# 5 n8n AI Agent Workflows That Run My Business

**I run an AI automation agency. These five n8n workflows handle 80% of my daily operations — from email triage to multi-language content pipelines. Here's how they work and how you can set them up.**

---

**Table of Contents**

1. [Workflow 1: The Email Triage Agent](#email)
2. [Workflow 2: Multi-Language Blog Pipeline](#blog)
3. [Workflow 3: CRM Auto-Enrichment](#crm)
4. [Workflow 4: Server Health Monitor](#server)
5. [Workflow 5: Social Media Cross-Poster](#social)
6. [How to Get Started](#start)

---

## Workflow 1: The Email Triage Agent

**Problem:** 50-80 business emails/day. 60% are newsletters, 20% are invoices, 15% are client messages, 5% are urgent.

**Solution:** An n8n workflow that reads Zoho Mail via IMAP, classifies every email using an AI agent (Qwen 3.8), and routes accordingly.

**How it works:**

```
[IMAP Trigger: New Email]
    → [AI Agent: Classify]
    ├── Spam → [Delete]
    ├── Newsletter → [Archive]
    ├── Invoice → [Extract PDF → Upload to WorkDrive → Delete]
    ├── Client → [Create CRM Contact → Notify Telegram]
    └── Urgent → [Forward to Steffen's Phone]
```

**Key nodes:**
- **IMAP Trigger:** Polls every 30 seconds
- **AI Agent (Qwen 3.8):** Classifies with this prompt:
  ```
  Classify this email: subject + first 200 chars of body.
  Return ONLY one of: SPAM, NEWSLETTER, INVOICE, CLIENT, URGENT.
  ```
- **HTTP Request (Zoho WorkDrive):** Uploads invoice attachments
- **Telegram:** Sends alert for CLIENT/URGENT with subject + 2-line summary

**Result:** 45 minutes/day saved. I now check email twice instead of constantly.

**Setup time:** 2 hours  
**Running cost:** ~€3/month (AI API calls)  
**Complexity:** ⭐⭐⭐

---

## Workflow 2: Multi-Language Blog Pipeline

**Problem:** I need weekly blog posts in German, English, and Spanish. Writing manually takes 4-6 hours per post.

**Solution:** An n8n workflow triggered by a Webhook that generates, translates, and deploys blog posts in all three languages.

**How it works:**

```
[Webhook Trigger: "Generate post about {topic}"]
    → [AI Agent: Write German post (2000 words)]
    → [AI Agent: Translate to English]
    → [AI Agent: Translate to Spanish]
    → [HTTP Request: Generate images (kie.ai)]
    → [Git: Commit all 3 MDX files to repo]
    → [Telegram: "Post live at {URLs}"]
```

**Key nodes:**
- **AI Agent Chain:** Three sequential calls, each with language-specific system prompts
- **HTTP Request (kie.ai):** Generates dark-mode neon featured images
- **Execute Command:** Runs `git add && git commit && git push`
- **Telegram:** Sends all three URLs

**Result:** One webhook call → 3 blog posts live in 12 minutes.

**Setup time:** 4 hours  
**Running cost:** ~€8/month (AI + image generation)  
**Complexity:** ⭐⭐⭐⭐

---

## Workflow 3: CRM Auto-Enrichment

**Problem:** New contacts enter Zoho CRM with just name and email. Manually adding company, role, LinkedIn takes 5 minutes each.

**Solution:** A webhook-triggered enrichment workflow.

**How it works:**

```
[Zoho CRM Webhook: New Contact]
    → [HTTP Request: Hunter.io (email → company)]
    → [HTTP Request: LinkedIn Sales Navigator (company → industry, size)]
    → [AI Agent: Generate contact summary]
    → [Zoho CRM: Update contact fields]
    → [Zoho CRM: Create follow-up task if CEO/CTO/VP]
```

**AI Agent prompt:**
```
Given this contact: {name}, {email}, {company}, {role}
Write a 2-sentence summary for CRM.
Generate 3 potential conversation openers based on company news.
```

