Five n8n Workflow Templates, Free to Download

Five workflows we actually run, as importable JSON. Lead qualification, supplier bills into Xero, quote chasing, expiry warnings and client onboarding. No email, no signup.

Time
10 minutes to first run
Difficulty
Beginner to intermediate
Tools
n8n, plus the apps each workflow touches
Cost
Free, hosting and tokens only

Five n8n workflows we actually run, free to download and import. No email. No signup. No drip sequence waiting for you on the other side of a button.

Most template libraries are full of workflows that demo well and break on contact with a real business. One happy path, no error handling, and a prompt doing work that should have been arithmetic. These five are the opposite. Boring where it counts.

Each one downloads as a JSON file. Import it, open the Configuration node, and the rest is wiring up your own credentials.

1. Qualify and route inbound leads

Qualifying inbound leads is the job everyone automates first and gets wrong in the same way. The usual build hands the whole thing to a prompt, read this enquiry and give it a score out of a hundred, and then the same enquiry scores 71 on Monday and 64 on Thursday with nobody able to say why.

This one splits the work. The model reads the enquiry and returns seven structured facts, and a Code node does the arithmetic. Same enquiry, same tier, every time, with the reasons written onto the record.

The field normaliser handles about thirty-five different names for the same seven fields, so pointing it at someone else's form doesn't break it. That's the most common reason an imported template dies on first run.

12 nodes. Form or Webhook, any AI model, HubSpot, Slack, Gmail.

Download Workflow (.json) Free, no email required.

2. Read supplier bills into Xero, ABN and GST checked

Bills arrive as PDFs and someone types them into Xero. A model can read them, and that is where most builds stop, which is exactly how bad data gets in.

This one checks the numbers before anything is written anywhere. It runs a real ABN checksum, the modulus 89 algorithm the ATO publishes, so a transposed digit fails rather than being trusted because the model sounded confident. Then it cross-checks GST at ten percent, confirms the subtotal and GST add up to the total, and throws out dates that aren't dates.

Bills that pass become drafts in Xero. Bills that fail go to Slack naming the check that failed. Nothing on a maybe.

11 nodes. Gmail, any AI model, Xero, Slack.

Download Workflow (.json) Free, no email required.

3. Chase unanswered quotes until someone replies

Quotes go out and then sit there. This reads a register every morning, works out which ones have gone quiet, and sends a follow-up worded differently each time. Three identical emails read as a robot. Three that acknowledge time passing read as a person.

It stops the moment someone replies. Change the status in the sheet and that quote is never chased again.

9 nodes. Google Sheets, Gmail, Slack.

Download Workflow (.json) Free, no email required.

4. Warn before a licence or insurance lapses

Licences, insurances, white cards, trade tickets, first aid certificates. Every one has a date on it and none of them announce themselves. This warns at sixty days, thirty days and seven, then daily once something has actually lapsed.

The part worth stealing is the suppression. Each item records which band it was last warned about, so sixty days sends once rather than every morning for a month. A warning nobody reads is worse than no warning, because it feels like cover.

9 nodes. Google Sheets, Gmail, Slack.

Download Workflow (.json) Free, no email required.

5. Run new client onboarding from one form

A client signs and five things need doing. A folder, a kickoff booking, a welcome email, a row in the tracker, a message to the team. Usually one person does all five, usually late, usually missing one.

Everything derives in a single node, so the folder naming can't drift. Two people naming folders by hand will produce Acme Pty Ltd and acme-pty-ltd-2026 in the same Drive inside six months, and sorting that out later means touching every link that ever pointed at either of them. That one is expensive to undo.

8 nodes. Google Drive, Google Calendar, Gmail, Google Sheets, Slack.

Download Workflow (.json) Free, no email required.

How to install them

Same five steps whichever workflow you picked. First run takes about ten minutes, most of it spent connecting credentials.

  1. Download the workflow you want

    Pick a workflow above and download its JSON file. Nothing is gated, so no email is required.

  2. Import it into n8n

    In n8n, open the workflows list, use Import from File, and choose the JSON you downloaded. The whole workflow appears on the canvas with its documentation attached as sticky notes.

  3. Connect your own credentials

    Every node that talks to an outside service shows a credential warning. Open each one and pick your own connection. None of our credentials ship inside the files.

  4. Edit the Configuration node

    Each workflow has a single Set node named Configuration holding the things you are meant to change: thresholds, channels, schedules, wording. Everything downstream reads from it.

  5. Run it once by hand before switching it on

    Use Execute workflow and watch what each node returns on real data. Confirm the branches behave before you activate anything on a schedule or a live webhook.

Some of these won't suit you

If you send four quotes a month, the chaser solves a problem you don't have, and a calendar reminder does the same job for nothing. The bill reader earns its keep somewhere north of thirty invoices a month. Under that, typing them is faster than explaining the exceptions. We'd rather give a straight answer than have you import five workflows and run two.

They're built on n8n rather than a per-task tool, which starts to matter the month a workflow fires often enough to notice. A process running fifty times a month and one running fifty thousand cost the same to operate. If you're weighing that up, we've compared n8n against Zapier and against Make in more detail.

When it doesn't work

Every node shows a red credential warning after import
That is correct behaviour. Credentials are deliberately stripped from the files, so open each flagged node and pick your own connection.
A Google Sheets node returns nothing, or writes to the wrong row
The column headers have to match the names listed in the sticky note exactly, including underscores. A header of "Client Email" will not match client_email.
Dates come back a month out in the bill reader
Australian invoices are DD/MM/YYYY and some models read them as US dates. The extraction prompt says so explicitly, so if it still happens, try a stronger model before editing the prompt.
The expiry monitor sends the same warning every morning
The last_notified_band column is not being written. Confirm the Record the Warning node is matching on item_id and that every row has a unique one.
The quote chaser emails someone who already replied
Its stop condition is the status column, not the inbox. A reply has to move the status out of sent or following_up, either by hand or by a rule that does it for you.
An AI node errors with a schema or parsing failure
Small models struggle to return structured output reliably. These were written against mid-sized models and above; if you are pointing at a small local model, expect to simplify the extracted fields.

n8n template FAQs

Yes, and there is nothing to sign up for. The files are plain JSON on this page, no email gate and no trial. We build automation for a living, so a workflow you can run yourself is an honest sample of how we work rather than a lead magnet with the useful parts removed.

No, though it helps for the Code nodes. Every workflow is designed so the only node you need to edit is Configuration, which is a plain list of values. The JavaScript inside the Code nodes is commented line by line if you want to change the logic, and safe to leave alone if you do not.

Both. Every node used is a standard n8n node with no community packages, so nothing here depends on a self-hosted instance. Self-hosting only matters if your data cannot sit on someone else’s cloud, which is a compliance question rather than a technical one.

Hosting plus model tokens. n8n Cloud starts around $30 AUD a month, and the AI calls in these workflows are small, usually a fraction of a cent each. A business processing 200 bills a month would spend a few dollars on tokens. There are no per-task fees, so volume does not change the shape of the bill.

Yes to both, including inside client work. They are meant to be pulled apart. If one gets you most of the way and the last stretch is awkward, that last stretch is usually the interesting part, and we are happy to look at it.

Check the troubleshooting list above first, because most first-run failures are a credential that was not set or a Google Sheet whose column headers do not match. If it is something else, tell us what the failing node returned and we will take a look.

Got one of these running and hit the awkward ten percent? Tell me which node is fighting you and I'll give you a straight answer on whether it's worth building properly.

Book a Free Call