Clozd

Count how many decision driver mentions fall into each category. Use when the user asks "How many driver mentions are in the Pricing category?" or wants a category-level frequency breakdown. Returns: an array of { categoryName, driverCount } sorted by driverCount descending. Data layer: PUBLISHED INTERVIEWS only — includes only deals with at least one published interview. Counts are not comparable to ALL CRM DEALS tools (e.g. get_workflow_deals).

get_decision_driver_category_counts

List individual decision drivers with aggregated sentiment distribution across a four-point scale (strong negative, negative, positive, strong positive). Use when the user asks which specific drivers are mentioned most often or most negatively, "What are the most common decision drivers?", "Which drivers come up most often?", or "What is the sentiment breakdown per driver?" Returns an array of drivers with driverName, totalMentions, and a sentimentBreakdown object containing strongNegative, negative, positive, and strongPositive counts. Sorted by totalMentions descending. Pass limit to return only the top N drivers. Supports filtering by outcome, competitor, date range, and other shared filters. Data layer: PUBLISHED INTERVIEWS only — includes only deals with at least one published interview. Counts are not comparable to ALL CRM DEALS tools (e.g. get_workflow_deals).

get_decision_drivers

Retrieve verbatim quotes from interview or survey responses linked to a specific decision driver or theme. Each quote includes the deal it came from plus respondent context such as title, channel, and publish date. Use when the user needs direct evidence, for example "Show me quotes about pricing" or "What did customers say about our support?" Provide either driverName for a case-insensitive text match or driverId for an exact UUID match. Data layer: PUBLISHED INTERVIEWS only — includes only deals with at least one published interview. Counts are not comparable to ALL CRM DEALS tools (e.g. get_workflow_deals). NOTE: PII redaction mode is enabled — personal names, emails, phone numbers, addresses, and company names in results are replaced with generic anonymized placeholders such as [Sales Rep], [Participant 1], [Company Name], and [Email Address].

get_driver_quotes

Retrieve Insight Gems — deals flagged with an actionable insight: At-Risk (likely to churn), Win-Back (lost but recoverable), Expansion (upsell opportunity), Feature Request (buyer asked for a specific product change), Sales Coaching (a coachable moment for the rep), or Case Study (a quotable customer success story). A program can also define its own custom gem types; they are returned alongside the built-ins and filtered the same way, by their custom_ key (e.g. custom_pricing_concern). Each gem carries gemKey (its type key) and gemName (its type's display name, which a program can change for a built-in type too; identify types by gemKey); the program's custom gem types also appear in counts by their custom_ key. list_filter_values with field gemTypes lists the program's valid types, and an unknown or unconfigured type returns an error listing them. Each gem has an open/resolved workflow status and by default only OPEN gems (still needing action) are returned — pass status to see resolved or all gems. Use when the user asks "Which deals are at risk?", "Show me win-back opportunities", or "What insight gems have we resolved?" Pass gemType to filter to one type (a built-in key or a custom gem's custom_ key), or omit for all types. Each gem includes its rationale (the reason it was flagged), who added it and how (addedByType: ai, clozd_consultant, or user), its owners (owner-field labels and literal emails) plus the resolved notification-recipient emails with notify/notificationDate, verbatim supporting quotes, any notes, and action taken / resolved-by details when resolved; counts per type are included by default. Data layer: ALL CRM DEALS — includes every CRM deal regardless of interview status. Counts are not comparable to PUBLISHED INTERVIEWS tools (e.g. get_deals, get_responses). Filtering by product or region is exact-match only. First call list_filters to see the filterable fields, then list_filter_values to enumerate a field's real values, and pass the exact values you want (arrays are supported). Do not guess names or use partial/wildcard strings. NOTE: PII redaction mode is enabled — personal names, emails, phone numbers, addresses, and company names in results are replaced with generic anonymized placeholders such as [Sales Rep], [Participant 1], [Company Name], and [Email Address].

get_insight_gems

List the Clozd programs available to the authenticated user. Call this first to discover programId values for later program-scoped tools.

get_programs

Retrieve AI-generated response summaries for published win-loss feedback. Pass responseId to fetch one response summary directly, or use shared deal and response filters to browse summaries in bulk with pagination. Data layer: PUBLISHED INTERVIEWS only — includes only deals with at least one published interview. Counts are not comparable to ALL CRM DEALS tools (e.g. get_workflow_deals). NOTE: PII redaction mode is enabled — personal names, emails, phone numbers, addresses, and company names in results are replaced with generic anonymized placeholders such as [Sales Rep], [Participant 1], [Company Name], and [Email Address].

get_response_summaries

Retrieve a paginated list of win-loss responses (interviews and surveys). Each response includes the linked deal, outcome, respondent role, channel, and publish date so the LLM can find specific feedback submissions without loading full transcripts. Data layer: PUBLISHED INTERVIEWS only — includes only deals with at least one published interview. Counts are not comparable to ALL CRM DEALS tools (e.g. get_workflow_deals). NOTE: PII redaction mode is enabled — personal names, emails, phone numbers, addresses, and company names in results are replaced with generic anonymized placeholders such as [Sales Rep], [Participant 1], [Company Name], and [Email Address].

get_responses

List tags applied to deals with their usage counts. Tags are user-defined labels (e.g. "churn risk", "enterprise", "product gap") that categorize deals for filtering and analysis. Use when the user asks for a summary of tags, "What tags do we have?", "Which tags appear most often?", "Give me a breakdown of our tags", "Summarize deals by tag", or "How are deals tagged?" — even if they use words like "summary", "breakdown", or "overview". Do NOT use get_decision_drivers for tag-related questions. Set includeDeals=true to also return the deal IDs associated with each tag. Returns a paginated list of { tagName, count, dealIds? }.

