Beyond AI Productivity: Enterprise AI Systems, Human–AI Responsibility, and the Future of Work
Year: 2026
Type: preprint
Abstract: A prior U.S. review, Enterprise AI at the Decision Point, established that enterprise artificial intelligence (AI) as-sistants have task-dependent productivity effects and that public evidence does not identify an independent financial return or a complete chain from AI cost to verified workforce replacement. The human side of that productivity record, and any organiza-tional consequences of workforce change, were left open. This study examines responsibilities, training, job insecurity and well-being, human oversight, and performance after U.S. enterprise AI adoption from November 2022 through September 2026. The design is a critical integrative review. Firms that report a task effect most often report exclusive augmentation. Frequent AI use is associated with greater fear of job loss. In one worker panel, displacement concern, rather than use frequency itself, is associated with lower engagement and higher burnout. The contribution is a responsibility-transfer framework with five constructs and four testable propositions. A proposed computa-tional specification of that framework, Human–AI Responsibility-Aware Workflow Optimization (HARWO), places verification and accountability on the workflow graph. Gross task-time savings are separated from net employee time. HARWO is not solved or estimated in this paper. The evidence gap reflects a measurement disconnect: existing studies separately examine task productivity, workforce changes, and employee well-being, but the evidence examined here does not adequately link these outcomes within the same organizations over time. Consequently, whether AI-driven efficiency gains reduce human workload or redistribute responsibilities remains unresolved. The U.S. record is consistent with the separation of task change from employment change, and with anticipatory strain. The performance of organizations that combine AI with workforce reduction remains unmeasured, and the propositions are not empirically validated.
Source: Research Square
Authors: Audrey Rah, David W. Johnson
Institution: University of Houston
Language: English
Topic: AI and HR Technologies
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7225759837