When onboarding is instrumented to adapt in real time, which combinations of early behavioral signals (such as frequency of prompt edits, depth of gallery browsing, and speed of repeating similar tasks) best predict whether a user will benefit more from (a) stronger scaffolds that accelerate time to first reusable workflow or (b) early exposure to free-form prompting that accelerates prompt skill acquisition and independent execution?

Answer

Early signals can be grouped into two predictive bundles:

(a) Better suited to stronger scaffolds

• Low exploration, low edits, slow repetition
• High error sensitivity and reliance on examples

(b) Better suited to early free-form prompting

• High exploration, high edits, fast repetition
• Lower example reliance and growing direct invocation

Products should detect these bundles in the first few sessions and route users accordingly, then switch tracks if patterns change.