Available Context & Tools

@_platform-references/org-variables.md @_platform-references/capabilities.md

Deal Intelligence Summary

Goal

Generate a comprehensive, narrative intelligence summary for a deal that synthesizes health data, relationship status, risk signals, activity patterns, and recommended actions into a coherent story. This is not a dashboard of metrics -- it is a strategic briefing that answers: "What is the current state of this deal, why is it in that state, and what should I do about it?"

Why Intelligence Summaries Matter

Sales reps are drowning in data but starving for insight. The average CRM deal record has 40+ data points, 20+ related records, and dozens of activities. Converting this into actionable intelligence requires synthesis, context, and interpretation -- exactly what humans are slow at and AI excels at.

Research shows:

Reps spend 21% of their time on data entry and review (Salesforce State of Sales, 2023) -- much of that time is trying to figure out "what does all this mean?"
Context switching costs 40% of productive time (American Psychological Association) -- jumping between CRM tabs, email, calendar, and notes to piece together deal status
Teams with AI-driven deal intelligence close 19% faster (McKinsey Sales Analytics Study) because reps spend less time diagnosing and more time acting

The Deal Intelligence Summary delivers what a rep would take 15 minutes to manually compile in under 3 seconds.

Required Capabilities

CRM: To fetch deal data, health scores, relationship scores, activities, and contacts

Inputs

deal_id: The deal identifier (required)
include_history: Whether to include health score trend history (optional, default: true)

Data Gathering (via execute_action)

Gather comprehensive intelligence data:

Deal record: execute_action("get_deal", { id: deal_id, include_health: true }) -- stage, value, close date, contacts, company
Deal health score: Query deal_health_scores table:
- overall_health_score, health_status, risk_level, risk_factors
- sentiment_trend, avg_sentiment_last_3_meetings
- days_in_current_stage, days_since_last_meeting, days_since_last_activity
- meeting_count_last_30_days, activity_count_last_30_days
- predicted_close_probability, predicted_days_to_close
- Component scores: stage_velocity_score, sentiment_score, engagement_score, activity_score
Health history (if include_history = true): Query deal_health_history table:
- Fetch last 5 snapshots ordered by snapshot_at descending
- Calculate trend: improving, stable, or declining
Relationship health: Query relationship_health_scores for all contacts on deal:
- Focus on primary contact: overall_health_score, is_ghost_risk, ghost_probability_percent
- days_since_last_contact, response_rate_percent, sentiment_score
Recent activities: execute_action("get_deal_activities", { deal_id, limit: 20 }) -- last 30 days
Recent meetings: Filter activities for type = 'meeting', extract outcomes and sentiment
Open tasks: execute_action("list_tasks", { deal_id, status: "open" }) -- planned actions

If any data is missing, note it in the narrative: "Health score not available -- recommend running health recalculation."

Narrative Structure

The intelligence summary should follow this structure:

1. Executive Snapshot (2-3 sentences)

Current stage, value, and expected close date
Overall health status in plain language
Primary risk or opportunity flag

Example: "The Acme Corp Enterprise License deal ($185K, Negotiation stage, closes Feb 28) is showing warning signs with a health score of 43/100. The primary contact has gone dark for 12 days and sentiment from the last demo was lukewarm (55%). This deal needs immediate re-engagement to prevent slippage."

2. Health Analysis (3-4 sentences)

Current health score and trend direction (if history available)
Breakdown of what is driving the score (stage velocity, sentiment, engagement)
Comparison to healthy deals in the same stage

Example: "Health score has declined from 58 to 43 over the past 14 days, driven by a 20-point drop in engagement score and a sentiment decline. Stage velocity is concerning -- the deal has been in Negotiation for 28 days vs. the 14-day average. Sentiment from recent meetings (55% avg) indicates buyer hesitation that has not been addressed."

3. Relationship Health (2-3 sentences)

Primary contact engagement status
Ghost risk assessment
Multi-threading status (are other stakeholders engaged?)

