How to Evaluate Whether a Workflow Should Be Automated at All

Some workflows should not be automated yet. The task might be too vague, too unstable, too political, or too dependent on judgement that the team has not even defined clearly.

A workflow is a good automation candidate when the value is clear, the steps are legible, the failure cost is tolerable, and the inputs are available in a repeatable way.

Start with frequency, friction, and payoff

If the workflow barely happens, automation is often a vanity project. You are adding complexity to avoid a cost that barely exists.

The best candidates are recurring tasks with obvious friction and a meaningful payoff if they become faster, cleaner, or more consistent.

Check whether the workflow is actually stable enough

Some work changes every time. The inputs vary, the owners vary, the definition of “done” varies, and the edge cases dominate. That is a bad automation starting point.

If the workflow itself is still unstable, standardise the process before you automate it.

Know where the judgement still belongs

Automation is not just about whether a model can produce an output. It is about whether the surrounding business is comfortable letting that output move forward.

That means identifying review gates, risk levels, and failure costs before you touch tooling. Otherwise the automation project becomes a fight about trust instead of a design exercise.