Board Narratives
The Board Narrative engine assembles a comprehensive data bundle from across the PILLAR platform and uses Claude to generate structured, board-ready narratives. Each narrative contains multiple sections with prose content, supporting data points, trend indicators, and severity classifications.Board Narratives are powered by Anthropic’s Claude AI. The engine enforces pre-computed KPIs to ensure generated numbers match PILLAR’s verified data exactly.
How It Works
Narrative Types
Each narrative type generates 5 core sections plus an optional “Strategic Recommendations” section when
includeRecommendations is enabled.
Tone Options
Data Assembly
The engine assembles data from the following sources in a single parallel fetch:Pre-Computed KPIs
To prevent the AI from hallucinating numbers, the engine pre-computes all aggregate KPIs and injects them into the prompt with explicit instructions to use them verbatim:- Total Portfolio ARR
- At-risk account count and ARR (with % of portfolio)
- Total renewable ARR and at-risk renewal ARR (with % of renewable)
- Critical and warning signal counts
- Top territory metrics (name, accounts, ARR, health)
- Lowest health territory metrics
- Play execution stats (active, completed, completion rate)
- Average health and risk scores
Narrative Sections
Each section in the generated narrative contains:
Data points include trend indicators:
Data Model
PILLAR stores generated narratives with full data snapshots for auditability. Each narrative includes the type, period, generated sections with content and data points, the complete data bundle used for generation (audit trail), tone, model version, and generation metadata.Detailed data model schemas are available in the PILLAR Implementation Guide provided to active customers.