Skip to main content

​
AI Intelligence Layer

PILLAR integrates Anthropic’s Claude AI to provide contextual intelligence across scoring, signals, account analysis, and recommended actions. All AI features use the Claude API with organization-scoped data isolation.

​
AI Capabilities

​
1. Pre-Computed Signal Actions

When CRITICAL signals are generated, Claude analyzes the full account context and produces role-specific recommended actions.
ComponentDetail
ModelClaude (via Anthropic API)
TriggerScoring cron detects CRITICAL signal
InputSignal data + account context (ARR, health, contacts, deals, renewal date)
OutputContextual action, reasoning, urgency, confidence score
Storagesignals.ai_action (JSONB)
Cost~$0.01 per signal, capped at 10 per cron run
FallbackRule-based actions from role-actions.ts
Example output: 
{
  "action": "Schedule executive sponsor call with Springfield SD. Their renewal is in 45 days with $350K ARR at risk. Superintendent Dr. Williams has been unresponsive since January.",
  "reasoning": "High-value renewal with declining engagement and no executive touchpoint in 8 months. Executive intervention increases save rate by 3x on deals over $100K.",
  "urgency": "immediate",
  "confidence": 92
}

​
2. Account Intelligence

On-demand AI analysis accessible via Account 360.
EndpointPurpose
/api/ai/account-intelligenceDeep account analysis with deal strategy, risk assessment, expansion opportunities
/api/ai/narrativeAI-generated account narrative for meeting prep and QBRs
/api/ai/action-planStrategic action plan generation for specific accounts
/api/ai/coachingRep coaching suggestions based on deal patterns
/api/ai/board-narrativeBoard-ready narrative generation for executive reporting

​
3. Role-Specific Action Routing

Each signal type has role-appropriate recommended actions defined in role-actions.ts:
RoleAction StyleExample
CRO/CEOStrategic, executive-level”Schedule executive sponsor call. Your involvement increases save rate by 3x.”
VP SalesCoaching, pipeline-focused”Coach rep on deal strategy. Review next steps and decision timeline.”
VP CSRetention, relationship-focused”Assign save play to senior CSM. Review renewal terms.”
RevOpsAnalytical, data-focused”Pull renewal history and usage data. Prepare executive briefing package.”
Rep/CSMTactical, action-oriented”Execute save play tasks. Schedule customer check-in within 48 hours.”

​
4. Weekly Digest Intelligence

The weekly digest email includes AI-curated highlights:
  • Top risk accounts with context
  • Pipeline movement summary
  • Recommended focus areas for the week

​
Configuration

SettingLocationDefault
ANTHROPIC_API_KEYVercel Environment VariablesRequired for AI features
AI enrichment capScoring cron10 signals per org per run
TemperatureAI action generation0.2 (low, for consistency)
Max tokensAI action generation300

​
Data Privacy

  • AI prompts include only the authenticated organization’s data
  • No cross-org data leakage in AI context
  • AI outputs are stored per-signal, scoped by org_id
  • Anthropic does not train on API data (per their data policy)

​
AI-Generated Fields

TableFieldTypeDescription
signalsai_actionJSONBAI-generated recommended action with reasoning and confidence
accountsai_adoption_summaryTEXTStarbridge AI adoption analysis
accountsstartup_friendliness_summaryTEXTStarbridge startup friendliness analysis
accountspropensity_to_spend_summaryTEXTStarbridge spending propensity analysis
accountsprocurement_summaryTEXTStarbridge procurement navigation guide