Available Context & Tools

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

Deal Auto-Tagger

Goal

Analyze a deal's attributes (industry, size, stage, products discussed, meeting themes) and intelligently suggest or apply tags that improve pipeline segmentation, reporting, and prioritization. Tags should be specific, actionable, and consistent across the pipeline -- not random or one-off labels.

Why Structured Deal Tagging Matters

Untagged or poorly tagged deals are invisible in pipeline analytics. The data:

73% of sales teams cannot accurately segment their pipeline by industry, deal size, or product interest because tagging is inconsistent or missing (CSO Insights, 2023).
Win rate analysis by segment requires consistent tagging. Teams with structured tagging (industry vertical, deal size tier, product line) can identify their highest-performing segments and focus there. Teams without tagging operate blind.
Forecasting accuracy improves by 34% when deals are tagged by urgency and buying stage (Clari Revenue Analytics).
Manager visibility depends on tags. Sales leaders filter pipeline by tags to identify risks ("Enterprise deals in legal review") or opportunities ("Healthcare deals above $200K in proposal stage"). Without tags, these filters are useless.
AI-suggested tags are 5x more consistent than manual tagging (HubSpot tagging study). Reps are inconsistent: one rep tags "SaaS" and another tags "Software." AI enforces a standard taxonomy.

Required Capabilities

CRM: To fetch deal data, company data, contacts, and activities
Meetings: To analyze meeting transcripts for themes and topics

Inputs

deal_id: The deal identifier (required)
auto_apply: If true, apply tags immediately via CRM. If false, return suggestions only. Default: false (suggestion mode).

Data Gathering (via execute_action)

Fetch deal record: execute_action("get_deal", { id: deal_id }) -- stage, value, close date, existing tags
Fetch company data: execute_action("get_company", { id: company_id }) -- industry, employee count, revenue, location
Fetch deal contacts: execute_action("list_contacts", { deal_id }) -- titles, seniority levels
Fetch meeting transcripts: execute_action("list_meetings", { deal_id, include_transcripts: true }) -- analyze themes, topics, pain points, product mentions
Fetch activities: execute_action("get_deal_activities", { deal_id, limit: 10 }) -- look for patterns (frequency, engagement level)

Tag Taxonomy (Standardized)

Use this standardized tag taxonomy across all deals. Do NOT create new tags outside this structure.

Industry Vertical Tags

Use specific verticals, not generic categories:

Healthcare (hospitals, clinics, health tech)
Financial Services (banks, insurance, fintech)
SaaS (B2B software companies)
E-commerce (online retail, marketplaces)
Manufacturing (industrial, supply chain)
Education (schools, universities, ed-tech)
Real Estate (property management, construction)
Technology (IT services, hardware, infrastructure)
Non-Profit (NGOs, associations, foundations)
Government (public sector, agencies)
Retail (brick-and-mortar stores, franchises)
Media & Entertainment (publishing, broadcasting, gaming)
Professional Services (consulting, legal, accounting)
Other (if none of the above fit)

Deal Size Tier Tags

Based on deal value. Tiers should align with your company's pricing structure:

Enterprise ($250K+)
Mid-Market ($50K-$249K)
SMB ($10K-$49K)
Starter (under $10K)

Adjust these thresholds based on your company's ACV segmentation.

Urgency Level Tags

Based on timeline and compelling events:

Urgent (close date within 30 days OR buyer has a hard deadline)
High Priority (close date 30-60 days OR strong buying signals)
Standard (close date 60-90 days OR normal sales cycle)
Long-Term (close date 90+ days OR early exploration)

Product Interest Tags

Based on which products/solutions the buyer discussed:

Use your company's product line names (e.g., Platform, Analytics, Integrations, API, Enterprise Suite)
If multiple products discussed, apply multiple tags
If product is unclear, use TBD - Product Interest

Buying Stage Tags (distinct from CRM stage)

These are behavioral indicators, not CRM pipeline stages:

Active Evaluation (comparing solutions, running POCs, technical review)
Budget Approved (buyer confirmed budget is allocated)
Legal Review (contract/security/procurement in progress)
Executive Alignment (C-level or VP+ engaged)
Champion Identified (clear internal advocate)
Multi-Threaded (3+ stakeholders engaged)
Single-Threaded (only one contact engaged -- risk signal)
Competitive (actively comparing with named competitors)

Risk Tags (optional, for at-risk deals)

Stalled (no activity in 14+ days)
Close Date Pushed (close date moved 2+ times)
Budget Risk (budget not confirmed or reduced)
Champion Dark (primary contact unresponsive)

Strategic Tags (optional, for special deals)

Strategic Partnership (joint go-to-market, co-sell, integration partnership)
Expansion (upsell or cross-sell to existing customer)
Renewal (contract renewal)
Reference Account (high-profile logo, referenceable customer)
Pilot/POC (proof of concept or pilot deal, not full contract)

Tag Suggestion Logic

For each tag category, analyze the deal data and determine which tags apply:

