X-SYNTH Framework Uses Human Attention for Enterprise Context

X-SYNTH is a framework for enterprise context synthesis that leverages human attention, the digitally observable interaction signatures of workers, to improve AI agent task performance. Current retrieval methods match request content to stored system state, which works for narrow requests but fails for complex agentic tasks, resulting in low True Lead Rate and high False Lead Rate. X-SYNTH encodes not just what workers did but the sequence of their actions, capturing behavioral patterns absent from retrieval indices. The framework is presented in arXiv paper 2605.15505.