**Result:** Every new contact gets enriched within 30 seconds. Zero manual data entry.

**Setup time:** 3 hours  
**Running cost:** ~€12/month (Hunter.io API + AI)  
**Complexity:** ⭐⭐⭐

---

## Workflow 4: Server Health Monitor

**Problem:** I run several servers. Need to know immediately if one goes down — not when a client notices.

**Solution:** A Cron-triggered watchdog that checks all servers every 5 minutes.

**How it works:**

```
[Cron Trigger: Every 5 minutes]
    → [Loop: For each server]
    ├── [HTTP Request: GET /health]
    ├── [SSH: Check disk, memory, CPU]
    └── [IF: Any check fails]
        ├── [AI Agent: Diagnose from logs]
        ├── [Telegram: Alert Steffen with diagnosis]
        └── [IF: Critical: Auto-restart service via SSH]
    → [IF: All OK → Log to database (rolling 7 days)]
```

**What it catches:**
- HTTP 5xx errors → restarts web server
- Disk >90% → cleans logs, alerts to upgrade
- Memory leak → restarts service, logs pattern
- SSL cert <7 days → emails to renew

**Result:** Zero unplanned downtime since deploying this. One 3 AM incident auto-resolved before I woke up.

**Setup time:** 5 hours  
**Running cost:** ~€1/month (minimal API calls)  
**Complexity:** ⭐⭐⭐⭐

---

## Workflow 5: Social Media Cross-Poster

**Problem:** A blog post is live. Now it needs to go on LinkedIn, Telegram, and optionally TikTok. Doing this manually takes 20 minutes per post.

**Solution:** One webhook triggers a cross-posting pipeline.

**How it works:**

```
[Webhook: "Post {slug} to social"]
    → [HTTP Request: Fetch blog post content]
    → [AI Agent: Generate LinkedIn post (1200 chars, professional)]
    → [AI Agent: Generate Telegram summary (300 chars)]
    → [AI Agent: Generate TikTok script (if video content)]
    → [HTTP Request: upload-post.com (LinkedIn)]
    → [HTTP Request: Telegram Bot API]
    → [IF: TikTok → HTTP Request: upload-post.com (TikTok)]
    → [Telegram: "Posted to {platforms}: {URLs}"]
```

**AI prompt for LinkedIn:**
```
Write a LinkedIn post about this blog article. 
- Max 1200 characters
- Hook in first 2 lines
- No emojis
- No em-dashes
- Link in first comment
- Call to action: "What's your experience with {topic}?"
Article: {extract}
```

**Result:** One click → every platform updated. Consistent brand voice across all channels.

**Setup time:** 3 hours  
**Running cost:** ~€5/month (upload-post.com API + AI)  
**Complexity:** ⭐⭐⭐

---

## How to Get Started

**You need:**
1. **n8n instance** — Self-hosted (free) or n8n Cloud (€20/month)
2. **API access to an LLM** — OpenRouter ($5 credit to start)
3. **Your existing tools' API keys** — CRM, email, Telegram

**My recommended starter stack:**
- **Hosting:** Hetzner VPS (4 vCPU, 8 GB RAM, €12/month) + Docker
- **LLM:** OpenRouter → Qwen 3.8 Flash for German tasks, DeepSeek V4 for code
- **Monitoring:** Health Monitor workflow (Workflow 4) as your first automation

**Pro tip:** Start with one workflow. Get it stable for 2 weeks. Then add the next. Don't build all five at once — automation debt is real.

[→ n8n Agency services — we build this for you](https://madebybrain.com/n8n-agentur)

[→ Hermes AI Agent setup — for workflows that learn](https://madebybrain.com/en/hermes-agent/install)

[→ Book a free strategy call →](https://madebybrain.com/en/contact)

---

*Steffen Hartmann runs automated business operations with n8n, Hermes Agent, and a multi-model AI infrastructure. He builds AI agents that work — not demo, production.*

---

*Originally published on [madebybrain.com](https://madebybrain.com/en/blog/n8n-agent-workflows-2026).*