Trustworthy Driver Compensation Automation: Human-in-the-Loop Reconciliation of Telematics, Trip, and Payroll Data

Abstract

Compensation for commercial drivers can be based on data from various telematics sources, duty status, trip execution, payroll policies and rules, location policies, and manual adjustments. A complete manual reconciliation is labor intensive and subject to quality issues, while a complete autonomous adjustment processing can introduce wage, compliance and explainability risks. This paper discusses a human-in-the-loop architecture for trustworthy driver-compensation automation. Allows the ingestion of source events without modifying the original records; provides a way to align telematics and trip timelines; brings compensation rules to the forefront; computes proposed adjustments; classifies by confidence and exception categories; routes uncertain cases for approved review. All suggested and accepted changes remain in the source evidence, calculation path, change rule version, change by reviewer and downstream payroll status. The architecture is analysed with respect to typical examples with breaks, unmatched trips, missing duty events, overlapping activities, delayed source records, and changes in policy, and an evaluation methodology is suggested based on the consistency of the calculations, the accuracy of the exceptions, the reduction of the records for manual review, the processing time, the auditability and the ability of the system to reject incomplete cases safely. The contribution moves beyond efficiency and incorporates all of the elements of explainability, source provenance, human authority, and employee-impact controls as integral parts of the architecture, not mere add-ons.