The scoring framework
Judge the system.
Not the model.
Our framework separates commodity generation from the workflow, data and action layers that can make an AI product durable.
Replaceability0—100
How much of the core value can a general-purpose LLM reproduce? High is vulnerable.
Product necessity formula
Workflow depth
Does the product own the stages before and after an AI response?
Domain UX
Does its interface outperform a chat box for the job?
Data advantage
Does it have proprietary or accumulated context the model lacks?
Action capability
Can it safely do the work instead of merely recommending it?
Switching cost
Does value accumulate in state, integrations and collaboration?
Feedback loop
Does the product improve from real user outcomes?
Editorial principle
“A wrapper is not automatically bad. A thin workflow is.”
Every score is editorial judgment grounded in observed product behavior. We separate facts from interpretation, explain why and pair criticism with constructive recommendations.
See the method in practice