Most lists of n8n workflow examples show a screenshot of forty connected nodes, call it "AI-powered lead nurturing" and move on, leaving you to guess what it does, what it costs and whether it would survive a busy Monday. This guide goes the other way: twelve workflows that small teams actually run, each with its trigger, the nodes in the middle, roughly how many executions it uses a month, and the thing that usually breaks it.
Node names match n8n 2.x, and where a ready-made template exists on StackLedge, we link to it.
What is an n8n workflow, in plain terms?
An n8n workflow is a chain of nodes that starts with one trigger: a schedule, a webhook, a form submission, a new email or a new spreadsheet row. Every node after it takes a list of items, does one job and passes them on. Same idea as Zapier or Make, except that you can self-host n8n for free, branching and loops cost nothing extra, and the AI nodes sit on the same canvas as everything else.
An execution is one complete run, however many nodes it passes through. On n8n Cloud that is what you pay for, so watch the last column below.
The 12 workflows at a glance
| # | Workflow | Trigger | Main nodes | Typical runs a month |
|---|---|---|---|---|
| 1 | Daily metrics digest | Schedule Trigger | Postgres or HTTP Request, Code, Slack | 30 |
| 2 | Error alerts for every workflow | Error Trigger | Slack or Gmail | Only when something fails |
| 3 | Lead capture, scoring and routing | n8n Form Trigger or Webhook | Remove Duplicates, HTTP Request, Text Classifier, CRM node, Slack | One per lead |
| 4 | Email triage with drafted replies | Gmail Trigger | Text Classifier, Gmail | One per check that finds new mail |
| 5 | Invoice and receipt extraction | Gmail Trigger or Google Drive Trigger | Extract from File, Information Extractor, Google Sheets | One per document |
| 6 | Competitor price watch | Schedule Trigger | HTTP Request, HTML, Google Sheets, Slack | 30 to 720 |
| 7 | Content repurposing queue | RSS Feed Trigger | HTTP Request, Basic LLM Chain, Notion | One per new post |
| 8 | Support ticket summaries | Webhook | Basic LLM Chain, Switch, Slack | One per ticket |
| 9 | Landing page audit on request | n8n Form Trigger | HTTP Request, HTML, AI Agent, Gmail | One per audit |
| 10 | Chat with your database | Chat Trigger | AI Agent, Postgres | One per message |
| 11 | Personal assistant on Telegram | Telegram Trigger | AI Agent, Gmail, Google Calendar | One per message |
| 12 | Nightly backup of your workflows | Schedule Trigger | n8n, GitHub | 30 |
Which n8n workflows should a beginner build first?
1. Daily metrics digest
A Schedule Trigger fires at 8am. A Postgres node counts yesterday's sign-ups, the Stripe node pulls revenue, and a Code or Aggregate node turns it all into one message for the Slack node. Thirty runs a month, and it teaches credentials, expressions and merging data.
Time zones break it. Your database probably stores UTC while the trigger follows the workflow's time zone, so "yesterday" quietly means two different things. Set the time zone in the workflow settings. And add an IF node that alerts you when a number is zero, because a broken tracker looks exactly like a slow day.
2. Error alerts for every workflow
Build this before anything that matters. A separate workflow starts with an Error Trigger and sends a Slack message or email with the workflow name, the failed node, the error and a link to the execution. Then pick it as the error workflow in every other workflow's settings.
It only fires when a workflow runs on its own, not on a manual test, so check it with the real trigger or a temporary Stop and Error node.
Which workflows bring in or protect revenue?
3. Lead capture, scoring and routing
An n8n Form Trigger starts it, or a Webhook node if your site already has a form. Remove Duplicates, keyed on email and set to drop items seen in previous executions, stops one person becoming two deals. An HTTP Request node enriches the lead from the email domain, a Text Classifier sorts it into Hot, Warm or Cold, and a Switch node sends Hot leads to your CRM node and a Slack ping.
Vague categories break it. "Hot: a good lead" gives the model nothing to work with. Write each one as a rule a new hire could follow ("Hot: 10 or more staff and asked about pricing or a demo") and test it on twenty past leads. The LeadTriage Kit is an importable workflow that already does this scoring.
4. Email triage with drafted replies
A Gmail Trigger checks the inbox every few minutes and only starts a run when there is new mail. A Text Classifier labels each message as billing, support, sales or noise, and for support and sales a chat model writes a reply that the Gmail node saves as a draft in the thread.
Drafts, not sends. An AI that emails a customer the wrong refund policy costs far more than the minute it saved. If you want it to send later, n8n can pause and ask a person first, as covered in how to build an AI agent in n8n. Use the trigger's search filter to skip newsletters and auto-replies, which otherwise eat tokens.
5. Invoice and receipt extraction
Trigger on emails with attachments or on new files in a Google Drive folder. Extract from File pulls the text out of the PDF, an Information Extractor returns the supplier, invoice number, date, net amount, VAT, total and currency, and Google Sheets gets a new row. An IF node flags any invoice where net plus VAT does not equal the total.
Scanned PDFs break it. A photographed receipt has no text layer, so the extractor gets nothing and may invent a supplier rather than admit it. Send empty extractions to a person, or to a vision-capable model.
6. Competitor price watch
A daily Schedule Trigger reads a list of competitor pricing pages from Google Sheets, an HTTP Request node fetches each one, and the HTML node pulls out the price with a CSS selector. If it differs from the stored value, the workflow posts to Slack and updates the row.
Redesigns break it. The selector stops matching, and an empty result looks just like "no change", so treat "price not found" as an alert in its own right. If you would rather not maintain selectors, RivalProbe does the same job as a hosted tool.
