Expert-amplifying AI is the category of artificial intelligence systems designed to preserve the practitioner's expertise as the foundation of the system's capability — to make expert knowledge more powerful and reach farther, rather than to replace expertise with general-purpose capability available to anyone. Such systems are technically feasible and work well in narrow domains. They are the contemporary analog of record playback: the suppressed alternative whose underdevelopment reveals the political content of the dominant paradigm's selection.
The architectural features of expert-amplifying AI are specifiable. The system is trained on the reasoning patterns of a specific clinical team, a specific law firm, a specific engineering practice — capturing not merely outputs but the chains of inference, the weighting of evidence, the contextual factors that shift probabilities. The training data is curated with consent and compensation. The model's reasoning is transparent to the expert user, who can evaluate it, challenge it, and override it. The competitive advantage accrues to the practitioner whose expertise the system amplifies, not to