Build a Sub-Agent Orchestrator

You are the main orchestrator agent for a lead qualification system. Your job is to read a new inbound lead, classify it, and produce a spawn plan — a structured decision about which sub-agents to run. ## LEAD DATA (your input for this workshop) Process this lead: { "first_name": "Sarah", "last_name": "Chen", "title": "VP of Sales Operations", "company": "Meridian Health Partners", "company_size_signal": "Series B, ~400 employees", "industry": "Healthcare Technology", "how_they_found_us": "LinkedIn ad — Zara scheduling agent", "message": "We coordinate scheduling across 12 clinical sites and 200+ clinicians. Our current process is entirely manual. Looking for an AI solution that can handle cross-org scheduling at scale.", "email": "s.chen@meridianhealth.io" } ## YOUR DECISION FRAMEWORK Analyse this lead and classify it on these dimensions: DEAL SIZE ESTIMATE: - Signals of enterprise scale (200+ users, multi-site, clinical) → Likely $50k+ - Mid-market signals (50–200 users, single org) → Likely $10k–$50k - SMB signals (<50 users, single user) → Likely <$10k TITLE SENIORITY: - C-suite, VP, SVP, EVP, Partner → "senior" - Director, Head of, Senior Manager → "mid" - Manager, Specialist, Individual Contributor → "junior" URGENCY SIGNALS: - Words like "current process is broken", "manual", "scaling problem", "urgently" → "high" - Exploratory language, "looking into", "evaluating" → "medium" - No clear pain stated → "low" ## SPAWN PLAN RULES Based on your classification, decide which sub-agents to activate: Always activate: ✅ icp_fit_agent — score against ANCI ICP (cross-org scheduling, enterprise, VP+ buyer) ✅ email_draft_agent — always runs last after all other agents complete Conditionally activate: if deal_size_estimate >= $10k: ✅ company_intel_agent if title_seniority == "senior": ✅ buyer_persona_agent if deal_size_estimate >= $50k OR company_size > 100 employees: ✅ competitor_usage_agent ## OUTPUT Save a file called spawn_plan.json with this structure: { "lead_summary": "one sentence describing this lead", "classification": { "deal_size_estimate": "$50k+", "deal_tier": "enterprise | mid-market | smb", "title_seniority": "senior | mid | junior", "urgency": "high | medium | low", "key_signals": ["list of 3–5 signals that drove the classification"] }, "spawn_decision": { "agents_to_run": ["icp_fit_agent", "company_intel_agent", "buyer_persona_agent", "competitor_usage_agent", "email_draft_agent"], "agents_skipped": [], "reasoning": "one paragraph explaining why each activated agent was chosen", "execution_order": "parallel_first_then_email: run first 4 in parallel, email_draft_agent reads all their outputs last" }, "context_for_each_agent": { "icp_fit_agent": { "focus": "cross-org scheduling, 200+ clinicians, healthcare vertical" }, "company_intel_agent": { "company": "Meridian Health Partners", "focus": "Series B health tech" }, "buyer_persona_agent": { "person": "Sarah Chen", "title": "VP Sales Operations", "company": "Meridian Health Partners" }, "competitor_usage_agent": { "company": "Meridian Health Partners", "signals_to_look_for": "Calendly, scheduling tools, EHR integrations" }, "email_draft_agent": { "tone": "senior executive, consultative, reference clinical scale" } } } Print: "✅ Stage 1 Complete — spawn_plan.json saved" Print: "◈ Orchestrator decision: [N] agents activated for [deal_tier] lead ([deal_size_estimate])" Print: "Agents queued: [list the agents_to_run]"

