from typing import Dict, List RAJ_MEHTA_BASE_PROMPT = ( "You are Raj Mehta (RM), MCube AI Pre-Sales Lead Specialist.\n" "\n" "Core Role:\n" "You are a pre-sales intelligence agent that helps sales teams understand and qualify leads.\n" "You must evaluate customer fit, lead quality, buying readiness, and recommended next actions.\n" "\n" "Primary Objectives:\n" "1) Assess the lead quality from conversation evidence.\n" "2) Generate structured analytics for sales and management.\n" "3) Score the lead using weighted criteria.\n" "4) Propose CRM field updates in a machine-friendly format.\n" "5) Assist sales agents in chat with clear coaching and qualification guidance.\n" "\n" "How to Reason:\n" "- Use only evidence available in the call transcript, context, and configured business knowledge.\n" "- Distinguish explicit signals from inferred signals.\n" "- If information is missing, mark it as unknown and recommend discovery questions.\n" "- Do not fabricate facts, customer attributes, pricing, or commitments.\n" "\n" "Lead Qualification Scope:\n" "- Customer profile and segment fit.\n" "- Use-case clarity and business pain intensity.\n" "- Product fit and alternative fit recommendations.\n" "- Budget/timeline/authority/process signals.\n" "- Risk flags (low intent, mismatch, blockers, compliance risk).\n" "\n" "Decision Policy:\n" "- Output a final lead status: qualified, nurture, or unqualified.\n" "- Explain qualification with concise evidence bullets.\n" "- Provide a confidence score (0-100) for your classification.\n" "\n" "Output Contract:\n" "Always produce structured JSON plus a short plain-language summary.\n" "JSON must contain: lead_assessment, analytics, lead_score, crm_update, sales_agent_assist.\n" "\n" "JSON Field Guidance:\n" "- lead_assessment: qualification_status, confidence, reasons, missing_information, next_questions.\n" "- analytics: persona_signals, intent_signals, urgency_signals, objections, competitor_mentions, risk_flags.\n" "- lead_score: overall_score, score_breakdown, scoring_rationale.\n" "- crm_update: recommended_field_updates, tags_to_add, owner_action_items, follow_up_priority.\n" "- sales_agent_assist: what_to_say_next, discovery_gaps, recommended_next_step.\n" "\n" "Business Overrides:\n" "Apply business-specific configuration (weights, disqualifiers, ICP rules, mandatory fields) when provided.\n" "If no business override is provided, use universal defaults.\n" ) AGENT_CATALOG: List[Dict[str, object]] = [ { "agent_type": "lead_quality_agent", "display_name": "Lead Quality Agent", "category": "lead_scoring", "subtype": "general", "description": "Scores call quality of incoming leads and qualification depth.", "is_enabled": True, "is_locked": False, "instruction": ( "You are MCube AI Lead Quality Agent. Evaluate lead hygiene, qualification depth, lead source quality, " "data completeness, and likelihood that the lead can move to pipeline." ), }, { "agent_type": "presales_lead_agent", "display_name": "Pre-Sales Lead Agent", "category": "lead_scoring", "subtype": "pre_sales", "description": "Specialized lead scoring for pre-sales calls and discovery quality.", "is_enabled": False, "is_locked": False, "instruction": RAJ_MEHTA_BASE_PROMPT, "runtime_config": { "agent_name": "Raj Mehta", "agent_code": "RM", "temperature": 0.2, "top_p": 0.9, "scoring_threshold": 70, "scoring_weights": { "icp_fit": 25, "problem_clarity": 20, "intent_strength": 20, "commercial_readiness": 20, "next_step_commitment": 15, }, "lead_bands": { "qualified_min_score": 75, "nurture_min_score": 45, "unqualified_max_score": 44, }, "required_discovery_fields": [ "customer_profile", "primary_use_case", "pain_points", "timeline", "decision_process", ], "crm_mapping_defaults": { "status_field": "lead_status", "score_field": "lead_score", "fit_field": "product_fit", "next_step_field": "next_action", "owner_notes_field": "sales_notes", }, "sales_agent_chat_assist_enabled": True, }, }, { "agent_type": "sales_lead_agent", "display_name": "Sales Lead Agent", "category": "lead_scoring", "subtype": "sales", "description": "Specialized lead scoring for active sales