{
  "title": "The semantic layer: computing truth instead of collecting it",
  "slug": "semantic-layer",
  "description": "Leadership waits days for numbers that arrive pre-negotiated. The fix is one canonical definition per metric, computed from data nobody thinks to fake.",
  "archetype": "pattern",
  "solution_id": "semantic-layer",
  "pillar": "visibility",
  "reader": "technical-peer, readable by coo-ops-leader",
  "language": "en",
  "anonymisation_level": "mixed (GoStudent L0; manufacturer + retail group L2)",
  "publish_date": "",
  "canonical": "",
  "internal_links": [
    "outreach-sequencer"
  ],
  "key_takeaways": [
    "A metric without a written definition belongs to whoever reports it.",
    "Prefer machine exhaust over human input: byproduct data that nobody thinks to falsify.",
    "Reconciliation beats blending: when bookings, chat analysis, CRM and revenue disagree, the divergence is the finding.",
    "Agents should query a semantic layer, never raw tables, or they improvise a fresh definition per question."
  ],
  "sources": [
    "01-intake/examples/filled-intake-A.md"
  ],
  "visual_briefs": [
    "semantic-layer-v1"
  ],
  "cta_variant": "quiet-door",
  "opening_move": "scene",
  "skeleton": "pattern",
  "register": "v2-engineering-essay",
  "target_query": "what is a semantic layer for leadership reporting",
  "author": "Antonio Marques, Brutal AI",
  "updated": "",
  "external_citations": [],
  "faq": [
    {
      "q": "Is this the same as the metrics layer in a BI stack?",
      "a": "Same family, three differences in emphasis: definitions are versioned with a named owner per metric; disagreement between sources is surfaced as output rather than cleaned away; and AI agents query only this layer, never raw tables. Existing warehouses and BI tools stay in place."
    },
    {
      "q": "How long does a first version take?",
      "a": "The pilot shape we run is 8 weeks against one data domain: read-only access, canonical definitions for the first metrics, and an evaluation set of 30 to 50 real leadership questions checked against trusted reports before access widens."
    },
    {
      "q": "Does leadership need to learn a new tool?",
      "a": "No. The layer is exposed through AI clients a leadership team already uses, such as Copilot or Claude, through a controlled gateway. The interface is a question in plain language; the layer decides what the model is allowed to compute from."
    }
  ]
}