Moving Up the Stack: Analytics Engineering in the Age of Agents
Essay by Jason Ganz. Argues the analytics engineering profession is hitting a phase change comparable to the pre-dbt to post-dbt transition, and practitioners need to move up the stack again.
Curated Topics
- Phase change parallel: Pre-dbt, Jason hand-wrote hundreds of SQL queries for a board meeting; dbt automated that entire skillset. AI agents are triggering the same kind of displacement-then-elevation cycle
- Adoption signals: Hex reports 50%+ of new cells are agent-created. dbt MCP server growing 40% month-over-month. Ramp deploying agentic analysts to exponentially increase data value
- dbt’s “moving up the stack” value: Core company value since 2016 — replace yourself with processes/technology/documentation to free capacity for higher-value work
- Open questions for the profession: What does an analytics engineer do when AI writes SQL? How do we maintain institutional knowledge about AI-generated data models? Why does knowledge need curation in an agent world?
- Timeline acceleration: The dbt adoption wave took 5 years. The agent transition will not take that long
- Semantic layer as open question: Repeatedly flagged as unresolved — does anyone have opinions about the role of a semantic layer in agentic data work?
RDCO-Relevant Links
- Analytics engineering role evolution — directly relevant to how we position consulting services and hire
- Semantic layer as unsolved problem — opportunity for RDCO thought leadership and consulting differentiation
- dbt MCP server as central data infrastructure — validates our agent-ready pipeline architecture positioning
- Ramp’s agentic analysts — case study for client conversations about ROI of agent-enabled data teams