One dark heavy stone set apart from a row of pale stones, representing picking the one task with the biggest payoff to automate first.

Guide

What should I automate first?

You don't automate everything — you automate the one thing with the biggest payoff and the lowest risk. This guide walks you through picking that one, in four steps, without buying a thing.

List your repetitive tasks 30 minutes

Write down everything you or your team do each week that's manual and predictable. You're looking for tasks that repeat and that you could describe in a sentence.

Common ones for a small business:

Answering the same customer questions
Drafting quotes, proposals, or invoices
Chasing late payments
Writing social posts or email replies
Transferring data between systems
Booking and confirming appointments

How to do this

Don't try to be exhaustive — five to ten items is enough. The point is to capture the tasks that genuinely eat time, not the ones you've heard are automatable.

Done when: you have a written list of your repetitive tasks.

Score them 20 minutes

Give each task two scores out of ten: how many hours it costs you a week, and how much you dread it.

The second score matters more than it sounds. A task you hate is one you'll actually hand over to a tool — and a tool you use beats a perfect tool you avoid.

How to do this

For each task, write two numbers:

Hours per week. Be honest and conservative. If a junior staffer does it, that's real time too.
Dread (1–10). How much you'd rather not do it at all.

Then add the two numbers. That's the task's score. The higher, the better a candidate it is.

Done when: every task on your list has a combined score.

Pick the winner 10 minutes

Choose the task with the highest score and the clearest steps. A task that's easy to describe is easy to automate — the opposite also holds.

Skip anything with lots of judgment calls, or that involves high-stakes mistakes (legal, finance, sensitive customer matters). Those aren't first candidates.

Why this matters

The single biggest predictor of AI success is starting narrow. One task, one clear metric, one month to prove it. Automating your three least-favorite tasks at once means you can't tell what's working, and one failure poisons the whole effort.

Done when: you've picked one task to start with.

Test it 30 minutes

Set up a free tool for that one task and run it on a real example. You're proving it can do the job before you commit any time or money.

How to do this

Match your task to the right kind of tool:

Your task	Start with
Writing / drafting	A general AI assistant, free tier
Customer questions	A simple chatbot
Invoices / bookkeeping	AI features already in your software
Data between systems	An automation tool, free plan

Feed it a real example, run one real output, and see if the result is usable. If it is — great. If not, either the task or the tool wasn't the right fit.

Done when: you know whether this task works with AI.

Pick the winner, prove it, then expand. One task, done well, beats three started.

The 30-second takeaway

List your repetitive tasks, score them by hours and dread, pick the one with the highest score and clearest steps, then test it with a free tool. Start narrow, win fast, and expand only once it works.