Automatic enrichment with scoring
Automation 12 min
0/21
Business problem
Manual lead enrichment is 10-15 minutes. At 50 leads a day that's 8+ hours of pure administrative work. Plus the data is inconsistent (the same company as 'X Sp. z o.o.' and 'X'), duplicated (the same lead twice), and scoring 'from the head' isn't repeatable. The result: a CRM full of dirty data that no one uses.
Proven method
The enrichment workflow in 6 steps:
- NEW LEAD: a trigger (form, import, manual add) — the lead enters the system,
- ENRICH COMPANY: a tool (Apollo/Hunter/Clay/API) adds: industry, size, location, technology, website,
- ENRICH CONTACT: email verification (SMTP), LinkedIn, role, phone,
- AI RESEARCH: AI generates context (signals, pain hypotheses, a brief) — from the data in steps 2-3,
- ICP SCORE: AI scores the lead by your model (structured output — JSON: score, qualified, priority, reason),
- SAVE TO CRM: the workflow saves the lead + all data + score to the CRM (normalization: unified names, formats). TECHNICAL TOPICS: NORMALIZATION ('X sp. z o.o.' = 'X' — rules and AI), DEDUP (the same lead doesn't enter twice — by email/NIP/name), AI CLASSIFICATION (lead → segment/category), LEAD SCORING (by the model with weights), STRUCTURED OUTPUT (AI returns JSON — so you can route deterministically). AI provides: research, classification, scoring, name normalization. Tools provide: hard data (email, LinkedIn). The rule: hard data from tools, interpretation from AI, both saved together.
Market example
- BAD
- Manual, inconsistent execution or chaotic automation lacking data validation.
- GOOD
- Lead 'Nowak' comes from a form. Workflow: Enrich Company → 'MebelSoft Sp. z o.o., 40 employees, e-commerce, Shopify'. Enrich Contact → verified email (95% status), LinkedIn found. AI Research → 'signals: hiring 2 people in customer service, no chat on the site, email contact; pain hypothesis: slow responses'. ICP Score (JSON): {
score
: 87,qualified
: true,priority
:HIGH
,reason
:ICP match + hiring + no tools
}. CRM → a saved lead with full data + a task 'verify and call'. All in 2 minutes instead of 15.
Ready-to-use prompt
You are an enrichment architect. Context: I have [X] leads/month. Sources: [FORM/IMPORT/MANUAL]. I use: [TOOLS — Apollo/Hunter/Sheets/CRM]. My ICP: [ICP CARD]. Task: Design the enrichment workflow: 1. A schema: New Lead → Enrich Company → Enrich Contact → AI Research → ICP Score → Save to CRM (with tools per step) 2. Fields for enrichment: company and contact (what I collect) 3. Normalization: rules and how AI unifies names/data 4. Dedup: what I check for duplicates (email/NIP/name) + what I do with a duplicate 5. AI classification: what AI classifies (segment, industry, category) — structured output (a JSON schema) 6. Lead scoring: the model + how AI returns score/qualified/priority/reason (JSON) 7. Save to CRM: what goes into the CRM and what tasks the workflow creates 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-tools version: Sheets + AI API + manual verification.
Full version: Clay/Apollo + n8n/Make + a CRM with automation.
Practical exercise
Design the enrichment workflow (a schema with tools per step). Define the fields to collect (company/contact) and the dedup rules. Write the JSON schema AI should return for scoring (score, qualified, priority, reason).
Zapisano w przeglądarce