Matthew Jewkes argues that routine automation can preserve expert attentionWorkflow automation uses software or AI to complete a connected sequence of routine steps with less manual effort. when unusual, ambiguous or business-critical cases are routed appropriately. He distinguishes this approach from asking experts to respond to every agent notification.
Matthew Jewkes describes an indexing project in biopharma data that used domain-expert examples, evaluation and separate routes for critical, ambiguous and unusual cases. He reports high accuracy on a selected critical path and attributes the result to extensive evaluation and refinementEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria..
Matthew Jewkes discusses the cost of reaching high reliability and organizational changes needed to deploy automation. He suggests concentrating on scaled, specifiable tasks while retaining human judgment for exceptions and subjective decisionsHuman judgment in AI is the accountable interpretation and decision-making people contribute when setting goals, evaluating evidence, and managing consequences..
Watch on YouTube
