Judgment work
Judgment work is the part of a job that needs a person to decide, because it depends on context, experience or taste that isn't written down.
Judgment work is the part of a job that needs a person to make a call. It's the step where the right answer depends on things nobody wrote down: what the business cares about this quarter, what a client will put up with, what you learned the last time you tried it.
The term comes up when teams decide what to hand to AI tools and what to keep. AI is good at the steps with clear rules. Judgment work is where you still want someone accountable.
How to spot it
A good test is to ask whether two people who both know the job well could disagree. If they could, it's judgment work.
Pulling last week's visibility numbers into a report has a right answer. Deciding whether a 4-point drop is worth telling the client about doesn't. Neither does choosing which of three competitors to go after first, or whether a wrong claim in an AI answer is worth a call to the publisher.
Judgment work in AI search
AI search tools do a lot of the counting for you. They collect the answers, read which brands are named, and flag what changed. The decisions after that are still yours.
For example, a tool might tell you that answers keep saying Fernhill Coffee "only ships within the UK." Is that a mistake to correct, or true and a reason to start shipping abroad? Only someone at Fernhill can say.
Judgment work in AnswerOrbit
Some decisions are left to you on purpose. In Perception, you mark each objection Mark as true if it's fair, or Mark as untrue if it's wrong, and your recommendations take that into account. In Claims, you can Change verdict when you think we've judged a claim wrong. And an assistant connected to AnswerOrbit can suggest prompts, but someone on your team has to approve them.