Classifier and decision model
Picks one answer from a fixed set of options, with a confidence score, instead of writing free text.
What it is
Classifiers and decision models answer structured questions: is this email spam, which team should handle this ticket, is this request safe. Because they choose from known options, they are fast, cheap, easy to measure and their confidence can be checked.
How it works
The model reads the input and outputs a probability for each allowed option. Classic classifiers are small encoders fine-tuned on labeled examples; newer decision models are language models that take the options in the prompt and return a probability for each in one call.
What it is good for
- Routing tickets and messages
- Diagnosing app bugs in the browser, as GenClass does
- Moderation and spam filtering
- Choosing which tool or model to call
- Yes/no checks inside automated workflows
Limits to know
- Only as good as the options you define
- Confidence can be miscalibrated
- Edge cases outside the options get forced into one
