Aaron Levie distinguishes personal assistants from agents that execute enterprise tasks continuously without a person initiating each step. In conversation with Sophia Dew, he describes parallel software review, scientific research and processing of contracts or invoices as workloads that could drive much greater demand. His estimate of a hundredfold to thousandfold increase is a forecast, not a measured near-term result.
Aaron Levie says enterprises need fine-grained permissions, audit trails, anomaly detection and ways to stop agents before large-scale deployment is safe. He describes controlling read and write access to organizational data and treats autonomous agents as potential insider risks. Stronger models alone do not supply those safeguards or connect existing business systems.
Aaron Levie frames this transition as a process that could unfold over one or two decades. Physical compute capacity, data integration and workflow redesign must advance together, with technical teams helping organizations decide where humans remain in the loop. The discussion separates eventual economic incentives from any confident prediction of rollout speed.
Watch on YouTube
