Most customer-facing AI features follow one of four patterns. An interface or API action layer lets users query a product and trigger actions. A data processing layer searches, summarises, and extracts knowledge. A content creation layer produces drafts or assets. A decisioning layer scores, ranks, forecasts, or recommends.

Classifying the feature before scoping infrastructure matters because each pattern creates different requirements for compute, networking, data preparation, security, evaluation, and governance. A clear pattern helps a team focus on the infrastructure risks created by the feature it plans to build.

The selector takes about two minutes and asks eight questions about the feature, its data, its output, and the actions it can take. The result identifies a dominant pattern or a mixed result, then highlights the infrastructure layer to assess first.