Decision infrastructure.
Aurelia is the governed layer between commercial data and the people, applications and AI agents making the next marketing decision. Connect context once; reuse the same meaning, evidence and tools everywhere.
The four layers
| Layer | What Aurelia governs | What it gives AI |
|---|---|---|
| Meaning | Metric definitions, formulas, owners, sources, freshness and versions. | Company-specific context instead of generic marketing definitions. |
| Evidence | Reported attribution, MMM, response curves, experiments and cohort value. | The right measurement basis with scope and uncertainty intact. |
| Tools | Typed semantic functions, guardrails and read-only MCP access. | Stable capabilities that survive a change of model or client. |
| Memory | Recommendation, approval, implementation, outcome and frozen lineage. | What the business tried, why it tried it and what happened next. |
Governed metrics are portable meaning
A number becomes infrastructure only when its meaning travels with it. Aurelia stores each headline metric as a versioned contract: the approved definition and formula, structured parameters, accountable owner, source lineage and approval history.
The workspace, Aurelia Auto and MCP tools all resolve the same active contract. If the company changes how MER is defined, downstream applications and agents receive the approved meaning rather than maintaining separate prompt instructions.
A measurement mix, not one magic number
Different questions require different evidence. Aurelia preserves the distinction rather than collapsing every source into a universal ROAS figure.
- Store-settled reporting describes recognised commercial performance.
- Ad-platform attribution describes what each platform claims credit for.
- MMM estimates period-level incremental contribution and, when verified curves exist, marginal return at a proposed spend level.
- Controlled experiments test a precommitted causal question against a holdout and retain uncertainty.
- Cohort economics connects acquisition to rebuy, payback and lifetime contribution.
AI agents are consumers, not competitors
Models are good at reasoning and interaction. Aurelia supplies what a general model does not own: the company’s governed definitions, measurement state, permissions and decision history. The built-in Aurelia Auto agent is one client of that infrastructure; compatible clients can use the same tools through MCP.
This separation keeps the infrastructure durable. Teams can change from one reasoning model to another without rebuilding metric logic, measurement caveats or action guardrails in every system prompt.
The next decision compounds
A recommendation is not the end of the workflow. Aurelia freezes the evidence used, records the reviewer and approved move, tracks implementation and compares the observed outcome with the expectation. That complete record becomes context for the next decision.