What to automate first: the matrix that keeps you from spending on what does not matter.
Before choosing a tool you have to choose a process. How to measure hours, frequency and complexity to decide where to start.
The question that reaches our office is almost always phrased backwards: "Would Power Automate work for us?" or "Should we use Apps Script?". The right answer starts two steps earlier, with which specific activity is consuming hours it should not.
The inventory is a four-column table: activity, who performs it, how many times a month, and how many minutes each time. From that you calculate the real annual cost of the task, valuing the hour of the role that performs it. A 20-minute activity repeated 60 times a month is 240 hours a year: if an analyst does it, the cost comfortably exceeds what automating it would cost.
The second dimension is complexity. Here it pays to be honest with three questions: does the input data always arrive in the same format? can the decision rules be written without ambiguity? can the result be verified automatically? Every "no" multiplies the implementation effort and, more importantly, the maintenance effort afterwards.
Crossing both dimensions gives four quadrants. High savings and low complexity is where you always start: those are the wins that convince the organization. High savings and high complexity gets planned for later, with real budget and time. Low savings and low complexity waits until there is spare capacity. Low savings and high complexity never gets done, even if it is technically possible.
There is one filter almost nobody applies and we put before all the others: should this activity exist at all? In a high share of cases, the analysis reveals that the task was born to compensate for a problem in another process. Consolidating three files that come from three different areas is usually a symptom that the data should be captured once. There, the solution is not to automate the consolidation but to remove the need to consolidate.
That filter is the difference between an automation project and a process-redesign project. The second almost always produces more savings, and the first without the second produces, at best, a faster mess.