Automating Tasks: Find What to Automate First

Most advice on automating tasks starts with a list of tools. I start with a smaller question. Where do you type the same thing twice? That one spot is usually your first win.

It's 4:50 on a Friday. A customer filled in your website form at lunch. You read it on your phone, meant to reply, and then a supplier called. Now you're typing their name into a spreadsheet, typing it again into your invoicing tool, and drafting an email that says sorry for the wait.

You don't need a list of the ten best tools to fix that. You need to see the pattern, because the pattern tells you what to automate.

I run a small stack of my own products on n8n, Supabase, Next.js and Docker on a VPS, plus the Anthropic API. Below is the method I use to decide where automating a task pays off and where it just creates a new chore.

Stop making lists. Follow the copy-paste.

The usual advice says to write down everything you do in a week and rank it. That works. It's also slow, and most people quit by Wednesday.

A faster test is to watch your hands. Notice when they do one of these things:

Copy text from one window and paste it into another.
Retype a number you just read somewhere else.
Forward a message with a one-line note on top.
Open the same three tabs in the same order.

Each one is a handoff between two systems, with you standing in the middle as the cable. Automation is good at replacing cables. It is bad at replacing judgment. That makes the copy-paste moments the cleanest candidates you have.

Why this beats a time audit

A time audit finds big, vague blocks like "admin." A copy-paste hunt finds small steps you can actually build. Nobody can build "admin." You can build "when a form is submitted, add a row to the sheet and create a draft invoice."

Most small businesses that ask me about this have the same three problems. Leads don't get followed up fast enough. Data gets typed twice into two systems. Calls go unanswered. Two of those are literally copy-paste problems. The third is a trigger problem.

A worked example: the landscaping quote

Say you run a landscaping crew. Quote requests arrive three ways: a web form, a text, and a voicemail. Here's what happens today, with the decision at each step.

1. You read the request. This needs your eyes the first time. Leave it alone. 2. You type it into your job spreadsheet. Copy-paste. Automate it. 3. You email the customer to confirm you got it. Same words every time. Automate it. 4. You price the job. This is judgment. A site visit, a soil problem, a tricky slope. Keep it manual. 5. You create the invoice after the job. Copy-paste from the quote. Automate it, but only after you approve the quote.

Three of the five steps are mechanical and two are yours. That ratio is typical. You don't automate the job. You automate the cable between its parts.

Look at what I did at step 4. I wrote down the rule for where the machine stops and you start. That boundary matters more than any tool you pick.

The decision rule I use

For each copy-paste moment, I ask two questions.

Is the input always the same shape? A form with fixed fields qualifies. A rambling email does not, at least not yet.
Is a wrong result cheap to catch? A wrong row in a spreadsheet gets noticed. A wrong email to a customer gets noticed too late.

Two yeses means build it now. One yes means build it with a human approval step. Zero means leave it manual and look again in six months.

Start with the boring one that has a clock on it

When I'm choosing between several copy-paste candidates, I pick the one where a delay costs money. In my experience the first automation that pays for itself is usually the boring one. Lead follow-up within minutes, not a chatbot.

The reason is timing. A research article in Harvard Business Review on how quickly online sales leads go cold found that a fast response made a large difference to whether a company reached the lead at all. It's an older study, so treat it as a direction and not a promise. The logic still holds. A lead you answer in minutes is warmer than one you answer tomorrow.

The automation itself is small. A form submission fires a trigger. The trigger sends a short, honest reply saying you got the message and when a human will follow up. It also drops the lead into your tracking sheet so it can't get lost.

What not to automate first

Skip anything that speaks for you in a long, open-ended way. Skip anything that moves money until you've watched a simpler flow run clean for a few weeks. And skip anything where you can't say in one sentence what "done" looks like.

Pick the tool last, and know what you're signing up for

Once you know the steps, the tool choice gets easy. A hosted tool like Zapier or Make suits you if you want to be running by tonight and never think about servers. A self-hosted tool like n8n suits you if you want more control and lower running costs.

I live with a trade-off. Self-hosting n8n on a VPS is cheap and reliable, but somebody has to own the updates, backups and secrets. On a managed tool, that somebody is the vendor. On mine, it's me. If nobody in your business wants that job, pay for the managed version and don't feel bad about it.

If you do self-host, the Docker documentation covers the container basics. Write down where your backups go before you build your second workflow, not after.

Check it like you'd check a new hire

A new automation deserves a probation period. For the first two weeks, look at every run. Read the output. Check the spreadsheet row against the form. Check the email that went out.

After that, set one alert for the failure that would hurt most. For lead follow-up, that's "a form came in and no reply went out." A silent failure is worse than a loud one, because you keep trusting something that stopped working.

The U.S. Small Business Administration keeps general guidance on running a business day to day. It won't tell you which workflow to build. It is a fair place to sanity-check the parts of your operation that have nothing to do with software.

Your next ten minutes

Open your inbox and your spreadsheet side by side. Find the last time you copied something from one to the other. Write down the trigger, the input, and what "done" looks like. Congratulations, you've just written the spec for your first automation.

If you'd like a hand turning it into something that runs, get in touch. Tell me what you keep retyping and I'll tell you honestly whether it's worth automating.

---

*Written by Mario Lucas. Mario runs Utomat, an automation studio that builds AI voice agents and workflow automation for small businesses.*

Automating Tasks: Follow the Copy-Paste to Find What to Automate First