Automatic qualification with JSON and rules
Automation 11 min
0/21
Business problem
Manual qualification isn't repeatable: 2 salespeople qualify the same lead differently, and the same salesperson qualifies differently on Monday than on Friday. AI without rules is also inconsistent: once it says 'qualified', once not — depending on the prompt. Without structured output you can't automate further steps.
Proven method
Automatic qualification in 3 layers:
- DATA: AI analyzes company (name, industry), industry, size, role (the contact's role), website, description (from enrichment),
- AI OUTPUT (structured JSON): {
icp_score
: 87,qualified
: true,reason
:target industry + size in range + hiring signal
,priority
:HIGH
}, - DETERMINISTIC RULES (workflow, not AI): if icp_score >= 70 → go to campaign; if 50-69 → nurture; if < 50 → don't contact (mark); if qualified=false → reject with a reason; if priority=HIGH → mark for quick contact. Rules:
- AI scores, rules decide (AI doesn't route itself — the workflow interprets JSON through IF/THEN),
- the schema has to be stable (the same JSON structure always),
- the 'qualified' definition is yours (what 70 points means in your model),
- the reason (justification) is critical — without it you don't know why AI scored it that way.
Market example
- BAD
- Manual, inconsistent execution or chaotic automation lacking data validation.
- GOOD
- Lead: 'MebelSoft, e-commerce, 40 people, Head of Customer Service, shop.meble.pl, sells furniture online'. AI JSON: {
icp_score
: 87,qualified
: true,reason
:e-commerce in ICP, 40 people in range (10-100), decision role, no chat on the site — a pain hypothesis
,priority
:HIGH
}. Rules: score 87 ≥ 70 → campaign; priority HIGH → task 'contact today'. Another lead: 'Janex, construction, 3 people, no website' → score 31 → rule: < 50 → don't contact, mark in the CRM with a reason.
Ready-to-use prompt
You are a qualification architect. Context: My ICP: [CARD]. My thresholds: [HOW I DEFINE A GOOD LEAD]. I use: [AI + MAKE/N8N + CRM]. Task: Design automatic qualification: 1. Input data: what AI analyzes (company, industry, size, role, website, description) 2. A JSON schema: the output structure (icp_score, qualified, reason, priority) with types 3. A qualification prompt: how to ask AI for a justification (the reason must be specific) 4. Deterministic rules: IF/THEN based on score and priority (workflow) 5. Thresholds: what 70+ / 50-69 / <50 means in my business + actions per threshold 6. Error handling: what if AI returns a bad format / no data (fallback) 7. Test: 5 sample leads through the system (input → JSON → rule decision) Format: 7 sections.
Model rekomendowany: Claude 3.5 Sonnet / GPT-4o Copy and paste into ChatGPT, Claude or Perplexity
ALTERNATYWNE SCENARIUSZE
Variations & alternative scenarios
No-rules version: AI scoring + manual routing (less risk).
Full version: rules + routing + marking in the CRM.
Practical exercise
Write the qualification JSON schema (icp_score, qualified, reason, priority). Write thresholds and IF/THEN rules for your business. Test on 3 sample leads (manually simulating AI).
Zapisano w przeglądarce