Example: "Primary contact Sarah Chen (Director of Ops) has not responded to 3 emails over 12 days, flagging a 68% ghost probability. No other stakeholders are actively engaged, making this a single-threaded deal at high risk. If Sarah continues to be unresponsive, the deal will need multi-threading or escalation within 72 hours."

4. Risk Signals (bullet list, top 3-5 signals with severity)

Ranked by severity (critical → high → medium)
Each signal includes: what it is, why it matters, and when it was detected

Example:

Critical: No response from champion in 12 days (detected Feb 15) -- single-threaded deals that go dark for 14+ days close at <8%
High: Stage duration 2x average (28 days vs 14-day avg) -- extended negotiation signals hidden objection or missing stakeholder
High: Sentiment declining trend (from 72% to 55%) -- buyer enthusiasm is waning, likely due to unaddressed concern
Medium: No next meeting scheduled -- momentum requires a committed next step with a date

5. Recent Highlights (bullet list, 3-5 key events)

Significant activities from the last 30 days
Meeting outcomes, key emails, proposal sent, etc.
Include dates for context

Example:

Feb 3: Demo completed with Sarah Chen and technical team -- positive initial feedback but security questions raised
Feb 5: Follow-up email sent with security documentation -- opened but no reply
Feb 10: Pricing proposal sent ($185K, 3-year contract) -- opened twice, no response
Feb 12: Voicemail left for Sarah -- no callback
Feb 15: LinkedIn message sent -- read but no reply

6. Recommended Next Actions (3 actions with reasoning)

Ranked by impact and urgency
Each action includes: what to do, why it matters, estimated time
Connect actions to the specific risk signals and deal context

Example:

Multi-thread into technical buyer (~20 min): Sarah's silence is critical. Reach out to the VP of Engineering (mentioned in demo) to check project status. If the deal is still active, request an alignment meeting. Single-threaded deals die when the thread breaks.
Send breakup email to Sarah (~10 min): After 12 days of multi-channel silence, send the honest check-in: "I want to be respectful of your time. It seems like timing may not be right for this initiative. If things change, I'm here." This re-engages 15-20% of stalled deals.
Address security concerns explicitly (~30 min): The demo surfaced security questions that were answered via email but not discussed. Schedule a technical deep-dive with your security team and their technical evaluator to remove this blocker.

7. Bottom-Line Assessment (1-2 sentences)

Honest evaluation: is the deal salvageable, healthy, or dead?
Recommended focus level: high priority, monitor, or disqualify

Example: "This deal is salvageable but fragile. Without re-engagement within 48 hours, it will likely slip past the Feb 28 close date or die entirely. Recommend making this a high-priority rescue this week."

Narrative Tone and Style

Direct and specific: Use names, dates, numbers. "Sarah Chen has not responded in 12 days" not "the contact is unresponsive"
Contextual: Explain why things matter. "Stage duration 2x average signals hidden objection" not just "deal has been in stage 28 days"
Honest: If the deal looks dead, say so. False optimism wastes time.
Actionable: Every insight should connect to a recommendation. "Sentiment is declining" → "Address the unresolved concern before proposing next steps"
Avoid jargon: Write for a sales rep, not a data scientist. "Health score of 43" is less useful than "Health score of 43, indicating the deal is at risk of stalling"

Output Contract

Return a SkillResult with:

data.narrative_summary: The full structured narrative (string, formatted with markdown headers and bullets)
data.health_snapshot: object
- overall_health_score: number (0-100)
- health_status: "healthy" | "warning" | "critical" | "stalled"
- risk_level: "low" | "medium" | "high" | "critical"
- trend_direction: "improving" | "stable" | "declining" | null (if no history)
- trend_change: number (e.g., -15 for 15-point drop) | null
data.relationship_snapshot: object
- primary_contact_name: string
- primary_contact_health_score: number (0-100)
- is_ghost_risk: boolean
- ghost_probability_percent: number | null
- days_since_last_contact: number
- response_rate_percent: number | null
- multithreading_status: "single" | "limited" | "good" (based on number of engaged contacts)
data.risk_signals: array of objects (top 5, ranked by severity)
- signal_type: string (e.g., "no_activity", "sentiment_decline", "stage_stall")
- severity: "critical" | "high" | "medium" | "low"
- description: string (plain language)
- detected_date: string (ISO date)
- impact: string (why this signal matters)
data.recent_highlights: array of objects (last 5 significant events)
- date: string (ISO date)
- event_type: "meeting" | "email" | "call" | "proposal" | "task" | "note"
- description: string
- outcome: string | null (for meetings)
data.recommended_actions: array of objects (top 3)
- action: string (specific, actionable)
- reasoning: string (why this action, tied to deal state)
- estimated_time_minutes: number
- priority: "critical" | "high" | "medium"
data.bottom_line_assessment: object
- salvageable: boolean
- recommended_focus: "high_priority" | "monitor" | "disqualify"
- assessment: string (1-2 sentence honest evaluation)

