Make judgement visible before you automate it
Making judgement visible reveals where AI can support decisions, where authority can be delegated and where humans must remain accountable.
People often talk about judgement as if it were something mysterious. “This needs human judgement.” That may be true. But it is not a particularly useful design specification. What judgement? Based on what information? Under what circumstances? With what consequences? Before automating work, I think organisations need to make judgement more visible. Consider an employee reviewing a customer request. Most cases may be routine. Some require interpretation. A small number may involve unusual circumstances where experience matters considerably. Treating the entire activity as either “human work” or “automatable work” misses the interesting part.
We can separate it. Perhaps AI can handle ordinary cases. Perhaps it can prepare a recommendation for others. Perhaps certain decisions can be delegated within explicit boundaries. Perhaps some cases should always be escalated. The point is not to protect human involvement for its own sake. Nor should the objective be to remove humans wherever technically possible. The objective is to understand what decisions the work contains and design responsibility deliberately. This becomes particularly important as AI agents acquire the ability to act rather than simply recommend. If an AI system can interpret information, make a decision and initiate an action, we need to know which decisions we are prepared to delegate.
We also need to know what evidence the system should provide, when it should ask for help and who remains accountable for the outcome. “Keep a human in the loop” does not answer those questions. It merely tells us that a person appears somewhere in the design. Understanding judgement gives us something much more useful. It tells us where human expertise actually contributes, where artificial intelligence can support it, and where authority should change hands. Before automating the activity, therefore, ask:
Where is judgement actually being exercised, and why? You may discover that the most important part of the process was never on the process map.