Article & News

A model that answers well in a chat window is not yet usable in a system. The moment the output feeds a database write, an API call or a decision, it has to arrive in a predictable shape every single time, and prose does not do that. Structured output is what closes that gap. Getting structured output right is the least glamorous part of building with language models, and it accounts for a large share of the reliability problems that reach production. Asking politely does not scale The first approach everyone tries is instructing the model to reply in JSON.

Every automated process has a decision to make about people: which steps still involve one, what they are actually reviewing, and what happens when they disagree with the system. Teams tend to answer this by reflex. Either everything gets reviewed, which removes most of the benefit, or nothing does, which works until the first expensive mistake. The useful question is narrower than “should a human check this”. It is which specific decisions carry enough cost when wrong to justify the delay of checking them. Sort the decisions by what being wrong costs Three categories cover most processes. Reversible and cheap: