Six n8n automations worth building first

Everyone agrees that automation saves time. Far fewer people can name the exact workflow they would build on Monday morning, which is why most automation projects stall at the enthusiasm stage.

So here are six we build repeatedly, each with its trigger, its steps and the manual work it removes. Pick the one that happens most often in your week. Not the one that annoys you most: the one with the highest repetition count, because that is where the hours are.

1. Invoice to appointment, end to end

Trigger: an intake or booking form submission hits a webhook.

Steps: create the client record; generate a PDF invoice from a template; email it from Gmail; watch the inbox for the payment confirmation; on payment, create the calendar appointment with a meeting link and send it; sync the record to Notion and Google Sheets; alert a human on any failed branch.

What it replaces: the chain of copy-paste that follows every new client. Count the minutes you spend per client on invoice, inbox-watching, booking and updating two spreadsheets, then multiply by new clients per month. For a clinic running several a day, that is most of an admin role.

The slowest step is usually the one nobody thinks of as work: checking whether the money arrived. Automating the payment detection is what unlocks the rest of the chain, because every later step depends on it.

2. Proposal generation from a short form

Trigger: a sales rep submits a five-field form after the call.

Steps: an AI step maps the notes onto your proposal structure and writes the personalised sections; a copy of the master deck is created in Google Drive; the placeholders are replaced in Google Slides; Gmail sends it from the rep’s own account so the thread stays with them; the inputs are logged to a sheet.

What it replaces: an hour or more of deck building per proposal, and the two days of cooling-off while the client waits. If your team sends five proposals a week, that is most of a working day back, every week.

Two details make it work. The trigger is a form the rep already fills in, so nobody has to adopt a new habit. And the AI writes into a fixed template rather than generating a document from scratch, which keeps everything on-brand.

3. Lead scraping and enrichment

Trigger: manual run with parameters (business type, location) or a weekly schedule.

Steps: search a source such as Google Maps for matching listings; loop over the results; visit each website; extract emails and any other contact detail; de-duplicate against your CRM; filter out the irrelevant entries; write a clean list to Google Sheets ready for outreach.

What it replaces: days of a junior researcher tabbing between search results and contact pages. A list that took two days to assemble by hand lands in minutes.

Be honest about the limits. Scraped lists are raw material, not a campaign, and they go stale. Respect the terms of the sources you pull from and your market’s marketing rules before you send anything.

4. Competitor price monitoring

Trigger: a daily schedule, or on demand before a pricing meeting.

Steps: read your product and competitor URL list from a sheet; fetch each page; convert the HTML to Markdown; ask an AI step to extract the current price; write it back to the sheet with the date so you build a history.

What it replaces: whoever currently opens 40 tabs every Monday, or more likely the fact that nobody does and your pricing decisions are based on a snapshot from last quarter.

The AI extraction step is what makes this maintainable. Hard-coded CSS selectors break every time a competitor redesigns; asking a model to find the price in the page text survives most layout changes.

5. Review requests and drafted replies

Trigger: a job marked complete in the CRM (for the request), and a scheduled check of your Google Business Profile (for the replies).

Steps: send the review request by WhatsApp or SMS a set number of hours after the job; separately, fetch new reviews; store them with their status in Airtable; have an AI agent draft an on-brand reply following your written guidelines; send the draft to a manager who can publish, edit then publish, or skip; post the approved reply and update the record.

What it replaces: review chasing that never happens consistently, and reply writing that takes minutes per review and sounds different depending on who did it.

Note the human in the loop. Public replies are one of the few places where we always keep approval, because the cost of one bad automated reply is higher than the time it saves. If local visibility is the goal behind this, the automation only matters alongside the rest of the local SEO fundamentals.

6. Reporting digests

Trigger: a schedule, usually Monday 7am and the first of the month.

Steps: pull the numbers from ad platforms, analytics and the CRM; normalise them into one row per channel; compare against the previous period; have an AI step write three sentences on what changed and what looks wrong; post to Slack or email, with a link to the full dashboard.

What it replaces: the half-day someone spends assembling a report nobody reads, and the fortnight it takes to notice that a campaign broke.

Keep the digest short. Five numbers and the exceptions. The purpose is to get people to look, not to be comprehensive.

When n8n is the right tool, and when it is not

Situation	Use	Why
Everything happens inside one platform	Native workflows (GoHighLevel, HubSpot, Shopify Flow)	No credentials to manage, your team can edit it, nothing to host
Two tools, one trigger, no branching	Zapier or Make	Faster to build, easier for a non-technical owner
Three or more systems, branching logic, loops	n8n	Real control flow, code steps where you need them, no per-task cost cliff
An AI step in the middle of the process	n8n	Prompt, model and output handling all live in the workflow
High volume or sensitive data	Self-hosted n8n	Predictable cost, data stays on your infrastructure
A process nobody has written down yet	Neither	Document it first, then automate it

That last row is the one we enforce. Automating an undocumented process just makes the confusion faster. If a process lives only in one person’s head, write it out as a numbered list first, and expect to find two or three steps that exist for no reason any more.

Native workflows are genuinely better inside their own walls. If your follow-up, reminders and pipeline all live in one CRM, build there and keep n8n for the things it cannot reach, which is roughly the dividing line we draw in GoHighLevel against HubSpot too.

The maintenance cost nobody mentions

Workflows are not furniture. They are small pieces of software, and they rot:

• Credentials expire. OAuth tokens, API keys, app passwords. Put renewal dates in a calendar.
• APIs change. A field gets renamed and your mapping silently writes nulls.
• Scrapers break. Any workflow that reads a page you do not control will need attention a few times a year.
• Volume changes. Something built for 20 records a day behaves differently at 2,000, especially inside loops.
• Models change. Prompts that worked on one model version need re-testing on the next.

So three things are part of the build, not extras. Every branch gets error handling that alerts a person instead of failing quietly. Every run gets logged somewhere you can read it, with the input that caused the failure. And every workflow gets an owner by name. Across a stack of six to ten workflows, expect a few hours a month of upkeep. That is cheap against the hours saved, but it is not zero, and anyone who tells you it is has not maintained one for a year.

What this looks like when we do it

The clinic case is the clearest. Staff were touching four tools per patient: copy the intake form, create an invoice, watch the inbox for payment, book the appointment, update two spreadsheets. We built the whole chain in n8n with error handling and human approval where it mattered, and admin time fell by 90% with zero manual data entry: the write-up is the clinic invoice-to-appointment automation. On the sales side, our n8n AI proposal generator turns post-call form notes into a personalised Slides proposal sent within seconds of the call ending, and the Google Maps lead scraper builds filtered, de-duplicated prospect lists for any niche and city.

Where to start

Take your week and find the task you do more than ten times. Write down its trigger, its steps, and what should happen when a step fails. If it stays inside one platform, build it there this afternoon. If it crosses three, and especially if a human is currently reading something and deciding what to type next, that is an n8n automation worth building, and the ones with a judgement step in the middle are where AI agents earn their place.

If you would rather have someone map the process with you and tell you which two workflows pay back fastest, that is the free audit: 30 minutes, and you keep the action plan whether you hire us or not.