Quality Checklist

Before returning the intelligence summary, verify:

Examples

Good Narrative Summary (Opening)

The GlobalTech Analytics Suite deal ($92K, Evaluation stage, closes Mar 15) is healthy with a score of 74/100, up from 68 two weeks ago. Momentum is building -- we completed a successful demo last week with positive sentiment (78%) and the champion is actively driving internal alignment. However, the deal is single-threaded through a mid-level PM, which is a structural risk worth addressing.

Why this works: Specific deal name, value, stage, date. Health score with context (trend). Sentiment data. Risk is flagged with context.

Bad Narrative Summary (Opening)

This deal is doing well. Health score is good. The demo went fine. We should keep working on it.

Why this fails: Vague. No specifics. No actionable insight. No risk assessment.

Good Risk Signal

{
  "signal_type": "sentiment_decline",
  "severity": "high",
  "description": "Sentiment dropped from 78% (Feb 10 demo) to 55% (Feb 15 follow-up)",
  "detected_date": "2026-02-15",
  "impact": "Declining sentiment signals an unresolved concern or objection. If not addressed, buyer enthusiasm will continue to erode and the deal will stall in evaluation."
}

Why this works: Specific trend with dates. Clear severity. Impact explains why the rep should care.

Bad Risk Signal

{
  "signal_type": "risk",
  "severity": "medium",
  "description": "There is some risk",
  "detected_date": null,
  "impact": "Could be a problem"
}

Why this fails: No specifics. Vague severity. No context. Not actionable.

Error Handling

Health score not available

Note it in the narrative: "Health score not available for this deal. Recommend running health recalculation to enable intelligence-driven insights. Based on available activity data, [manual assessment]."

No recent activity

Frame it as a finding: "No activity recorded in the last 30 days. This deal is either inactive or activity logging is incomplete. Recommend confirming deal status with the champion and updating the CRM before investing more time."

Single activity data point

Work with what you have but flag the limitation: "Only 1 meeting recorded in the last 30 days. Limited activity data makes trend analysis unreliable. Sentiment from that meeting (Feb 10 demo) was positive (78%), but without follow-up interactions, we cannot assess whether momentum is maintained."

Deal has no close date

Include as a risk signal: "No expected close date set. This ambiguity itself is a risk -- deals without a target date are 3x more likely to slip indefinitely. First action: confirm a realistic close date with the buyer."

Missing contact health data

Note it: "Relationship health data unavailable for primary contact. Based on activity logs, last contact was [X] days ago. Recommend confirming engagement status and response patterns."

Deal is clearly dead

Be direct: "This deal shows all the hallmarks of a lost opportunity: 45+ days with no response, close date 3 months overdue, and health score of 12/100. Recommend marking as closed-lost and reallocating time to active opportunities. If there is reason to believe the deal is still viable, document that reasoning in the CRM."

Tone and Presentation

Write as if briefing a sales manager who has 60 seconds to understand the deal
Lead with the most important information (health status, primary risk)
Use plain language, not CRM jargon
Be honest about bad news (managers respect candor over optimism)
Connect data to actions (every insight should lead to a recommendation)
Include numbers and dates for credibility and specificity
Acknowledge uncertainty when data is incomplete (do not fill gaps with assumptions)