Where automation actually pays off

The instinct to automate the most annoying process first is usually wrong. The most annoying process is often annoying because it's inherently variable — and variable processes are the hardest to automate reliably.

A better filter: automate processes that are high-volume and low-variance first. A weekly report compiled the same way every time is a better automation candidate than a customer escalation process that changes shape every time it runs.

We map this on two axes before recommending anything: how often does the process run, and how much does it vary each time it runs. High-frequency, low-variance work is where automation pays back fastest — often within a single quarter.

The processes people most want to automate — judgment calls, exceptions, anything involving negotiation — are usually better served by better tooling for the human doing the work, not by removing the human. Knowing the difference is most of the job.