---
title: CRM Pipeline & Lead Scoring
url: https://blog.guigpap.com/en/workflows/crm-automation/
url_md: https://blog.guigpap.com/en/workflows/crm-automation.md
category: automation
date: '2026-03-28'
maturite: production
techno:
  - n8n
  - odoo
  - claude
application:
  - automation
  - business
---

# CRM Pipeline & Lead Scoring

> Odoo CRM pipeline automation with AI scoring, follow-ups and lead enrichment

## 1. What? — Definition and context

A lead lands in Odoo — web form, Cal.com, manual import. And then? Without automation, it sits in "New" until someone remembers to handle it. Scoring is subjective, follow-ups are forgotten, and lost leads are never reached out to again.

The **CRM Pipeline Automation** brings together two complementary workflows: a system for scoring and automatic pipeline progression, and AI enrichment of incoming leads.

> **Note - Lead scoring**
>
> **Lead scoring** assigns a numeric score (0-100) to each prospect based on objective criteria: complete profile, acquisition source, estimated budget, recent activity. The higher the score, the more attention the lead deserves.

### The two workflows

| Workflow | Nodes | Trigger | Role |
|----------|-------|---------|------|
| **CRM Pipeline Automation** | ~35 | Hourly + daily + webhook + monthly | Scoring, progression, follow-ups, reactivation |
| **Lead Enrichment** | ~15 | Webhook (lead creation) | AI enrichment via website |

### Architecture

```mermaid
flowchart TD
  Lead["New Odoo lead"]

  subgraph Enrich["Lead Enrichment · creation webhook"]
    direction TB
    Fetch["Fetch website"]
    Claude["Claude analysis"]
    Update["Update Odoo · industry, size, budget"]
  end

  subgraph Pipe["CRM Pipeline Automation · hourly schedule"]
    direction TB
    Score["Score 0-100"]
    Stage["Auto stage"]
    Follow["AI follow-up"]
    Reactivate["Monthly reactivation"]
  end

  Lead --> Enrich
  Lead --> Pipe
  Fetch --> Claude --> Update
```

---

## 2. Why? — Stakes and motivations

### Problems solved

| Problem | Without automation | With automation |
|---------|--------------------|-----------------|
| **Subjective scoring** | "This lead looks good" | 0-100 score on objective criteria |
| **Forgotten leads** | Sit in "New" for weeks | Automatic progression by signals |
| **Manual follow-ups** | Forget to follow up after 7 days | Claude generates the follow-up email |
| **Lost leads** | Never recontacted | Monthly reactivation of 3+ month leads |
| **Empty leads** | No info on the company | Auto enrichment via the website |

> **Tip - Four cadences**
>
> The pipeline runs at four speeds: real-time enrichment (webhook), hourly scoring, daily follow-ups, and monthly reactivation. Each frequency matches a different need.

---

## 3. How? — Technical implementation

### Scoring (hourly)

Every hour, the workflow recomputes the score of every active lead along five axes:

| Axis | Max points | Criteria |
|------|------------|----------|
| **Profile** | 20 | Email (+5), phone (+5), company (+5), role (+5) |
| **Source** | 15 | Referral (15), Cal.com (12), web form (10), import (5) |
| **Budget** | 25 | +50k (25), 20-50k (20), 10-20k (15), 5-10k (10), -5k (5) |
| **Activity** | 15 | Email received within 7d (15), 14d (10), 30d (5) |
| **Tags** | 10 | Tags "decision-maker" (+5), "budget-confirmed" (+5) |

The score is stored in the `x_score` field with the calculation date.

### Automatic progression

When certain signals are detected, the lead automatically advances in the pipeline:

| Detected signal | Progression |
|-----------------|-------------|
| Email reply received | New → Qualified |
| Meeting scheduled (Cal.com) | Qualified → Proposition |
| Quote sent in Odoo | Proposition → Negotiation |
| Contract signed | Negotiation → Won |

### AI follow-ups (daily, 9 AM)

Every morning, the workflow identifies leads with no activity for 7+ days. For each one, Claude (haiku model) generates a personalised follow-up email based on the lead's activity history. The email is created as a draft in Odoo — it is never sent automatically.

### Reactivation (monthly)

On the first Monday of the month, leads marked "Lost" for more than 3 months are analysed by Claude. The AI excludes "Not interested" or "Competitor won" leads and proposes a re-engagement angle for the others. The `x_reactivation_count` counter prevents endless reach-outs.

### Lead enrichment

When a new lead is created in Odoo (or when the company name is missing), a webhook triggers enrichment:

1. Domain extraction from email (gmail, yahoo etc. are excluded)
2. Fetch the website (homepage + /about as fallback)
3. Claude analyses the content and extracts: industry, estimated size, products, pain points, estimated budget
4. Odoo fields are updated with a confidence score (0-1)

> **Caution - Confidence threshold**
>
> If the Claude confidence score is below 0.3, the enrichment is NOT applied. The data is logged but the lead is not modified — to avoid polluting the CRM with dubious information.

---

## 4. What if? — Outlook and limits

### Current limits

| Limit | Impact | Mitigation |
|-------|--------|------------|
| **Free email providers** | Enrichment not possible | LinkedIn fallback (DuckDuckGo search) |
| **Follow-ups as drafts** | No automatic send | Deliberate choice — human validation |
| **Static scoring** | No learning from conversions | Eventually: adjust weights from history |

### Evolution scenarios

**If lead volume grows**:
- Prioritise enrichment by score
- Batch processing instead of per-webhook
- Configurable rate limiting (currently 50/h)

**If a sales team is involved**:
- Round-robin assignment by score
- Differentiated notifications per salesperson
- Scoring dashboard inside Odoo

---

## Related pages

### Workflows
- [Cal.com → Odoo CRM](/en/workflows/cal-rdv-odoo/) — Cal.com lead source
- [Website Lead Notification](/en/workflows/website-lead-notification/) — Web-form notifications
- [Notification Hub](/en/workflows/notification-hub/) — Alert routing

### Infrastructure
- [Odoo 18 on Docker](/en/infrastructure/odoo-18-setup/) — CRM configuration
- [AI Stack](/en/infrastructure/ai-stack/) — Claude for enrichment and follow-ups

## Metadonnees agent

- Cet article est issu du blog GuiGPaP Lab.
- Contexte global du blog: https://blog.guigpap.com/llms.txt
- Contact auteur: https://odoo.guigpap.com/mon-cv
- Licence: CC-BY-SA 4.0