's no single "correct" answer. ADE-bench is open source, and the dbt Community Slack has a dedicated channel (tools-ade-bench) if you want to run your own tests or build on top of it. Full post linked in the comments.

The 2026 State of Analytics Engineering report is out, and one finding is hard to ignore: AI is accelerating output faster than foundations can keep up. Join us live on LinkedIn April 29 with Katie Bauer (Hex), Jay Sobel (Ramp), and Jason Ganz (dbt Labs) to go beyond the numbers 📊 They'll get into: • How AI has moved from experimentation into everyday workflows • Where trust and governance are under pressure • The tension between moving fast and maintaining control • Where teams are investing and where they're pulling back Bring your questions. We'll see you there.

Aviva unlocked 1.3 petabytes of legacy data in under a year. One of the UK's largest insurers needed an enterprise-scale AI-ready data platform. They built it on Snowflake and dbt, and the results speak for themselves: • A shift to a product-led data operating model • Hundreds of governed data products shipped in five months • 8x increase in release frequency • A foundation that now underpins Aviva's entire AI strategy On May 13 at the Gartner Data & Analytics Summit in London, Niall Scott, Head of Data Engineering at Aviva, takes the stage to share the practical lessons behind the journey, from modernising legacy data at speed to building AI-ready foundations that actually deliver. Catch Niall's session, then find the dbt Labs team at booth 317 ☕ #GartnerDA

Most analytics engineers using agents today are getting a real speedup on model writing. That's also the wrong thing to focus on. The more important question is where agents *break,* and in AE work, the break point looks different than it does in software engineering. Agents handle the fiddly middle of the week well. Sessionization, SCDs, deduplication, rolling aggregations — one-shotted, clean, usable. But tacit knowledge breaks them. Clean SQL, passing tests, wrong joins — because two systems used the same field name to mean two different things, and that knowledge lived in engineers' heads, not in the code or the docs. The agent had no way to know. And it *looked fine*. The gap between "the code runs" and "the data is right" is where data incidents live. Agents are good at the first. Variable at the second. The real opportunity for AEs right now isn't getting faster at the middle, it's going deeper at the edges. Source system knowledge on the left side of the DAG. Business context on the right. That's where the leverage is. Where are you finding those edges in your own work? We write about topics like this every week in the Analytics Engineering Roundup. This week's piece is by Jason Ganz — link in my profile if you want to subscribe. #analyticsengineering #dbt #dataengineering #AIAgents

Meet the engineer who's been living inside the hardest problem in the dbt Fusion rewrite. Anders Swanson caught up with Chenyu Li to talk about what static analysis actually breaks when you try to map familiar dbt concepts onto a fundamentally different architecture. The answer: schema. Specifically, what "schema" even means when Fusion is analyzing your project before anything runs. They get into: • Why deferral gets dramatically more complex under static analysis • How the analyze phase adds a new layer to the task graph that didn't exist in dbt Core • Why you can't just point refs to the production schema when propagating from sources locally • What DAG optimization could look like when Fusion collapses shared operations across models into a single executor node

We're going live tomorrow to break down the 2026 State of Analytics Engineering report with Katie Bauer, Jay Sobel, and Jason Ganz. AI is accelerating output faster than foundations can keep up. We're getting into what that means for your team. Don't miss it https://lnkd.in/g_ZE34f6

The modern data stack has a fragmentation problem. Even with best-in-class warehouse solutions, mismatched ingestion and transformation tooling quietly erodes reliability, governance, and the overall value of your data. Open data infrastructure (ODI) is how leading teams are solving it. Kyle Dempsey (dbt Labs) and David A. Gelman, Ph.D. (Brooklyn Data (Velir's data studio) walk through what ODI looks like in practice, why organizations are making the shift, and how to build toward it. You'll leave with a clearer picture of: • What a modular, interoperable data architecture looks like across clouds and platforms • How to reduce vendor lock-in without sacrificing reliability or governance • Where dbt fits into a flexible, future-ready stack built for AI workloads Global-friendly sessions on May 5th & 6th. Save your seat: https://lnkd.in/eWaVnYmj

The analytics engineering frontier is moving fast, and it's not always obvious where the edges are. This week's issue of The Analytics Engineering Roundup features Jason Ganz's firsthand take on where coding agents succeed and fall short in complex analytics engineering work today.

The next generation of analytics engineers is out there, and we're going to them. Last month, Connor McArthur, Erin Vaughan, and Anna Lee headed to Temple University for dbt Labs' first in-person University Partners event. They covered: 🔸What dbt is and how it fits into the AI ecosystem 🔸How to get involved with the dbt community 🔸The path from university to the data workplace For students who couldn't make it to campus, "Analytics Engineering for Students" brings the same foundation online. It's a free, beginner-friendly virtual training that takes you from raw data to insights using dbt, covering the full analytics development lifecycle from the ground up https://lnkd.in/g6vRR3fJ