Data Engineer in 2026: AI Writes the SQL You Review

Elena opens her PR queue. A product manager posted in #data: 'I need MRR by cohort, split by acquisition channel, rolling 90-day.' Elena types the ask into Snowflake Cortex; it drafts a CTE-heavy query, joins four tables, and returns a first pass in 30 seconds. Elena rewrites two joins for correctness, adds a test in dbt, opens the dashboard, spots a data quality issue in the ad-channel attribution, files a ticket with the marketing team. By the time the PM checks back at 2 p.m., she has a working dashboard and a cleaner pipeline. In 2020, this ask was a two-week project.

• SQL generation from natural language (Snowflake Cortex, Databricks Assistant, BigQuery Duet AI).
• dbt model scaffolding — dbt Copilot drafts models with docs and tests.
• Data lineage — Collibra, Alation, Atlan auto-discover and document data flows.
• Pipeline monitoring — Monte Carlo and Bigeye detect data quality anomalies.
• ETL — Fivetran, Airbyte connectors + AI-suggested transformations.
• Semantic layer — Cube, LookML, dbt Semantic Layer with AI-generated metrics.
• Natural-language data access — business users query via chat, not SQL.

• Databricks Assistant — SQL and Python in notebooks with AI.
• Snowflake Cortex — native LLM functions + AI-assisted SQL.
• dbt Copilot — the transformation layer with AI first-drafts.
• Fivetran, Airbyte — managed ELT with AI-suggested mappings.
• Airflow / Dagster / Prefect — orchestration; AI-generated DAGs.
• Monte Carlo, Bigeye — data observability with anomaly detection.
• Atlan, Collibra, Alation — governance and catalog with AI search.

Designing the data model. Deciding whether a field should be an event or a dimension. Enforcing contracts across producer and consumer teams. Explaining to finance why the number they expected is not the number in the dashboard — and proving which one is right. Leading a migration from one warehouse to another. Setting retention and privacy policy. Negotiating ad-hoc exceptions without breaking the platform. AI can write SQL; it cannot design your company's data vocabulary.

• SQL mastery — window functions, CTEs, query optimization, execution plans.
• Data modeling — star/snowflake schemas, dimensional modeling, data vault.
• Python — ETL, orchestration, data science support.
• One warehouse deeply — Snowflake or Databricks.
• dbt or equivalent transformation framework.
• Governance and observability — the fastest-growing senior-engineer skill.

If you want to be a data engineer: In high school, take AP Statistics and AP CS. In college, major in CS, data science, or analytics; SQL is learned best by doing it on public datasets (Kaggle, BigQuery public data). Build an end-to-end data pipeline as a portfolio project — Fivetran → Snowflake → dbt → Looker. Data engineering is the infrastructure layer of analytics and ML; it is less fashionable than MLE and often better-paid mid-career. The role is stable precisely because AI produces more data that needs engineering, not less.