AI agents are on the verge of being recognized as full-fledged workers (www.lemonde.fr)

AI capabilities that were once written off as parroting internet text have matured into autonomous agents that can perform and coordinate real work, and the debate has shifted from “can they?” to “how should they be treated?” Industry leaders now forecast that AI will carry out a substantial share of enterprise work—Paul Harris of Slack/Salesforce estimated as much as 40% of Fortune 1000 work—and macroeconomic data (OECD growth estimates) are already pointing to AI as a driver of resilience. That momentum, plus rapid improvements since early conversational models like ChatGPT, is pushing organizations and regulators to consider recognizing AI agents as operational “workers” embedded in business processes. For the AI/ML community this has concrete technical and governance implications: models must move beyond occasional convincingly human outputs to consistent, auditable, and safe task performance. That means prioritizing agent architectures, robust evaluation metrics, tooling and API integrations for enterprise workflows, provenance and audit trails, and mechanisms for attribution, error-handling and human oversight. Practitioners will need to balance productivity gains with standards for reliability, explainability and liability as AI systems shift from assistants to autonomous contributors in production environments.