n8n Intermediate: Real Business Workflows

A Google Sheet can be the database behind an automation. n8n reads rows (Get row(s)), adds rows (Append row) and changes rows it finds by a key column (Update row). Each row comes into n8n as one item. To touch your sheet at all, n8n needs a credential: a saved, encrypted login you create once and pick from a list in every node after that.

Most small businesses already keep their clients, jobs and requests in a sheet. Using that sheet, instead of moving to new software, means the automation fits how the team already works, and anyone can still open the sheet and read it.

A small office keeps a request list in Google Sheets. n8n reads it each morning, appends new requests from a form only if the email isn't there yet, and updates each row with a short AI-written description, all through one Google Sheets credential.

A Schedule Trigger starts a workflow at a set time, like every weekday at 5pm, in the workflow's time zone. A Summarize node then turns many items into one, such as the total of a revenue column, so the report is one email, not fifty.

Recurring reports are the easiest hours to win back. Nobody has to remember to run them, and they arrive the same way every day.

At 5pm n8n pulls today's jobs from a sheet, totals the revenue with Summarize and emails the owner one line: today's jobs and today's total. Before, someone added it up by hand every evening.

An API is a door an app opens for other software. An HTTP request is a message sent to that door: a method (GET to read, POST to create), a URL, sometimes query parameters to narrow it down and a body with the data you send. The app answers with JSON, and n8n turns each record into an item.

n8n has nodes for hundreds of apps, but your business probably uses one it doesn't cover. The HTTP Request node connects anything with an API, so no app has to stay an island that someone copies data out of.

A marketing agency pulls its client list from its CRM's API, asks for one client with ?username=Samantha, then POSTs a follow-up note back to the CRM with the client's email filled in by an expression.

A webhook flips an API around. Instead of n8n asking an app for news every few minutes, the app sends the data to a web address the moment something happens: a form is sent, an order is paid, a call ends.

Speed. A new lead or a support request reaches your workflow in a second, not on the next scheduled check, and nothing runs when nothing happens.

A website contact form is pointed at a webhook address. A visitor sends a message, and the name, email and message arrive as JSON straight away, ready for the next step.

There are two quick ways to get form answers into n8n. On form submission gives you a form that n8n hosts for you, with the questions you add. A Webhook node gives you an address that your own website form, or any other app, can send to. The Webhook has two addresses: the Test URL only listens while you click Listen for test event in the editor, and the Production URL listens all the time once the workflow is published.

Every form the business runs, quotes, bookings, new clients, can land in one sheet without anyone copying it over. And mixing up the two URLs is the number one reason a form works in testing and then goes silent, so knowing which one goes in the live form saves a lost lead.

A business collects quote requests from an n8n form and from its website form. Both feed the same Google Sheet with Append row in sheet, mapping Name and Email by dragging them from the INPUT panel. The website form is pointed at the Production URL once the workflow is published.

An AI step can read a photo or a PDF and pull out exactly the fields you ask for, like the date, vendor, amount and category, as structured data that the next node can save.

Typing receipts, invoices and forms into a sheet is slow and full of typos. AI extraction does the reading; you check the result instead of doing the typing.

Someone drops a receipt photo into an n8n chat. An AI agent reads it, picks a category and adds one row to the expenses sheet.

A Gmail trigger starts the workflow when a new email arrives. An AI step writes a reply and saves it as a draft in the same thread, instead of sending it. A person reads it and clicks send.

Customers get a faster answer, and nothing goes out that a person hasn't seen. That human check is what makes AI safe to use on customer email.

A customer asks about prices. Within a minute a draft reply is waiting in the thread, using the business's own wording. The owner fixes one line and sends it.

When a node gets ten items, it runs ten times, once per item. An expression such as {{ $json.client }} fills in the current item's value, so the same node writes a different email for every row. You make one by dragging a field from the INPUT panel onto a parameter.

This is how one small workflow handles a whole client list. You build it once for one row, and it already works for a hundred.

Every Monday n8n reads the client tracker, and for each client AI writes a short status update from that row's project and next step, saved as a Gmail draft addressed to that client.

An AI Agent node gets a goal, a chat model and tools, such as a Google Sheets tool. It decides which tool to use and when, instead of following fixed steps. Memory lets it remember the conversation.

Some jobs need a judgment call in the middle: look something up, then decide what to save. An agent handles that step, while the rest of the workflow stays fixed and predictable.

An agent suggests social post ideas, and when you approve one it writes it to a Google Sheet through its Sheets tool.

Things fail: an API is down, a sheet was renamed, a rate limit was hit. An error workflow (Error Trigger) emails you when any workflow fails. Retry On Fail makes one flaky step try again by itself. On Error: Continue (using error output) sends failed items down their own branch, so the good ones still finish. Executions lets you retry a failed run.

A silent failure is worse than no automation: the price sheet goes stale and nobody knows. Error handling is what lets you stop checking on your workflows.

A nightly price update fails because the supplier's API timed out. The step retries twice on its own, and when it still fails, the owner gets an email with the workflow name and the error, then re-runs it from Executions in the morning.