Enhance AI Email Classifier in Make.com
Budget: $30 – $250 USD
Full detail in notion https://www.notion.so/lasus/Freelancer-email-classification-with-AI-33bebe43945d80ac8248e313f0a12e87?v=4fb19831bc7347529f7aa3cc6d30470e&source=copy_link
# Make.com Expert Needed — AI Email Classifier Improvements (3 Enhancements)
## Overview
We have a working Make.com scenario that automatically classifies incoming emails using GPT-4o-mini and moves them to the correct folder in Zoho Mail. The core flow is functional (Webhook → GPT → Router → move email). We need an experienced Make.com developer to implement 3 specific improvements directly in the Make visual editor.
**Estimated scope:** 3–6 hours total. Fixed price preferred.
---
## Current State
- Scenario #4543102 in Make.com (team ID: 1835732)
- Classifies emails from a single inbox into 7 categories: `cotizacion`, `prioridad_alta`, `tarea_notion`, `acceso`, `notificaciones`, `spam_promocional`, `otros`
- Classification categories and destination folder IDs are currently hardcoded for a single user
- GPT-4o-mini is already connected and working
- Existing Make connections you'll reuse:
- Zoho Mail: "Admin FULL" (ID: 8076348)
- Notion: "My Notion Internal connection" (ID: 7541936)
- OpenAI: "Clasificador Correos" (ID: 8073039)
- MySQL: "Lasus MySQL" (ID: 7295310)
---
## Deliverables
### Improvement 1 — Pull classification rules dynamically from Notion
**Goal:** Before GPT classifies an email, fetch active rules from a Notion database and inject them into the GPT prompt. Users can then add/modify rules in Notion without touching Make.
**Notion data source:**
- Data source ID: `3aa6bfe9-b321-4227-9dbf-c3988f010ff0`
- Relevant fields: `Patron remitente` (sender pattern), `Tipo condicion` (domain / exact_email), `Tipo de regla` (target classification), `Activa` (checkbox, filter active only)
- Already has 5 example rules populated
**What to build in Make:**
1. Add a **Notion → Search Objects** module between the Webhook and the OpenAI module
- Search Data Source Items on data source `3aa6bfe9-b321-4227-9dbf-c3988f010ff0`
- Filter: `Activa = true`
- Limit: 50
2. Add a **Tools → Text Aggregator** after the Notion Search
- Source module: the Notion Search module above
- Row separator: new line
- Text row format: `If fromAddress contains '[Patron remitente]' → classify as '[Tipo de regla]'`
3. Modify the OpenAI prompt to prepend:
```
PRIORITY RULES (from Notion — apply BEFORE classifying):[Text Aggregator output]INSTRUCTION: If the sender matches any of the above patterns, apply that classification directly. Otherwise, classify based on the categories below.
```
**Expected result:** Adding a rule in Notion (e.g., "if sender contains doofinder.com → spam_promocional") is automatically applied without touching Make.
---
### Improvement 2 — Make the scenario multi-user with dynamic folder IDs
**Goal:** The same scenario should route emails for multiple users, each with their own Zoho Mail folders. Currently, folder IDs are hardcoded for a single account.
**Notion data source (user-folder mapping):**
- Data source ID: `b9989857-8c66-44c4-b873-cbcc8c8d4daa`
- Fields: `Email usuario`, `Clasificacion`, `Folder destino ID`, `Account ID Zoho`, `Activo`
- Already has 7 records for the current user (one per classification)
**What to build in Make:**
1. After the JSON Parse module, add a **Notion → Search Objects** module
- Data source ID: `b9989857-8c66-44c4-b873-cbcc8c8d4daa`
- Filter (3 conditions, AND):
- `Activo = true`
- `Clasificacion = [accion variable from JSON Parse]`
- `Email usuario = [toAddress from Webhook]`
- Limit: 1
2. Replace the 7 hardcoded Zoho Mail routes in the Router with **a single dynamic route** that uses:
- Account ID: `Account ID Zoho` from the Notion lookup
- Folder ID: `Folder destino ID` from the Notion lookup
3. Keep the special-action routes separate in the Router:
- `tarea_notion` → create task in Notion (existing)
- `cotizacion` → insert record in MySQL (existing)
**Expected result:** Adding a new user requires only: (a) configure outbound webhook in their Zoho Mail, (b) create 7 mapping records in Notion with their folder IDs. No changes to Make needed.