Which workflows save the most time on content and support?
7. Content repurposing queue
An RSS Feed Trigger watches your blog, and an HTTP Request node fetches the full article, since many feeds only carry a summary. A Basic LLM Chain turns it into a LinkedIn post, a short thread and a newsletter blurb, saved to Notion as drafts for a person to post.
Left alone, the model opens every post the same way, so give it three of your own posts as examples. Long articles through a large model add up, and AICost compares LLM API prices across providers.
8. Support ticket summaries
Your help desk calls a Webhook node for each new ticket. A Basic LLM Chain with a Structured Output Parser returns a three-line summary, a category and an urgency level, and a Switch node routes it to the right Slack channel, with a louder alert for anything urgent.
Mind the personal data. Tickets carry names, email addresses and sometimes card numbers, all sent to your model provider, so strip obvious details with a Code node or run a local model through the Ollama nodes.
9. Landing page audit on request
An n8n Form Trigger asks for a URL and an email address. An HTTP Request node fetches the page, the HTML node extracts the headings, meta description and button text, an AI Agent writes ten specific suggestions, and Gmail sends them. On your website it doubles as a lead magnet. Analyze Landing Page with OpenAI is a published version of the same idea.
Limit requests per email, and trim the HTML before it reaches the model, because a 2 MB page makes a very expensive prompt.
What can AI agents do inside n8n workflows?
10. Chat with your database
A Chat Trigger opens a chat window, and an AI Agent with a chat model, Postgres Chat Memory and a Postgres tool answers questions like "how many trials converted last week?" Every message is an execution. The Chat with a database using AI template does this for Postgres, MySQL and SQLite.
Give the agent a read-only database user, never your app's main login, and describe your tables and columns in the system message. An agent left to guess column names guesses wrong with total confidence.
11. Personal assistant on Telegram
A Telegram Trigger receives a message, voice notes get transcribed, and an AI Agent with Gmail, Google Calendar and a task list as tools reads your mail, checks your diary or adds a to-do, then replies. Angie is a published template built this way.
Anyone who finds the bot can talk to it, and it can read your inbox. Put an IF node straight after the trigger that checks the sender's Telegram user ID and stops everyone else.
12. Nightly backup of your workflows
At 2am a Schedule Trigger runs the n8n node, using an API key from your instance settings, to fetch every workflow, and the GitHub node writes each one to a private repository as JSON. That gives you version history and a way back.
It matters most if you self-host, where one dead disk can take months of work with it (see self-hosting n8n in production). Keep the repository private: exports include credential names, which people often name after clients.
How many executions will these workflows use?
As of 18 October 2026, n8n Cloud Starter costs €20 a month billed yearly (€24 month to month) for 2,500 executions, and Pro costs €50 a month billed yearly (€60 month to month) for 10,000. The self-hosted Community Edition has no execution cap, and you pay for the server instead.
The schedule usually decides the plan. Every 15 minutes is about 2,880 executions a month, more than Starter allows on its own. Every 5 minutes is 8,640, hourly is 720 and daily is 30. Use the slowest schedule the job can live with, and batch the work: one run checking twenty pages is one execution, twenty workflows are twenty.
The digest, error alerts, lead routing for a few hundred leads, a daily price watch and the nightly backup come to roughly 400 executions a month, well inside Starter. Chat agents are the ones to watch, and model costs are billed separately by your AI provider. Our n8n vs Make vs Zapier comparison covers how each platform charges.
What goes wrong after a workflow goes live?
- Green runs that did nothing. The API returned an empty list and nothing was sent. Check for "no items" where silence is bad news. For scheduled jobs outside n8n, Crontinel flags runs that finish without processing any work.
- Expired logins. Google and Microsoft connections get revoked, and every workflow using them fails at once. Your error workflow is what tells you.
- Rate limits. Turn on Retry On Fail for anything calling an outside API, and use the HTTP Request node's batching option for long loops.
- Fixes that never went live. Since n8n 2.0, saving keeps a draft and only Publish changes what runs in production.
- One giant workflow. Past about thirty nodes, move repeated steps into a sub-workflow.
Where can you find n8n workflow templates?
n8n's own library at n8n.io/workflows holds well over 9,000 community templates of mixed quality, so read every node and test on fake data first. The StackLedge n8n templates category is smaller, and a person reviews each listing before it goes live. If you would rather describe a workflow than build it, n8nChat is a browser extension that builds and edits n8n workflows from plain-English prompts.
Frequently asked questions about n8n workflows
What is an n8n workflow? An automation built from connected nodes. It starts with one trigger, such as a schedule, webhook, form or new email, and each node after it fetches, filters, transforms or sends data.
Is n8n free to use? The self-hosted Community Edition is free with no execution limit, so you only pay for the server. n8n Cloud starts at €20 a month billed yearly for 2,500 executions.
What is a good first n8n workflow for a beginner? A daily digest: a Schedule Trigger, a node or two that fetch numbers, and a Slack or email node to send them. It teaches the basics without much that can go wrong.
Do you need to know how to code to use n8n? No. Most workflows use built-in nodes only. The Code node accepts JavaScript or Python for the odd job a built-in node cannot do.
How do you import an n8n workflow from JSON? Paste the JSON onto an empty canvas, or choose Import from File in the workflow menu, then reconnect credentials on each node.
Built one worth sharing?
If you have an n8n workflow other people could use, submit it to StackLedge. Workflows can be listed from their JSON alone, no website needed, and this guide explains whether yours belongs as a workflow, a node or a tool.