How to find what to automate first

Most automation projects go wrong at the very first step: someone picks a tool and then looks for a place to use it. A better starting point is a plain question. Where does your team repeat the same action several times a day, and where does that repetition cause mistakes, delays or irritation?

The answer is usually not glamorous. It is rarely a futuristic idea. It is a person copying order details from a message into a sheet, a manager chasing three people for the same update, or an accountant retyping figures from a bill. These small, boring jobs are exactly where automation pays off, because they are frequent, they follow rules, and nobody enjoys them.

Ask each person on the team to keep a short note for a week. Every time they do something they have done before, they write one line: what it was, roughly how long it took, and what they started from. You are not measuring productivity. You are building an honest inventory.

Do not rely on a meeting to produce this list. People forget the small things in a meeting, and the small things are the point. A shared note, or even a notebook at each desk, gives a more truthful picture than a workshop does.

Once you have the list, score each item simply. You do not need a spreadsheet model. Four questions are enough, and each can be answered high, medium or low.

The best first project is frequent, rule-based, low-risk and has data already available. Sending an acknowledgement when an enquiry arrives is a good example. Copying a form submission into a tracking list is another. Both are boring, repeatable and easy to check.

Avoid choosing the biggest pain first. The biggest pain is often the most tangled, with exceptions nobody has written down. It is a fine second or third project, once you have learned how automation behaves in your business and your team trusts it.

Automation does not have to mean removing people. For many tasks, the sensible design is that software prepares the work and a person approves it. A drafted supplier order that a manager confirms is safer and easier to trust than one sent silently.

Write down, for each task, what the system may do alone, what it should suggest, and what it should never touch. Money movements, customer-facing messages in new situations and anything legally sensitive usually belong in the 'a person approves' group.

Automation has running costs that are easy to forget. Hosting, software subscriptions, message fees from providers such as WhatsApp or SMS vendors, payment gateway charges and AI usage charges are all paid to third parties and continue after the build. Someone also needs to maintain the system when a process or a connected tool changes.

Set these against the time the task takes today. If a task takes ten minutes a week, almost no automation will justify itself. If it takes an hour a day across the team, the numbers are worth working through carefully. We do not promise a specific saving, because it depends on your volumes and how your team actually works.

After the first automation is live, compare it with your starting notes. Is the task actually faster? Are there new errors? Do people trust it, or are they quietly redoing the work by hand? Honest answers here shape what you build next.

Then repeat the scoring with the remaining items. Each cycle teaches you something about your own process, and often the next candidate looks different once the first one is out of the way. If you want a structured version of this exercise, our automation checklist tool walks through the same questions.