**Note to validate:** The "Admin FULL" Zoho Mail connection (ID: 8076348) needs to have permission to move emails across multiple organizational accounts. If it doesn't, a Zoho admin-level OAuth connection may be required — please flag this during implementation.
---
### Improvement 3 — Review and fix misclassifications
**Goal:** Audit recent execution history and improve classification accuracy for observed error patterns.
**Known issues:**
- Provider API notification emails classified as `notificaciones` instead of `otros`
- Some newsletters reaching the inbox not classified as `spam_promocional`
**What to do:**
1. Review execution history in Make scenario #4543102 and identify misclassification patterns
2. Adjust the GPT prompt in the OpenAI module based on the patterns found
3. Add specific rules to the Notion classification rules database for recurring cases
---
## Access You'll Need
- **Make.com:** Access to scenario #4543102 (team 1835732, org 6456907) — will be granted upon hiring
- **Notion workspace:** Read/write access to two databases (classification rules + user folder mapping) — will be granted upon hiring
- **No Zoho Mail access needed**
- **No MySQL access needed**
---
## Skills Required
- Make.com (intermediate to advanced) — Notion Search, Text Aggregator, Router, JSON modules
- Familiarity with OpenAI API prompt structure in Make
- Notion API via Make (Search Objects, filtering by data source)
- Basic understanding of Zoho Mail webhook flows (no configuration needed, just context)
---
## What We're NOT Looking for
- Rebuilding the scenario from scratch
- Python/JS development
- Changes to the Zoho Mail configuration
- Changes to the Notion database schemas
---
## Budget & Timeline
Fixed price proposals preferred. Please include:
- Your estimated hours per improvement
- Any blockers or questions based on this description (especially regarding multi-account Zoho permissions)
- Examples of Make scenarios you've built that involve multi-step Notion + OpenAI flows
# Make.com Expert Needed — AI Email Classifier Improvements (3 Enhancements)
## Overview
We have a working Make.com scenario that automatically classifies incoming emails using GPT-4o-mini and moves them to the correct folder in Zoho Mail. The core flow is functional (Webhook → GPT → Router → move email). We need an experienced Make.com developer to implement 3 specific improvements directly in the Make visual editor.
**Estimated scope:** 3–6 hours total. Fixed price preferred.
---
## Current State
- Scenario #4543102 in Make.com (team ID: 1835732)
- Classifies emails from a single inbox into 7 categories: `cotizacion`, `prioridad_alta`, `tarea_notion`, `acceso`, `notificaciones`, `spam_promocional`, `otros`
- Classification categories and destination folder IDs are currently hardcoded for a single user
- GPT-4o-mini is already connected and working
- Existing Make connections you'll reuse:
- Zoho Mail: "Admin FULL" (ID: 8076348)
- Notion: "My Notion Internal connection" (ID: 7541936)
- OpenAI: "Clasificador Correos" (ID: 8073039)
- MySQL: "Lasus MySQL" (ID: 7295310)
---
## Deliverables
### Improvement 1 — Pull classification rules dynamically from Notion
**Goal:** Before GPT classifies an email, fetch active rules from a Notion database and inject them into the GPT prompt. Users can then add/modify rules in Notion without touching Make.
