dataaaaa!

The AI market is shifting from a "possibility market" to an "accountability market" where economics, infrastructure, and organizational reality must reconcile. While enterprises adopt AI, many struggle to turn experimentation into measurable returns.

Data engineers often mistake database schemas, DDL files, or dbt folders for actual data models. Joe Reis argues that "data modeling is more than just schema design." A schema represents physical implementation, whereas true data modeling requires a deeper conceptual understanding of the business.

Steve Brown discusses why corporate AI initiatives fail due to poor communication and lack of cultural integration. He outlines a three-step path to AI transformation, emphasizing "21st-century labor amplification" over 20th-century cost-cutting.

The upcoming U.S. AI regulations will significantly impact enterprise adoption by 2026. Data engineers must prepare for stricter compliance, data governance, and auditing requirements when building AI pipelines.

As data artifact creation becomes faster and cheaper, the scarce resource is company-wide consensus. Data teams must build a singular reality to prevent globally inconsistent data. To scale self-serve safely, Ramp routes AI agents through an abstraction layer rather than the raw data lake.

Saks Global orchestrates its massive retail data pipelines using Apache Airflow to ingest 8 million SKUs into Snowflake. To streamline operations, they migrated from self-managed Kubernetes to Astro.

Nao Labs released an open-source analytics agent with a harness UI designed for data professionals. While traditional agents focus on execution, Nao optimizes the context layer using static assets like schemas. If an agent is the "brain," the harness is everything that lets it actually operate.

Ameet Talwalkar of CMU and Datadog discusses the evolution of time series foundation models like Toto V1 and V2. In observability, scale and data curation outperform public benchmarks. We have foundation models for text.

As Ramp scaled, its data team shifted focus to managing self-serve analytics and routing AI agents through the data lake. Ian Macomber argues that in this post-AI landscape, "the scarce resource is consensus" rather than data access.

AI agents fail in data engineering because they lack a harness to ground them. Without it, 78% of AI-generated queries contain silent wrong joins. The query runs, returns a result, and looks correct. It just isn't. Deterministic rules, not probabilistic LLMs, must handle validation.

Tiankai Feng, Director of Data and AI Strategy at Thoughtworks, discusses the human side of data and AI. He emphasizes the importance of humanizing AI strategy, focusing on how organizations can bridge the gap between technical data engineering and human-centric business outcomes.

Astronomer Staff Data Scientist Marion Azoulai details how they built their internal data and AI department. The team structured a modern data stack using Snowflake, Sigma, and Airflow as the backbone. This architecture supports machine learning industrialization and AI analytics.

Decathlon Chief Data and AI Officer Didier Mamma structures the company's global strategy around key data engineering initiatives. Their major technical projects focus on building a robust lakehouse architecture, improving data quality and governance, and deploying agentic workflows.

As AI agents generate the majority of database activity, traditional architectures are merging transactional and analytical systems into unified platforms like Lakebase and LTAP. This shift targets speed, scale, and a single copy of governed data.

DataBird launches three bootcamps, including AI Agent Builder, to train data professionals in generative AI and agentic workflows. Le marche evolue tres vite et les besoins explosent as they target data analysts, product managers, and ops to build practical AI agents.

Wes McKinney discusses his shift from AI skepticism to adopting coding agents. He explains why Go is the best language for AI agents and how he builds local data sovereignty tools. He shares his complete reversal from AI skepticism to being fully locked in on coding agents.

Data engineering fundamentals remain essential in the era of artificial intelligence. This episode argues against the notion that teams lack the time for core practices, asserting that AI actually makes data engineering more critical, not less for successful implementations.

Bob Seiner discusses his "Non-Invasive Data Governance" approach and its application to AI. The episode explores why he pushes back on "data enablement" and addresses the critical state of data security in the age of autonomous AI agents.

The 40-year adoption curve of the electric grid is a better analogy for the AI boom than the dot-com bubble. For data engineers, this highlights that most AI failures come down to the organization, not the technology, requiring a focus on long-term infrastructure and systemic integration.

Fivetran CEO George Fraser explains how the modern data stack has transitioned into open data infrastructure. He argues that data gravity is the most overrated concept in data management.

In the age of AI coding agents, traditional software code is no longer a defensible moat. Data engineers must focus on proprietary data pipelines and integration complexity to build real value. What's actually defensible in the age of coding agents, and what isn't is now the critical question.

Following Snowflake's acquisition of SelectStar, founder Shinji Kim discusses the evolution of data catalogs. The category is shifting from traditional documentation tools into a living AI context layer.

Joe Reis challenges the idea that data professionals must constantly maximize AI use. He advocates for minimizing AI during deep cognitive work to maintain critical thinking. Relying too much on automated tokens can erode the core analytical skills that actually prevent you from getting left behind.

AI coding agents fail at data engineering because they jump to conclusions and ship pipelines nobody understands. This leads to costly vibe-coded pipeline problems. As Josh Wills notes, agents create pipelines that lack the human context needed to navigate organizational data politics.

In this episode, Joe Reis and Zach Wilson meet in Stockholm to discuss "what it actually takes to be a data engineer in 2026." They explore the evolving technical skills, architectural patterns, and tools required to succeed in the future data engineering landscape.