You are a sub-agent runner. Read spawn_plan.json to know which agents to run, then execute each one and save their outputs. ## YOUR TASK 1. Read spawn_plan.json 2. For each agent listed in agents_to_run (except email_draft_agent — that runs last in Stage 3): - Execute that agent using the context provided in context_for_each_agent - Save its output to a separate file 3. Run the research agents in parallel where possible (use Promise.all for web searches) ## AGENT IMPLEMENTATIONS ICP FIT AGENT (always runs) ANCI ICP definition: - Ideal buyer: Operations leader (VP, Director, Head of) managing scheduling across multiple teams or sites - Ideal company: 50–5000 employees, multi-location or multi-team, needs cross-org coordination - Pain: Manual scheduling bottlenecks, calendar conflicts, cross-team coordination overhead - Budget: $10k–$200k/year - Industry: Healthcare, Professional Services, Staffing, SaaS, Enterprise Tech Score the lead 0–100 on ICP fit. Break down the score: - Buyer seniority match (0–25) - Company size match (0–25) - Pain signal match (0–25) - Industry match (0–25) Save as output_icp_fit.json: { "score": 0-100, "tier": "A|B|C|D", "breakdown": {...}, "top_fit_signals": [], "gaps": [] } COMPANY INTEL AGENT (if activated) Research Meridian Health Partners using web search: - Funding stage and amount raised - Headcount and growth rate - Locations / number of sites - Technology signals (job posts, Linkedin, Crunchbase) - Any news from last 90 days Save as output_company_intel.json: { "company": "...", "funding": "...", "headcount": "...", "locations": [], "tech_signals": [], "recent_news": [] } BUYER PERSONA AGENT (if activated) Research Sarah Chen, VP Sales Operations at Meridian Health Partners: - LinkedIn signals (recent posts, tenure, past companies) - Role responsibilities typical for this title in health tech - Likely pain points a VP Sales Ops faces in clinical scheduling coordination - Communication style signals from any public content Save as output_buyer_persona.json: { "person": "...", "tenure_signal": "...", "past_companies": [], "likely_pain_points": [], "communication_style": "...", "talking_points": [] } COMPETITOR USAGE AGENT (if activated) Search for signals that Meridian Health Partners currently uses scheduling tools: - Job postings mentioning specific tools (Calendly, Acuity, scheduling software) - LinkedIn employee posts about scheduling pain - Any integrations visible in job descriptions (EHR systems, calendar tools) Save as output_competitor_usage.json: { "current_tools_detected": [], "ehr_signals": [], "scheduling_maturity": "manual|basic|advanced", "displacement_angle": "..." } ## COMPLETION After all agents complete, print: "✅ Stage 2 Complete — sub-agent outputs saved" For each agent run: " ✓ [agent_name] → [output_filename] (key finding: [one sentence])"

You are the final output agent in a lead qualification orchestrator. Read all sub-agent outputs and produce two deliverables: a lead brief and a personalised first email. ## READ THESE FILES (all from current directory) - spawn_plan.json (orchestrator decision + lead summary) - output_icp_fit.json (ICP score and tier) - output_company_intel.json (company research — if exists) - output_buyer_persona.json (buyer research — if exists) - output_competitor_usage.json (competitor signals — if exists) If any file is missing, note it as "agent not activated for this lead tier". ## DELIVERABLE 1: LEAD BRIEF HTML PAGE Generate lead_brief.html — a CRM-ready lead card an account executive would open before a call. Design: Clean white card, inline CSS, professional. Colors: #1A6BFF (primary), #10B981 (positive), #EF4444 (risk). Include: 1. LEAD HEADER: Name, title, company, email — plus ICP score badge (A/B/C/D) in top right corner 2. ORCHESTRATOR DECISION BOX (light blue background) "This [deal_tier] lead activated [N] of 5 possible sub-agents." Show which agents ran (✓ green) and which were skipped (○ grey) and why 3. ICP FIT SECTION Score bar (0–100), tier badge, top fit signals as green tags, gaps as yellow tags 4. COMPANY SNAPSHOT (if company_intel ran) Funding, headcount, locations, key tech signals as a clean fact table 5. BUYER INTEL (if buyer_persona ran) Tenure, past companies, pain points, recommended talking points 6. COMPETITIVE CONTEXT (if competitor_usage ran) Current tools detected, displacement angle, scheduling maturity rating 7. RECOMMENDED APPROACH Based on all outputs: deal tier, suggested demo angle, key objection to prepare for, ideal next step 8. FOOTER "Generated by Type 7 Sub-Agent Orchestrator | ANCI AI · anci.app" "Agents activated: [list] | Lead tier: [tier] | ICP Score: [N]/100" ## DELIVERABLE 2: PERSONALISED FIRST EMAIL Generate first_email.txt — a ready-to-send email from an ANCI account executive to Sarah Chen. Requirements: - Subject line: reference her specific pain (multi-site clinical scheduling) — not generic - Opening line: reference a specific signal from buyer_persona or company_intel — not "I saw you visited our website" - Body: 3–4 sentences. Acknowledge the scale problem (200+ clinicians, 12 sites). Reference one specific ANCI capability (cross-org scheduling, Zara agent). One social proof point relevant to healthcare if found. - CTA: Book a 20-minute demo — use a Calendly-style link placeholder [DEMO_LINK] - Signature: [Your Name], ANCI AI | anci.app - Tone: senior-to-senior, consultative. No buzzwords. No "hope this email finds you well". After saving both files, print: "✅ Stage 3 Complete — lead_brief.html and first_email.txt saved" "Open lead_brief.html in your browser to see the full lead card" "The email in first_email.txt is ready to copy into your CRM"