When not to automate

Automating something nobody needs

The first question about any process is not how to automate it but whether it should exist. Plenty of recurring work is vestigial: a report that was needed for a decision nobody makes anymore, a form filled in for a person who left.

Automating that is worse than leaving it, because you have now spent money making something permanent that was on its way to being questioned.

Ask who consumes the output and what they do with it. If the honest answer is nothing, delete the process.

When judgement is the job

Some work looks repetitive and is actually a judgement call performed quickly by someone experienced. Automating the visible steps while discarding the judgement produces something that looks right and behaves badly in exactly the cases that mattered.

The tell is when the person doing it cannot fully explain the rule. That usually means the rule is real and contextual rather than absent, and the right move is a tool that supports the decision rather than replaces it.

When volume does not justify it

A process that runs twice a year is rarely worth building for, no matter how irritating it is when it comes round. Irritation and cost are not the same thing, and roadmaps built from irritation solve the wrong problems.

Total annual handling time is the number that matters. A twenty-minute task done weekly beats a two-day task done annually, even though the second one feels worse.

When the process is about to change

Automating a workflow that is being restructured next quarter is building on sand. The same applies to a process wrapped around a system you are actively considering replacing.

Wait, or automate deliberately at the layer that will survive. Otherwise the build is scrapped along with the process it encoded.

Saying so is part of the job

A consultancy that never recommends against building is not exercising judgement, it is selling. Every assessment we produce has a section on what to leave alone, and it is frequently the part clients quote back to us.

Being told what not to do is worth paying for, because it is the advice nobody with something to sell has an incentive to give.