I Analyzed 2,046 n8n Workflows. Here's What People Actually Automate.

Talha Athar — Fri, 02 Oct 2026 15:03:50 +0000

I got tired of starting every n8n automation from a blank canvas, so I did something slightly unhinged: I indexed 38,000 n8n workflows — every one I could find — organized by department and integration, with copy-paste JSON import.

Then I did what any data-brained builder would do: I analyzed a 2,046-workflow sample to find out what people actually automate. Not what vendors say. What the workflows say.

Here are 7 findings.

1. Automations are smaller than you think

The median workflow has 11 nodes. The average is 14.4. 76% are under 20 nodes.

We imagine automation as these sprawling Rube Goldberg machines. In practice, most production automations are: trigger → fetch → transform → notify. Four to eleven nodes, doing one job well.

Takeaway: if your workflow has 40+ nodes, you probably have 3 workflows wearing a trench coat. Split them — they're easier to debug, and a failure in one doesn't kill the others.

2. The #1 "integration" isn't an integration at all

The most-used node across all 2,046 workflows? HTTP Request — by a landslide (2,123 instances).

Not Slack. Not OpenAI. The humble HTTP node. Because the real world runs on APIs that don't have dedicated n8n nodes yet, and builders just wire them up directly.

Takeaway: learn the HTTP Request node properly — auth headers, pagination, error handling — and you can automate literally anything with an API. It's the highest-leverage node in n8n.

3. Google Sheets is the database of automation

After HTTP Request, the top services: Google Sheets (597), Telegram (390), Google Drive (290), Gmail (259), Airtable (255), Slack (198), Notion (163).

Read that again: a spreadsheet is used 2.3x more than Airtable and 3.6x more than Notion as automation storage. The "spreadsheet as backend" pattern isn't a beginner crutch — it's the dominant production architecture for small-to-mid automations.

Takeaway: don't feel guilty about Sheets-as-database. Just follow two rules: upsert on a unique key (never append blindly), and add a dedupe check before any send action.

4. Telegram beats Slack 2 to 1

Telegram appears in 390 workflows vs Slack's 198. Nearly double.

This is the indie-hacker and global-builder effect: Telegram's Bot API is free, simple, and needs no workspace admin approval. For notification and command-bot patterns, it's the path of least resistance — especially outside the US enterprise bubble.

Takeaway: if you're building notification workflows, prototype on Telegram first. You can always add Slack later.

5. The AI agent wave shows up in the triggers

181 workflows now start with a chat trigger — the signature of the AI-agent era (up from essentially zero two years ago). Meanwhile the classic triggers still dominate: schedule (311) and webhooks/forms.

But here's the telling stat: 771 workflows start with a manual trigger. That's builders testing before they wire up the real trigger — a healthy sign that people prototype before they automate.

Takeaway: the stack is shifting from "scheduled scripts" to "conversational agents," but the discipline hasn't changed: trigger → logic → action, tested manually first.

6. The long tail is real

Node counts range from 1 to 246. That 246-node monster exists — but it's the exception. The distribution is heavily skewed small, which means: most automation value comes from composing small workflows (via Execute Workflow — 109 instances) rather than building big ones.

Takeaway: build a library of small, single-purpose workflows and chain them. It's the microservices lesson, applied to automation.

7. What I'd build next (the gaps)

The data shows what's missing: very few workflows handle the unsexy production concerns — error workflows (only 18 error triggers in 2,046!), credential rotation, or execution monitoring. Everyone builds the happy path.

Takeaway: the highest-leverage template nobody's sharing is a production hardening pack: error trigger → Telegram alert → execution log to Sheets. Add it to every workflow on day one.

The index

I built all of this into a searchable vault: 38,000 n8n workflows organized by 8 business departments, filterable by 100+ integrations. Search → preview → copy JSON → Ctrl+V onto your canvas. Everything runs on your instance with your keys.

Free 30-minute full-access demo: https://n8n-enterprise-suite.pages.dev/

Methodology note: stats computed from a 2,046-workflow sample of the index (valid JSON, deduplicated). Counts are node instances across the sample.

DEV Community: Talha Athar