AI-generated “workslop” is destroying productivity? (hbr.org)

Companies are rapidly adopting generative AI—usage has roughly doubled since 2023 and the number of firms with fully AI-led processes nearly doubled last year—yet the payoff is starkly missing: an MIT Media Lab report finds 95% of organizations see no measurable return on their AI investments. The result is what critics call “workslop”: high-volume, low-value AI outputs that create more noise than productivity. Common drivers include poor integration with existing workflows, low-quality or hallucinated outputs that require human rework, fragmented tool stacks, and a lack of clear success metrics or governance that turns promising capabilities into operational drag. For the AI/ML community this is a wake-up call to move beyond capability demos and prioritize reliable, measurable systems engineering. Technical fixes include stronger human-in-the-loop designs, domain-specific fine-tuning, retrieval-augmented generation to ground outputs, robust evaluation and observability, and workflow orchestration that minimizes context loss and rework. Equally important are product and organizational changes—clear ROI metrics, change management, and feedback loops—to ensure models actually reduce friction rather than add it. Without that end-to-end focus, broader AI adoption risks scaling mediocre automation rather than delivering real productivity gains.