**Notion data source:**
- Data source ID: `3aa6bfe9-b321-4227-9dbf-c3988f010ff0`
- Relevant fields: `Patron remitente` (sender pattern), `Tipo condicion` (domain / exact_email), `Tipo de regla` (target classification), `Activa` (checkbox, filter active only)
- Already has 5 example rules populated
**What to build in Make:**
1. Add a **Notion → Search Objects** module between the Webhook and the OpenAI module
- Search Data Source Items on data source `3aa6bfe9-b321-4227-9dbf-c3988f010ff0`
- Filter: `Activa = true`
- Limit: 50
2. Add a **Tools → Text Aggregator** after the Notion Search
- Source module: the Notion Search module above
- Row separator: new line
- Text row format: `If fromAddress contains '[Patron remitente]' → classify as '[Tipo de regla]'`
3. Modify the OpenAI prompt to prepend:
```
PRIORITY RULES (from Notion — apply BEFORE classifying):[Text Aggregator output]INSTRUCTION: If the sender matches any of the above patterns, apply that classification directly. Otherwise, classify based on the categories below.
```
**Expected result:** Adding a rule in Notion (e.g., "if sender contains doofinder.com → spam_promocional") is automatically applied without touching Make.
---
### Improvement 2 — Make the scenario multi-user with dynamic folder IDs
**Goal:** The same scenario should route emails for multiple users, each with their own Zoho Mail folders. Currently, folder IDs are hardcoded for a single account.
**Notion data source (user-folder mapping):**
- Data source ID: `b9989857-8c66-44c4-b873-cbcc8c8d4daa`
- Fields: `Email usuario`, `Clasificacion`, `Folder destino ID`, `Account ID Zoho`, `Activo`
- Already has 7 records for the current user (one per classification)
**What to build in Make:**
1. After the JSON Parse module, add a **Notion → Search Objects** module
- Data source ID: `b9989857-8c66-44c4-b873-cbcc8c8d4daa`
- Filter (3 conditions, AND):
- `Activo = true`
- `Clasificacion = [accion variable from JSON Parse]`
- `Email usuario = [toAddress from Webhook]`
- Limit: 1
2. Replace the 7 hardcoded Zoho Mail routes in the Router with **a single dynamic route** that uses:
- Account ID: `Account ID Zoho` from the Notion lookup
- Folder ID: `Folder destino ID` from the Notion lookup
3. Keep the special-action routes separate in the Router:
- `tarea_notion` → create task in Notion (existing)
- `cotizacion` → insert record in MySQL (existing)
**Expected result:** Adding a new user requires only: (a) configure outbound webhook in their Zoho Mail, (b) create 7 mapping records in Notion with their folder IDs. No changes to Make needed.
**Note to validate:** The "Admin FULL" Zoho Mail connection (ID: 8076348) needs to have permission to move emails across multiple organizational accounts. If it doesn't, a Zoho admin-level OAuth connection may be required — please flag this during implementation.
---
### Improvement 3 — Review and fix misclassifications
**Goal:** Audit recent execution history and improve classification accuracy for observed error patterns.
**Known issues:**
- Provider API notification emails classified as `notificaciones` instead of `otros`
- Some newsletters reaching the inbox not classified as `spam_promocional`
**What to do:**
1. Review execution history in Make scenario #4543102 and identify misclassification patterns
2. Adjust the GPT prompt in the OpenAI module based on the patterns found
3. Add specific rules to the Notion classification rules database for recurring cases
---
## Access You'll Need
- **Make.com:** Access to scenario #4543102 (team 1835732, org 6456907) — will be granted upon hiring
- **Notion workspace:** Read/write access to two databases (classification rules + user folder mapping) — will be granted upon hiring
- **No Zoho Mail access needed**
- **No MySQL access needed**
---
## Skills Required
- Make.com (intermediate to advanced) — Notion Search, Text Aggregator, Router, JSON modules
- Familiarity with OpenAI API prompt structure in Make
- Notion API via Make (Search Objects, filtering by data source)
- Basic understanding of Zoho Mail webhook flows (no configuration needed, just context)
---
## What We're NOT Looking for
- Rebuilding the scenario from scratch
- Python/JS development
- Changes to the Zoho Mail configuration
- Changes to the Notion database schemas
---
## Budget & Timeline
Fixed price proposals preferred. Please include:
- Your estimated hours per improvement
- Any blockers or questions based on this description (especially regarding multi-account Zoho permissions)
- Examples of Make scenarios you've built that involve multi-step Notion + OpenAI flows