How to find which repetitive tasks to automate with AI

The question is not “where can we use AI?” but “where does our team’s time go on work that repeats?”. Start from the work, and only then ask whether AI is the right tool. Sometimes a simple form, a template, or a rule in an existing system is enough, and that is a good result too.

Sit with the people who do the work for a few hours, or ask each person to note what they do in a normal day. Managers often describe the process as it should run; the people doing it know the workarounds, the copying, and the chasing. Write down every task, however small.

A task is a good candidate when several of these are true:

• It happens many times a day or week, in roughly the same way
• It starts from the same kind of input: a call, a message, a form, a document
• It involves copying information from one place to another
• It means answering the same questions again and again
• It involves waiting for, chasing, or checking something
• Mistakes come from tiredness or volume, not from lack of judgement

For each candidate, estimate the minutes it takes and how often it happens, and multiply to get hours a week. Then score it on four questions: Is the information it needs available digitally, or could it be? Does it follow rules, or need judgement each time? What happens if the system gets it wrong? How hard would it be to connect to the systems involved? The worksheet below puts this in one table.

Choose the task with the most hours a week, low risk if something goes wrong, and information that is already digital or easy to capture. Do not start with the most complex process in the business. A first system that ships in weeks and saves real time builds trust, and teaches you what to automate next.

Even in the tasks you do automate, keep a checkpoint where a person approves or takes over. Your team keeps the work that needs judgement; the system handles the repetitive part.

• Tasks that happen rarely: the saving will not cover the cost
• High-stakes decisions where a person must own the outcome
• Work that is mostly about relationships and trust
• Processes that change every time, with no stable pattern

If you want a second opinion, our AI automation audit does this with your team and adds a cost and savings estimate for each opportunity, so you know what to build first and what to leave alone.