“How to Actually Move Up the Stack” — AE Roundup
Why this is in the vault
Practical career playbook from the dbt creator for analytics engineers navigating the agentic transition. Filed as K-sender thought-leadership because it blends personal narrative, tactical advice, and organizational insight. Directly relevant to RDCO’s consulting positioning and the “boring AI” thesis.
Core argument
Analytics engineers are being called to move up the stack into agentic workflows. The window where this is “just slightly early” is now, and it is narrower than previous transitions. The key difference from the dbt transition: this one requires organizational change, not just individual adoption.
Key themes
Timing window is narrow. The sweet spot is when it feels like surfing a wave, not getting overtaken. The agentic transition is moving faster than any prior paradigm shift, compressing the window for early adopters.
Hands-on experience is non-negotiable. Working with real data agents on production data is so different from demos or toy datasets that without it you are flying blind. If your org blocks access, that is serious data about your environment.
Pattern: immerse, act, share. Handy describes his own path — tracking MCP at conferences, vibe-coding the dbt MCP server on a weekend, launching dbt agent skills after seeing a Claude skills talk. The flywheel is: absorb interesting patterns, build locally, share publicly.
Change management is the real barrier. Commenter Salim’s insight: the dbt transition could happen in isolation (version control your transforms, nobody else needs to change). The agentic transition requires the whole org to rethink knowledge management — business context must be captured in agent-consumable formats. This is culture work, not context engineering.
Recommended reading list. Ethan Mollick, AI Daily Brief, AI Engineer conference, METR, Redwood Research, Hyperdimensional (Dean Ball), Don’t Worry About the Vase, Andrew Curren on X.
RDCO mapping
- Consulting positioning — Handy’s framing of “change management as the biggest barrier” is exactly where RDCO advisory value lives. Technical implementation is necessary but insufficient; the organizational metabolism problem is the real engagement.
- Skills/harness validation — his personal narrative of building the dbt MCP server and agent skills after conference exposure mirrors the thin-harness-fat-skills thesis. The harness (MCP server, skills, orchestration) is where the moat forms.
- Sanity Check content angle — the “change management > context engineering” insight is a strong article hook. Most AI content focuses on the technical layer; the organizational layer is underserved.
- Information diet quality — his recommended reading list is a useful cross-reference for our own newsletter intake pipeline.
Related
- 2026-04-11-garry-tan-thin-harness-fat-skills — the architecture pattern Handy is describing from a practitioner’s perspective
- 2026-04-12-alphasignal-claude-code-leak-harness-engineering — the Claude Code leak as evidence for the harness engineering thesis
- 2026-04-12-lindstrom-board-ai-governance — the governance/accountability layer that complements the change management challenge