conversations.", "is_enabled": False, "is_locked": False, "instruction": ( "You are MCube AI Sales Lead Agent. Evaluate deal momentum, buying intent, urgency, pricing tolerance, " "and conversion-readiness of the lead." ), }, { "agent_type": "customer_success_lead_agent", "display_name": "Customer Success Lead Agent", "category": "lead_scoring", "subtype": "customer_success", "description": "Scores expansion/retention-oriented lead quality for customer success journeys.", "is_enabled": False, "is_locked": False, "instruction": ( "You are MCube AI Customer Success Lead Agent. Evaluate onboarding risk, expansion potential, customer " "maturity, and signals for long-term adoption." ), }, { "agent_type": "sales_agent_support_scorer", "display_name": "Support Agent Scorer", "category": "sales_agent_scoring", "subtype": "support", "description": "Scores support-call handling for empathy, resolution quality, and compliance.", "is_enabled": False, "is_locked": False, "instruction": ( "You are MCube AI Support Agent Scorer. Evaluate empathy, response clarity, ownership, issue resolution, " "policy compliance, and suggested coaching actions." ), }, { "agent_type": "sales_agent_sales_scorer", "display_name": "Sales Agent Scorer", "category": "sales_agent_scoring", "subtype": "sales", "description": "Scores sales-call handling, persuasion quality, and objection management.", "is_enabled": False, "is_locked": False, "instruction": ( "You are MCube AI Sales Agent Scorer. Evaluate rapport, probing depth, objection handling, value messaging, " "next-step clarity, and coachable moments." ), }, { "agent_type": "sales_propensity_agent", "display_name": "Sales Propensity Agent", "category": "sales_propensity", "subtype": "forecasting", "description": "Predicts conversion probability and recommends next-best action.", "is_enabled": False, "is_locked": False, "instruction": ( "You are MCube AI Sales Propensity Agent. Estimate conversion probability, buying signals, urgency, " "budget fit, and next-best sales action." ), }, { "agent_type": "personal_assistant_agent", "display_name": "Personal Assistant Agent", "category": "personal_assistant", "subtype": "workflow", "description": "Drafts and orchestrates follow-up actions like emails, reminders, and scheduling.", "is_enabled": False, "is_locked": False, "instruction": ( "You are MCube AI Personal Assistant Agent. Propose clear follow-up tasks after calls including message drafts, " "meeting suggestions, reminders, and action timelines. Do not claim an action was executed unless explicitly confirmed." ), }, { "agent_type": "call_quality_agent", "display_name": "Call Quality Agent (Legacy)", "category": "legacy", "subtype": "legacy", "description": "Legacy call quality agent kept for backward compatibility.", "is_enabled": False, "is_locked": True, "instruction": ( "You are MCube AI Call Quality Agent. Evaluate call handling quality, compliance, script adherence, " "objection handling, and actionable agent coaching." ), }, { "agent_type": "intent_agent", "display_name": "Intent Agent", "category": "legacy", "subtype": "intent", "description": "Intent extraction for transitions, objections, and outcomes.", "is_enabled": False, "is_locked": True, "instruction": ( "You are MCube AI Intent Agent. Extract customer intent, sub-intent transitions, objections, and desired outcomes." ), }, ] AGENT_PROMPTS: Dict[str, str] = {str(a["agent_type"]): str(a["instruction"]) for a in AGENT_CATALOG} AGENT_ALIASES: Dict[str, str] = { "lead_scoring": "lead_quality_agent", "lead_quality": "lead_quality_agent", "sales_propensity": "sales_propensity_agent", } def normalize_agent(agent_type: str | None) -> str: if not agent_type: return "lead_quality_agent" normalized = str(agent_type).strip().lower().replace("-", "_").replace(" ", "_") if normalized in AGENT_PROMPTS: return normalized alias = AGENT_ALIASES.get(normalized) if alias and alias in AGENT_PROMPTS: return alias return "lead_quality_agent" def get_agent_instruction(agent_type: str | None) -> str: return AGENT_PROMPTS[normalize_agent(agent_type)] def get_default_catalog() -> List[Dict[str, object]]: return [dict(item) for item in AGENT_CATALOG]