// the tuesday guide

AI automation without coding: where to start

2026 · 09 · 15 — THE TUESDAY GUIDEAI automation without coding

“I'd like to automate something, but I can't code.” It's the sentence I hear most often from freelancers and small business owners, and it usually comes after a day spent copying data from an email into a spreadsheet, answering the same question for the tenth time, hunting for an invoice among emails from three months ago. The good news is that in 2026 you no longer need code. The less comfortable news is that you need to know what to automate, and no tool decides that for you.

This guide does exactly that: a test for choosing, three automations that pay for themselves in the first week, the tools with their real costs, and the rule for not getting hurt. All without writing a single line of code.

1 · The “always the same” test

The most common mistake is starting from the tool: “I saw a video about Make, what can I do with it?”. You start from the opposite end. Take a sheet of paper and write down the jobs you redo every week, the same way, without thinking. Not “managing customers” — too big — but small, precise things: “I copy the name and phone number from the email into the spreadsheet,” “I reply to people asking whether we deliver,” “I save the courier's receipt in the expenses folder.”

Then apply the test: if you have to make a decision every time you do it, leave it alone. If you could do it with your eyes closed, it's the right candidate. A custom quote requires judgment and can't be automated; the email that says “thanks, I'll get back to you by tomorrow with the quote” can. After ten minutes you'll have a list of three or four items. That's enough: you add the others once the first ones work.

2 · Automation 1: email that sorts itself

The first automation I recommend to everyone, because it affects everyone. It works like this: every incoming email is read by an AI model, which gives it a label — quote, invoice, support, advertising — and summarizes it in one line. You decide the labels, using the words you already use. The model receives only the text of the email and replies with the category: for most inboxes it's a job that costs a few dozen cents a month.

The extra step, the one that really pays off, is for quote requests: the AI also prepares a draft reply in your tone, saved in your drafts, with the three questions you always ask to understand what the customer needs. You open the draft, correct it if necessary, and send it. It never sends on its own. Response time goes from “when I get to it” to ten minutes, and in the morning the inbox is no longer a wall: it's a list already sorted.

3 · Automation 2: from the website form to a reply in two minutes

The website's “Contact us” form and WhatsApp messages share the same problem: they arrive when you're not there. In the evening, on Saturday, while you're at a customer's. And meanwhile the person who wrote has also written to two others. The automation is simple: when a request arrives, the data goes into a row of a spreadsheet (name, contact, what they're asking, date) and the customer immediately receives a reply that says what happens next: “Thanks, I've received your request. I'll get back to you by noon tomorrow with a quote; in the meantime, if you have photos or measurements, send them here.”

It's not a chatbot pretending to be you. It's an honest confirmation with a precise time, and at the end of the chain the AI can add a useful line read from the request: if the person is asking for a repair, the reply reminds them to attach a photo; if they're asking for a quote, it reminds them what information is needed. Anyone who wants to go further — answering frequently asked questions directly in chat, at night — can connect a WhatsApp or Telegram bot to their price list, but that's the next step, not the first.

4 · Automation 3: incoming invoices that file themselves

The third one is the one that makes your accountant happy. Every invoice or receipt that arrives by email — the supplier, the courier, the subscription, the phone — is saved in the right folder of your Drive or your archive, renamed in a readable way (“2026-09-14 Supplier Rossi 340.00”) and recorded in a row of the expenses spreadsheet: date, supplier, amount, category. This is exactly where AI comes in: it reads the PDF, pulls out the four pieces of data and writes them into the spreadsheet, something that two years ago required software costing thousands of euros.

Two precautions. First: the automation doesn't pay anything and doesn't touch the bank; it only files and records; decisions about money stay yours. Second: in Italy, electronic invoices go through the SDI exchange system and you already have them in your tax portal, so this automation is for everything else — foreign receipts, online subscriptions, photographed paper receipts — which today gets lost or reconstructed at year's end.

5 · The tools and what they really cost

The tools are called Make, Zapier and n8n, and they all work the same way: on screen you connect blocks — “when an email arrives,” “ask the AI to classify it,” “write a row in the spreadsheet” — and the chain runs on its own. No code is written: you pick apps from a list, fill in the fields, test with a real email. Each has a free plan that's enough to get started and experiment; for serious daily use you're looking at between zero and €20 a month, plus the pay-as-you-go cost of the AI, which for these three automations stays in the range of a few euros. Your time, the first time around, is the biggest expense: budget an afternoon for the first one and an hour for each of the others.

Which one to choose? Zapier is the simplest and the most expensive as you grow; Make has a clearer visual logic and costs less; n8n is the most powerful and can even be installed on your own server, but asks for a bit more patience. For anyone starting from scratch: Make. If one day the automations become many and have to talk to your management system, that's the moment it pays to have them custom-built, with AI automation for your business designed around your process rather than a list of apps.

6 · The golden rule: watch it for a week before you trust it

Any automation, as soon as it's switched on, should be kept in “prepare and wait” mode for at least a week: the AI labels, summarizes, fills in the draft — and you approve. Only when you've gone seven days without correcting anything do you let it run on its own, and even then with a copy in a “to check” folder that you open now and then. A faulty automation running on its own doesn't make one mistake: it makes two hundred, all identical, before anyone notices.

And there are things you never automate: replies to an angry customer, quotes that involve judgment, anything touching payments or health data. Also keep in mind what you pass to the AI: the text of a sales email is fine, the customer's entire record is not. If you have doubts about what can leave your office, we covered it in our guide on business data and AI.

In short

You don't need to know how to code, you need a list. Write down the jobs you redo every week the same way without thinking: those get automated, the ones that require a decision every time don't. Start with email that sorts itself with a label and a ready draft, then the request from your website or WhatsApp that lands in a spreadsheet and gets a reply in two minutes, then incoming invoices that file and record themselves without you opening a PDF. The tools are Make, Zapier or n8n, from zero to €20 a month, an afternoon for the first one. And for the first week the automation prepares and you approve: only after that do you let it run on its own.

Frequently asked questions