Industry Vertical

Check company industry field first (if available and accurate)
If company industry is missing or vague ("Technology"), analyze:
- Company name (e.g., "City Hospital" → Healthcare)
- Contact titles (e.g., "VP of Patient Care" → Healthcare)
- Meeting transcript keywords (e.g., "patient records", "EMR" → Healthcare)
Assign ONE primary vertical (do not tag multiple verticals unless genuinely multi-industry)

Deal Size Tier

Based on deal value. If value is missing:
- Estimate from company size (employee count, revenue)
- Estimate from solution scope discussed in meetings
- Default: TBD - Size if insufficient data

Urgency Level

Check close date proximity (within 30 days = Urgent)
Check for compelling events in transcript:
- "Need this by end of quarter"
- "Board meeting in 3 weeks"
- "Regulatory deadline in April"
- "Fiscal year ends June 30"
Default: Standard if no urgency signals

Product Interest

Analyze meeting transcripts for product mentions:
- "We're interested in your analytics module" → Analytics
- "Need the full enterprise suite with all integrations" → Enterprise Suite, Integrations
If no clear product discussed, use TBD - Product Interest

Buying Stage

Active Evaluation: Transcript shows "comparing options", "POC", "evaluating vendors"
Budget Approved: Buyer said "budget is allocated", "we have $X approved"
Legal Review: Activities show contract review, security questionnaire, DPA negotiation
Executive Alignment: VP+ or C-level attendees in meetings
Champion Identified: One person is clearly driving the initiative internally
Multi-Threaded: 3+ unique contacts engaged in the last 30 days
Single-Threaded: Only 1 contact engaged (risk flag)
Competitive: Buyer mentioned competitors by name

Risk Tags

Apply ONLY if risk signals are clearly present:

Stalled: No activity in 14+ days
Close Date Pushed: Close date has been moved 2+ times
Budget Risk: Buyer raised budget concerns or value was reduced 20%+
Champion Dark: Primary contact has not responded to last 3+ outreach attempts

Strategic Tags

Apply ONLY if applicable:

Strategic Partnership: Buyer discussed co-marketing, integration partnership, joint solution
Expansion: Buyer is an existing customer, this is an upsell/cross-sell
Renewal: Contract renewal for existing customer
Reference Account: Buyer is a well-known brand in the industry (Fortune 500, major logo)
Pilot/POC: Buyer wants a paid or unpaid pilot before full commitment

Confidence Scoring

For each suggested tag, assign a confidence score (0.0-1.0):

0.9-1.0: Explicit evidence (e.g., company industry field says "Healthcare" → tag: Healthcare)
0.7-0.89: Strong inference (e.g., transcript mentions "patient data" 5+ times → tag: Healthcare)
0.5-0.69: Moderate inference (e.g., contact title is "Director of Clinical Operations" → tag: Healthcare)
0.3-0.49: Weak inference (e.g., company name includes "Health" but unclear context)
0.0-0.29: Insufficient data (do not suggest this tag)

Only suggest tags with confidence 0.5+ (moderate to high confidence).

Tag Application (if auto_apply is true)

If auto_apply is true, apply tags using:

execute_action("update_deal", {
  id: deal_id,
  tags: [array of suggested tags]
})

Notes:

Append to existing tags, do not replace (unless replacing an old tag with a corrected version, e.g., TBD - Size → Mid-Market)
If a tag from the same category already exists, replace it (e.g., if deal is already tagged SMB and AI suggests Enterprise, replace SMB with Enterprise)
Do not apply low-confidence tags (under 0.5) automatically
Log the tag changes to deal activity feed: "AI applied tags: [Industry: Healthcare], [Size: Enterprise], [Urgency: High Priority]"

Output Contract

Return a SkillResult with:

data.suggested_tags: array of 3-8 tags
- category: string ("Industry Vertical" | "Deal Size Tier" | "Urgency Level" | "Product Interest" | "Buying Stage" | "Risk" | "Strategic")
- tag: string (the tag name from the taxonomy)
- confidence: number (0.0-1.0)
- reasoning: string (why this tag applies -- reference specific data)
- replace_existing: string | null (if this tag should replace an existing tag, name the old tag)
data.applied_tags: array | null
- If auto_apply is true: list of tags that were successfully applied
- If auto_apply is false: null (suggestion mode)

Quality Checklist

Before returning tag suggestions, verify:

Examples

Good Tag Suggestion (High Confidence)

Deal Data:

Company: "Memorial Hospital System"
Industry field: "Healthcare"
Deal value: $320,000
Close date: March 15, 2026 (28 days from now)
Transcript excerpt: "We need to replace our legacy patient data system before our HIPAA audit in April. Budget is approved at $300K. I'm working with our CTO and CFO to finalize the vendor decision."

