Automation in a small team: where to start
Start with the task that is frequent, boring and rule-based, and where a failure breaks nothing serious. Copying data between tools, sending the same follow-up, building the same report. Do not start with the most complex thing or the one that annoys you most: start with whatever happens the most times per week.
The arithmetic that settles it
Multiply minutes per instance × instances per week. A two-minute task done forty times a week outweighs an hour-long task done once a month, and it is far easier to automate.
Intuition almost always picks wrong: it picks what irritates most, not what consumes most.
Three candidates that almost always work
Copying data between two tools. If someone reads something off one screen and types it into another, that is an automation with immediate return and near-zero risk.
The message that is always the same. Confirmations, reminders, the 48-hour follow-up email. Written once, fires on its own.
The Monday report. If someone builds the same summary every week from three places, automating it recovers half a day per month.
Three to leave alone for now
Anything requiring judgement. If the right answer depends on context and changes case by case, automating it produces confidently wrong answers.
Anything touching money without review. Invoicing, paying, collecting. Automate the preparation, leave approval with a person.
Anything one person does their own way. Before automating it you have to write down how it is done, and that exercise alone usually saves more than the automation.
A ceiling before you spend
Decide in advance how much you will invest and when you stop. A rule I use: three attempts, and if two usable results do not come out, that path closes.
Without that written ceiling, automations become a hole that swallows time, because it always looks like you are nearly there.
How to know it worked
Run the same arithmetic a month later. If minutes per week did not drop, it did not work, however elegant it looks.
And check whether new work appeared: maintaining the automation, reviewing its errors, fixing it when something changes. Those are minutes too, and they have to be subtracted.
Frequent questions
Do I need AI to automate?
For most first steps, no. Moving data between tools and sending rule-based messages is solved without AI, cheaper and more predictably. AI earns its place where there is text to interpret or classify, which is usually the second step, not the first.
What does it cost to start?
A small team’s first useful automation usually runs from a few hundred to a couple of thousand dollars depending on how many systems must be connected. If the minutes arithmetic does not pay it back in three months, you picked the wrong task.