What is a context layer?

A context layer is the shared record of what a company's numbers mean: which systems hold each metric, how each one is calculated, and which version is approved. It sits between finance data and AI tools, so an AI assistant can answer from agreed definitions instead of guessing from raw tables.

The term is new, and vendors use it loosely. This page explains what a context layer contains, how it differs from a semantic layer, and why finance teams benefit from one before they put AI in front of their numbers.

Why AI needs a context layer in finance

"Deposits" can mean one thing in the core banking system, another in the general ledger and a third in the FP&A model. Each version is correct for its purpose. An AI assistant pointed at all three may pick one without saying which, and the answer sounds equally confident either way. A context layer tells it which definition applies to which question.

Context layer vs semantic layer vs metric layer vs data catalog

These terms overlap, and vendors use them differently. This is how they typically compare.

• Context layer. Stores definitions, cross-system mappings, owners, approvals and lineage. Maintained by finance with data team support, used by people and AI tools, spans multiple systems, and is built for AI assistants.
• Semantic layer. Stores business terms mapped to warehouse tables. Maintained by the data team, used by BI tools and analysts, usually covers one warehouse, and is increasingly built for AI assistants.
• Metric layer. Stores metric calculations. Maintained by the data team, used by BI tools and analysts, usually covers one warehouse, and is partly built for AI assistants.
• Data catalog. Stores an inventory of data assets and metadata. Maintained by the data governance team, used by data teams, spans multiple systems as an inventory, and is not primarily built for AI assistants.

What goes into a context layer

1. Definitions. What each metric and business concept means, in plain language.
2. Mappings. Where each concept lives in each system, and how the versions relate.
3. Owners. Who is accountable for each definition.
4. Approval history. Who approved each definition and when it changed.
5. Lineage. How each figure traces back to its source.

How Preql builds one

Preql's Discovery Agent scans your data sources to find how each system represents the same business concept. The Translation Agent turns those representations into shared business logic. Your team reviews and approves the resulting mappings, and every change is version-controlled.

Who should own it

Finance owns what the numbers mean. Data teams own the pipelines, connections and infrastructure. A context layer works best when finance can read, approve and change definitions without filing a ticket.

FAQ

Is a context layer the same as a semantic layer?

They overlap. A semantic layer translates tables into business terms for BI tools. A context layer adds ownership, approval and cross-system mapping, and is built to be read by AI.

Do I need Snowflake to use Preql?

No. Preql is designed to work with your existing systems and infrastructure.

Do I need a data team?

Finance can define and approve metrics in Preql without writing code. A data or IT team still typically helps with system access and connections.