Use Case Triage
Describe an AI use case in plain language. This lightweight rules-based prototype extracts signals from the description, maps them across six decision dimensions, and suggests a next step: discover, experiment, scale candidate, or consider a non-AI intervention.
I deliberately kept this prototype transparent and rule-based rather than adding an LLM where one wasn't necessary. The point is to make the decision logic inspectable, and to demonstrate the very principle it evaluates: not every useful capability needs AI. It runs entirely in your browser, no data leaves the page. Treat it as a first-pass triage that exposes the questions worth asking next, not a substitute for validation.
The verdict matters less than the next question it raises.
How it reads a use case
The prototype does transparent, rules-based signal extraction: it scans the description for cues, then raises or lowers six dimensions, problem strength, AI advantage, trust & data quality, workflow fit, value potential and scale readiness. It weighs things like whether a real, frequent problem is named; whether the work needs judgment a model can genuinely add versus deterministic rules; whether the data is described as reliable or shaky; whether it fits where the work actually happens; and whether it's proven or still needs testing. It's intentionally simple and inspectable, not a machine-learning model.
Then it applies rules, not an average, the same philosophy as the decision framework: a weak problem sends you back to discovery; a real problem with no genuine AI advantage points to a non-AI fix; shaky trust or workflow fit caps it at experiment. Only a strong problem, a real AI edge, trustworthy data, a workflow it fits, clear value and readiness earns a scale recommendation. The value isn't the verdict, it's that it forces the assumptions behind a use case to become explicit.