Not every broken process is broken the same way, and the kind matters, because two of the three are worth automating once repaired and one almost never is.

1. Broken by inconsistency

The steps are known and correct. Different people do them differently, or the same person does them differently when tired. Results vary for no good reason.

Good automation candidate, after you write the steps down. This is automation's home ground. A machine's one real virtue is that it does not get tired or creative.

2. Broken by volume

The steps are correct and consistently followed, but there are too many of them for the time available, so things get skipped — usually the checks, because skipping a check has no immediate consequence.

Good candidate, with a caution. Automate the volume, but automate the checks first, not the work. The checks are what is currently being dropped; if you automate the work and leave the checks manual, you have just increased the volume of work being checked by a human who was already behind.

3. Broken by ambiguity

Nobody can quite say what the correct outcome is. The rule has exceptions, the exceptions have exceptions, and the person who does it says “it depends” and cannot finish the sentence.

Do not automate this. Not yet, possibly not ever. Ambiguity in a process is a decision that has not been made. Automating it does not make the decision — it freezes whatever the last person happened to do into a rule that nobody agreed to, and then applies it at scale to cases nobody considered.

The work here is not technical. Someone has to sit down and decide what the answer is. That is a conversation, not a script. When the conversation is done and the rule is real, come back — it will probably be a category 1 problem, and easy.

Why this matters more with AI in the room

An AI model will cheerfully produce an answer for an ambiguous process. It will not tell you the process was ambiguous. It will be fluent, confident, and internally consistent, and it will have invented the rule. This is the single most expensive mistake I see people make right now, and the whole of category 3 exists to keep you out of it.

Last modified: Monday, 17 August 2026, 1:05 PM