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.

Shown separatelyLLM
Replaceability
0—100

How much of the core value can a general-purpose LLM reproduce? High is vulnerable.

Product necessity formula

0125%

Workflow depth

Does the product own the stages before and after an AI response?

0215%

Domain UX

Does its interface outperform a chat box for the job?

0320%

Data advantage

Does it have proprietary or accumulated context the model lacks?

0415%

Action capability

Can it safely do the work instead of merely recommending it?

0510%

Switching cost

Does value accumulate in state, integrations and collaboration?

0615%

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