Exposure scores from Eloundou et al. (2023) define AI exposure as the share of occupational tasks large language models can assist with, becoming a central input in future-of-work debates. These static measures suffer from temporal, geographic, and ontological limitations that often fail to travel with them into policy analyses. The authors identify two primary gaps: structural mismatches between static scores and dynamic policy needs, and insufficient coordination between researchers and policymakers. To address measurement limits, the article surveys five research families including dynamic benchmarks, ensemble methods, task-framework extensions, worker-centered metrics, and adoption data. The second gap requires deliberate political work to reimagine future outcomes rather than relying solely on better measurement. Policymakers must widen their evidence base, engage workers as partners, and shift from prediction to preparedness. Researchers are urged to build data infrastructure, adopt participatory methods, and write with policymakers in mind.