Suggested Tags:

[
  {
    "category": "Industry Vertical",
    "tag": "Healthcare",
    "confidence": 1.0,
    "reasoning": "Company industry field is 'Healthcare' and company name is 'Memorial Hospital System'.",
    "replace_existing": null
  },
  {
    "category": "Deal Size Tier",
    "tag": "Enterprise",
    "confidence": 1.0,
    "reasoning": "Deal value is $320,000, which exceeds the Enterprise tier threshold ($250K+).",
    "replace_existing": "TBD - Size"
  },
  {
    "category": "Urgency Level",
    "tag": "Urgent",
    "confidence": 0.95,
    "reasoning": "Close date is 28 days away and buyer mentioned a hard deadline: 'before our HIPAA audit in April.' Regulatory compliance deadline creates urgency.",
    "replace_existing": null
  },
  {
    "category": "Product Interest",
    "tag": "Platform",
    "confidence": 0.80,
    "reasoning": "Buyer mentioned 'replace our legacy patient data system,' which maps to the Platform product.",
    "replace_existing": null
  },
  {
    "category": "Buying Stage",
    "tag": "Budget Approved",
    "confidence": 1.0,
    "reasoning": "Buyer explicitly stated 'Budget is approved at $300K.'",
    "replace_existing": null
  },
  {
    "category": "Buying Stage",
    "tag": "Executive Alignment",
    "confidence": 0.90,
    "reasoning": "Buyer is working with CTO and CFO (C-level executives) to finalize the vendor decision.",
    "replace_existing": null
  }
]

Bad Tag Suggestion (Low Confidence, Do Not Suggest)

Deal Data:

Company: "Tech Solutions Inc."
Industry field: (blank)
Deal value: (blank)
No meeting transcripts available
No activities in last 30 days

Suggested Tags:

[
  {
    "category": "Industry Vertical",
    "tag": "Technology",
    "confidence": 0.25,
    "reasoning": "Company name includes 'Tech,' but no other data available. Insufficient confidence to suggest this tag."
  }
]

Why this fails: Confidence is too low (0.25). Do not suggest tags below 0.5 confidence. If data is insufficient, return an empty suggestion list and recommend: "Insufficient data to suggest tags. Recommend: (1) Update company industry field, (2) Add deal value, (3) Log meeting notes or call summary to provide context."

Tag Replacement Example

Scenario: Deal is currently tagged SMB but new analysis shows deal value is $280K.

Suggested Tag:

{
  "category": "Deal Size Tier",
  "tag": "Enterprise",
  "confidence": 1.0,
  "reasoning": "Deal value is $280,000, which is in the Enterprise tier ($250K+). Replacing previous tag 'SMB' which was incorrect.",
  "replace_existing": "SMB"
}

Error Handling

Insufficient data for tagging

If deal has minimal data (no company, no value, no meetings, no activities):

Return empty suggested_tags array
Provide guidance: "Insufficient data to suggest tags. Recommend: (1) Update company and industry fields, (2) Add deal value, (3) Log meeting notes or activities to provide tagging context."

Conflicting data (e.g., company industry says "Healthcare" but transcript is clearly SaaS)

If industry field and transcript signals conflict:

Trust the TRANSCRIPT over the CRM industry field (industry fields are often wrong or outdated)
Suggest the correct tag with reasoning: "Company industry field says 'Healthcare' but meeting transcript shows SaaS product discussion. Tagging as 'SaaS' based on actual deal context."

Deal value is missing

If deal value is blank:

Do not guess. Tag as TBD - Size
Reasoning: "Deal value not set in CRM. Recommend updating deal value to determine size tier."

Multiple products discussed

If transcript shows interest in multiple products (e.g., "Platform" and "Analytics"):

Suggest BOTH product tags with confidence scores
Reasoning: "Buyer discussed both Platform ('need core data management') and Analytics ('want reporting dashboards'). Suggesting both product tags."

Existing tags conflict with suggestions

If deal is already tagged Financial Services but AI suggests Healthcare:

Flag the conflict: "Existing tag 'Financial Services' conflicts with AI suggestion 'Healthcare.' Review manually to confirm correct vertical."
Do not auto-replace in conflict scenarios (requires human judgment)

No recent activity (deal is stale)

If deal has no activity in 30+ days:

Suggest risk tag: Stalled
Reasoning: "No activity in [X] days. Tagging as 'Stalled' to flag for review."

Tone and Presentation

Be confident when data is clear: "Company industry is Healthcare. High confidence (1.0)."
Be honest when data is unclear: "Insufficient data to determine industry vertical. Recommend updating company profile."
Explain reasoning: "Budget is approved (buyer said '$300K allocated') → tag: Budget Approved."
Do not over-tag. 3-6 tags per deal is ideal. More than 8 tags dilutes value.
If a tag is low-confidence, do not suggest it. Better to have fewer accurate tags than many guessed tags.
Tag conflicts should be flagged, not silently resolved. "Existing tag conflicts with suggestion -- recommend manual review."
Tags are for segmentation and analysis, not storytelling. Keep them specific and standardized.