AI-Mediated Tacit Knowledge Transfer: A Falsifiable Protocol for Reducing Expert Support Time

This document presents a pre-registered, falsifiable validation protocol for testing whether structured AI-mediated tacit-knowledge transfer can reduce expert-support time while preserving or improving novice decision quality. The protocol evaluates one bounded expert workflow using a predefined comparator, task set, scoring rubric, outcome measures, and failure conditions. Its primary commercial and empirical threshold is a reduction of at least 20% in expert-support time without unacceptable deterioration in decision quality. Secondary measures include time-to-competence, first-pass success, decision consistency, rework, and escalation behavior. This deposit reports no original experimental results and does not claim that the intervention has been validated. It specifies the test to be conducted before outcomes are known. The tested claim concerns partial, workflow-specific transfer of decision support and does not imply complete transfer of expertise or replacement of experts. This protocol supplements the conceptual framework presented in “The Master-Embedded Device v3.0: Toward a Peer-Reviewable Framework for Tacit Knowledge Transfer into AI Agent Systems” (https://doi.org/10.5281/zenodo.20318003).