Nate B Jones argues that greater agent efficiency does not automatically reduce human work. As people run more agents in parallel, they spend less time watching individual steps but more time choosing jobs, supplying context, checking outcomes and intervening when an entire run goes off course.
Jones says small businesses struggle when agent management becomes another job for time-poor owners or an expensive responsibility delegated to vendors. He contrasts this with enterprises that can fund deployment engineers, security controls, monitoring, evaluation and deeper system integration across many employees.
Verifiable domains such as coding and legal work are advancing faster because experts already have ways to judge outputs. Jones recommends starting with tasks whose results people already review, while expecting managers to develop skills for allocating agents, defining permissions, recognizing failures and recovering safely.
Watch on YouTube