Skip to content
ki-automation · 1 min read

Where to start with automation (and where not to)

Not every task should be automated. A pragmatic guide to finding the right first steps – without spreading yourself thin.

The most common question I hear isn’t “can this be automated?” – almost anything can be, somehow. The better question is: is it worth it? And this is exactly where most people make the wrong call.

The two axes that matter

Before you build anything, sort your recurring tasks by two questions:

  1. How often does it happen? A task that comes up once a quarter is rarely worth the automation effort – no matter how annoying it is.
  2. How rule-based is it? The clearer the steps, the easier the automation. Tasks that need a lot of human judgment are better suited to AI assistance than to rigid automation.

The most rewarding candidates sit top-right: frequent and rule-based. These are almost always the invisible things – moving data, sending notifications, reconciling lists.

Start small, but start

The second big mistake is thinking too big. No one needs to automate their entire company at once. Pick one workflow that costs you time every week and automate exactly that. The first working workflow changes more than any slide deck about “digital transformation” – because your team suddenly sees what’s possible.

Where AI comes in

Classic automation follows rules. AI comes in where it used to be “a human has to decide this.” Roughly sorting an email, pulling the relevant data from a document, writing a first draft – these are no longer human tasks but groundwork you review and approve.

The key remains: AI doesn’t replace your judgment, it speeds up the path to it. The best automations let humans decide exactly where it counts – and take the rest off their plate.

— Felix Gudowius

nios.digital

Rather act than just read?

If a topic is on your mind, let's talk briefly. A short conversation is often enough to see the next concrete step.