The evolution of the Analytic Context Engineer (ACE)

LLMs have created the easiest way to access data and are creating an emerging role of Analytic Context Engineer (ACE). ACEs are responsible for making sure every AI answer and every dashboard stays accurate by capturing, governing, and continuously updating the business logic.

ACE extends the “analytics engineer”, popularized by Tristan Handy at dbt Labs . Where analytics engineers transform raw data into clean tables (driven by cloud data warehousing and ELT), context engineers take the next step: they capture business meaning - metrics, definitions, synonyms, permissions - inside a semantic layer so every tool, dashboard, and chatbot answers questions the same way.

Why Natural Language Needs Context

LLMs are intelligent but don’t know your business.

Even a simple prompt like “How many orders do we have?” can split into multiple valid answers:

Finance may exclude fraudulent or returned orders as they don’t contribute to revenue, while logistics includes them to plan shipping capacity. Industry practitioners have long flagged how such ambiguity trips up self‑serve tools and chatbots.

Natural language is an easy interface but inherently imprecise. Implicit in the user's question is an understanding of business context.

A semantic layer anchors these definitions so an LLM can translate everyday language into the right metric, not just any metric.

Business Change Is Accelerating

Generative AI is speeding up product cycles and operational improvements across nearly every domain, forcing analytics to keep pace. When operations shift this quickly, static models rot; context must evolve just as fast.

AI removes the friction from accessing data. Like Jevons paradox, far from automating away data teams, we are seeing even greater demand for ACEs to enable everyone to use reliable, interpretable data.

ACE: Collaborative & Continuous

Analytical context needs to be anti-fragile. As more users in an organization use data, the context improves through implicit signals and explicit engineering.

ACE extends analytics engineering with a product‑management mindset:

• Collect language from business users, encode logic with data teams.
• Promote governed definitions for the entire organization.
• Close the loop by mining usage and chat transcripts to spot new synonyms or edge‑cases.

Continuous feedback loops - common in ML and DevOps - are now being applied to semantic models so they strengthen, not break, under change.

Collaboration between data teams and business users is paramount. To enable accurate, interpretable natural language the technical definition of a metric must be tied to the language. If a marketer refers to return on ad spend as ad performance, LLMs need that context.

What Powers ACEs

• Collaborative modeling UI: ops experts suggest terms and create metrics; analysts refine metadata. 
• Versioned promotion paths: scratch → team → org with impact analysis before each merge.
• One semantic truth everywhere: models serve dashboards, workbooks, and chat, and can be written back to the data warehouse for cross‑tool reuse - no copy‑pasting.