What should you automate first?

It is easy to start thinking about automation by choosing a tool. Someone demonstrates a new AI feature, integration or document generator, and the idea of introducing it into the business quickly follows. The problem is that a tool will not fix a process nobody understands properly.

A better starting point is less eye-catching: choose one repeatable task and find out what actually happens from the moment it begins until it is complete.

Choose a task that recurs regularly

The easiest tasks to automate recur in a similar form. They might involve copying data from messages into a CRM, preparing a document from a form or sorting customer enquiries.

Not every case has to be identical. What matters is that most cases follow the same stages and use similar data. If every decision depends on a different context, the working rules need to be clarified first.

Count the work that does not appear in reports

The time spent carrying out the task tells only part of the story. Add the time spent waiting for data, correcting mistakes, looking for the right file and handing a case from one person to another. These small interruptions are often why a simple process takes several days.

For one week, record:

• how often the team carries out the task,
• how many minutes a typical case takes,
• where missing information or errors occur,
• which situations require a person to make a decision.

You do not need an elaborate report. A short table is usually enough to compare several automation ideas.

Separate rules from exceptions

A system handles a clearly defined workflow well. It can receive data, check required fields, create a record and send a notification. Cases that fall outside those rules should go to a person with an explanation of what is missing.

This distinction matters. Trying to handle every rare case from the outset increases the cost and lengthens the implementation timeline. The first version can automate routine cases and direct everything else to the person who can make the decision.

Check whether the data is available

Even a simple idea will stall if the information it needs exists only in scans, private inboxes or inconsistently named files. Before work begins, check where the system will obtain its data and where it should save the result.

It is also worth establishing whether the software you use provides an API or supports secure data exports. If it does not, a solution may still be possible, but the additional work and ongoing maintenance need to be included in the plan.

Define one simple measure of success

Before implementation, write down how you will know whether the change has helped. For one process, success may mean a shorter handling time. For another, the error rate or the number of cases the team can complete without copying data manually may matter more.

Choose a measure you can check both before and after implementation. Without a baseline, it is difficult to tell whether the automation solved the problem or merely moved it into a new system.

Start with a narrow scope

The first project should quickly show the team whether the chosen direction makes sense. Select one type of case, one data source and one clearly defined outcome. Once it is running, observe the exceptions, collect user feedback and only then extend the scope.

If you already have a process you would like to improve, see how we approach business process automation or describe it in our contact form. A few sentences about who carries out the work, what data they use and where delays occur most often are enough to get started.