Semantic Layer vs. Decision-Making Skills
A Semantic Layer Reports a Number. Fig Decides.
Semantic layers like dbt, Cube, LookML, and Snowflake Semantic Views make a number correct and consistent. That's the foundation — not the finish line. Fig is the decision-making skills layer on top: it knows what the number should be, catches it when it's off, remembers what happened, and acts.
Semantic Layer dbt · Cube · LookML · Snowflake Semantic Views | Fig Decision-making skills layer | |
|---|---|---|
| One consistent metric definition | ||
| Enforced on AI agents, not just BI | ||
| Causal map (what drives what) | ||
| Knows what a number should be (expectations) | ||
| Anomaly detection + automatic root cause | ||
| Remembers outcomes and learns | ||
| Acts within governed guardrails | ||
| Works with any AI agent (MCP) |
Fig knows what a number should be.
A semantic layer returns the number and stops. It has no forecast or expectation, so it can never tell you a number is wrong — only what it is.
Fig notices the moment a number is off — and who it affects.
A semantic layer is passive. Snowflake Semantic Views and dbt metrics wait to be queried; nothing watches for surprise or traces the cascade.
Fig remembers how past decisions turned out.
Semantic layers hold definitions only. There's no memory of outcomes, so nothing gets more accurate over time.
Fig acts within governed guardrails.
A semantic layer can't act — no policy, authority, budget, or preview-and-undo. It defines; it cannot decide.
Questions people ask
What is the difference between a semantic layer and a decision-making skills layer?
A semantic layer (dbt, Cube, LookML, Snowflake Semantic Views) defines what a metric means so every tool computes it the same way — it makes a number correct and consistent. A decision-making skills layer sits on top and decides: it knows what the number should be, detects when it's off and who it affects, remembers how past decisions turned out, and acts within governed guardrails. A semantic layer reports; a decision-making skills layer decides.
Can a semantic layer detect anomalies or forecast?
No. A semantic layer is definitional — it normalizes and serves consistent metric definitions. It has no concept of an expected value, no anomaly detection, and no forecasting. Detecting that a number is abnormal, and tracing why, requires a layer that holds expectations — a decision-making skills layer.
Do Snowflake Semantic Views replace Fig?
No — they're complementary. Snowflake Semantic Views define consistent business metrics inside Snowflake and feed tools like Cortex Analyst. Fig reads those definitions and adds what a semantic view doesn't: expectations, anomaly detection with root cause, memory of outcomes, governed action, and access for any AI agent via MCP. You keep your semantic views; Fig turns them into decisions.
Do I need a semantic layer to use Fig?
No. Fig builds a governed metric and causal model from your warehouse directly, and will use an existing semantic layer's definitions as input when you have one. Either way, Fig adds the decision-making skills layer on top.
Is Fig a semantic layer?
Fig includes governed, consistent metric definitions (the semantic-layer job) but is more than that. Fig is a decision-making skills layer: it adds expectations, surprise detection, memory, and governed action that a semantic layer does not provide.
See what deciding — not just reporting — looks like
Connect your data and put governed decision-making in front of your whole team and every AI agent.