The boring task almost always beats the flashy one
When automation comes up, most people picture something big and visible — a chatbot on the website, a sales forecast, "an AI that advises us." But the best first project is usually the opposite. It's the task someone in the company does by hand every day, that nobody enjoys and nobody brags about.
Retyping data from an email into a spreadsheet. Issuing the same invoice every month. Copying orders from one system into another. These tasks are boring precisely because they're predictable — and that's exactly what makes them ideal candidates. Computers love rules and repetition. Flashy projects sound great in a meeting, but boring projects give you time back as early as next week.
Three questions that reveal a good candidate
Before you automate anything, ask three simple questions about the task. First: does it happen often? A once-a-year task isn't worth solving, even if it takes an hour. A task you do ten times a day is worth solving even if it takes two minutes.
Second: does it have clear rules? If you can describe the task in sentences like "when this arrives, do that," a machine can handle it. If the decision depends on feel, context, and exceptions you can't even list yourself, automation will struggle.
Third: does a mistake hurt? If a typo in an account number means a misdirected payment or an angry customer, automation here doesn't just save time — it saves nerves and money. A task that is frequent, rule-based, and costly to get wrong is almost always the right pick.
A simple map of the week
Before making any decision, map your week for five days. It needs no tool — a notepad or a shared document is enough. People on the team jot down what they do repeatedly and roughly how much time it costs them.
You don't need minute-level precision. A rough estimate is fine: "reports for the accountant — about 40 minutes on Monday," "answering the same customer questions — half an hour a day." After a week you'll see patterns you could never pull from memory.
This map usually surprises people. A task you thought was marginal eats three hours a week. And the reverse — the thing that worried you in the meeting happens once a month and doesn't need fixing. The map replaces a hunch with data.
Where automation doesn't pay off
Being honest means also saying where to keep your hands off. If a task happens rarely and differently every time, you'll never earn back the time spent automating it. Setting up an exception that occurs twice a year costs more than just doing it by hand.
Watch out, too, for tasks where the decision is genuinely human — sensitive communication with a customer, negotiation, a judgment that rests on trust and context. Automation won't help here; it only creates a cold impression and mistakes that are hard to undo.
And one simple rule holds: if you can't describe the process precisely yourself, it isn't ready for automation. Clarify it first. Automating a mess only makes the mess run faster.
Start in one corner, not everywhere at once
The most common mistake isn't picking the wrong task — it's trying to automate half the company in one step. A grand plan sounds ambitious, but it collapses under its own weight. Something doesn't fit, someone gets lost, and the whole project stalls.
It's far better to pick a single task — the one your map flagged as the most frequent and the clearest — and solve only that. Make it work end to end, make people trust it, let it save the first few hours. Only then look at the next corner.
This path of small steps is slower on paper but faster in reality. Every finished step gives you confidence, data, and the appetite to continue. In practice, this is usually how we begin: one corner, one result, then onward.