get_tags

Retrieve full-text transcripts of win-loss interviews or surveys. Returns the complete conversation content. WARNING: Transcripts can be very long. Prefer get_response_summaries for quick insights or get_driver_quotes for targeted evidence. Use when the user explicitly requests raw transcripts, e.g. "Show me the full transcript for deal X." Pass responseId for a single transcript, or use filters with pagination for bulk retrieval. Data layer: PUBLISHED INTERVIEWS only — includes only deals with at least one published interview. Counts are not comparable to ALL CRM DEALS tools (e.g. get_workflow_deals). NOTE: PII redaction mode is enabled — personal names, emails, phone numbers, addresses, and company names in results are replaced with generic anonymized placeholders such as [Sales Rep], [Participant 1], [Company Name], and [Email Address].

get_transcripts

Calculate win/loss rates for deals in a program. Returns the percentage of deals won vs. lost, with optional breakdown by a single dimension (e.g. industry, product, sales rep). Use when the user asks questions like "What is our win rate?" or "Which industry has the highest win rate?" Returns: overall winRate, lossRate, totalDeals, and a breakdown array when groupBy is specified. Data layer: PUBLISHED INTERVIEWS only — includes only deals with at least one published interview. Counts are not comparable to ALL CRM DEALS tools (e.g. get_workflow_deals).

get_win_rates

Retrieve participants for a specific deal in a workflow, including each participant's approval status (pending/approved/rejected), outreach status, and who approved or rejected them. Returns the deal's current workflow stage and a list of participants with their individual status details. Use after get_workflow_deals to drill into participant-level detail for a specific deal. Example queries: "Who are the participants for this deal?" or "Has anyone approved this deal yet?" Data layer: ALL CRM DEALS — includes every CRM deal regardless of interview status. Counts are not comparable to PUBLISHED INTERVIEWS tools (e.g. get_deals, get_responses). NOTE: PII redaction mode is enabled — personal names, emails, phone numbers, addresses, and company names in results are replaced with generic anonymized placeholders such as [Sales Rep], [Participant 1], [Company Name], and [Email Address].

get_workflow_deal_participants

Retrieve a paginated list of deals in a workflow with their stage, status, and participant count. Each deal includes: name, outcome, amount, sales rep, stage (needs_attention, needs_approval, ready_for_distribution, in_progress, completed), and a participant_count. To get full participant details (approval status, who approved/rejected), use the get_workflow_deal_participants tool with a specific dealId. Example queries: "Show me deals needing approval in this workflow" or "Which deals are ready for distribution?" Data layer: ALL CRM DEALS — includes every CRM deal regardless of interview status. Counts are not comparable to PUBLISHED INTERVIEWS tools (e.g. get_deals, get_responses). Filtering by product or region is exact-match only. First call list_filters to see the filterable fields, then list_filter_values to enumerate a field's real values, and pass the exact values you want (arrays are supported). Do not guess names or use partial/wildcard strings. NOTE: PII redaction mode is enabled — personal names, emails, phone numbers, addresses, and company names in results are replaced with generic anonymized placeholders such as [Sales Rep], [Participant 1], [Company Name], and [Email Address]. Search parameters that match on personal or company values are disabled.

get_workflow_deals

Retrieve all workflows in a program with their configuration, filter rules, settings, and deal/participant counts. Each workflow includes: name, status, feedback channel, interview type, deal rules, participant rules, outreach settings, contributor settings, and counts by stage (needs_attention, needs_approval, ready_for_distribution, in_progress, completed). Use when the user wants to understand the workflow pipeline for a program. Example queries: "Show me all active workflows" or "What are the outreach settings for our buyer workflows?"

get_workflows

List the valid or observed values for a single deal filter field in a program. Pass a field returned by list_filters. The response has: valueSource (enum | reference | crm_observed | metadata_picklist | metadata_observed), exhaustive (true when the values are the complete valid set, false when they are only values seen in the data), truncated (observed lists are capped at 100 by deal count), and values (each an optional dealCount; a null value is the blank/unspecified bucket). Call this to get exact filter values before filtering get_deals, so you pass real values instead of guessing strings.

list_filter_values

List the filterable fields for a program's deals — the standard Clozd fields (product, region, industry, segment, outcome, businessType, decision, gemTypes, leadSource). Each entry has: field (the key to pass to list_filter_values and to get_deals), displayName, valueType, and exhaustive (whether the field has a complete known value set). Call this first to discover what you can filter by, then call list_filter_values for a field's values before filtering get_deals. The catalog also includes the program's custom deal-metadata fields (valueType metadata_picklist or metadata_observed) — filter get_deals by these via its customFilters parameter.

list_filters

Search Clozd win-loss knowledge for relevant deals, feedback summaries, themes, competitors, and quotes using a natural-language query plus optional filters. Use this first when the user asks to find, explore, cite, compare, or summarize Clozd feedback but does not already have a specific record ID. Returns compact canonical record IDs that can be passed to fetch for more detail. NOTE: PII redaction mode is enabled — personal names, emails, phone numbers, addresses, and company names in results are replaced with generic anonymized placeholders such as [Sales Rep], [Participant 1], [Company Name], and [Email Address]. Search parameters that match on personal or company values are disabled.

search