AI agents and the evolving workplace
A recent Stanford University study, conducted between January and May 2025, offers a comprehensive framework for understanding the future of work with AI agents. The research balances worker preferences with technological capabilities, introducing the Human Agency Scale (HAS) to quantify desired human involvement in tasks. This study highlights critical mismatches and opportunities for tailoring AI agents to meet human desires, ultimately reshaping core human competencies.
Points clés
- A Stanford University study, conducted between January and May 2025, provides insights into the future of work with AI agents.
- The study introduces the Human Agency Scale (HAS), a five-level system to quantify desired human involvement in tasks, ranging from H1 (full automation) to H5 (essential human augmentation).
- Approximately 80% of U.S. workers may see large language models (LLMs) affect at least 10% of their tasks.
- Usage data from Anthropic in early 2025 shows that workers in 36% of occupations were already using AI for at least 25% of their tasks.
- Workers often prefer higher levels of human agency than experts deem technologically necessary, indicating a gap in AI development priorities.
- For 46.1% of tasks, workers express positive attitudes toward AI agent automation.
- Conversely, 28% of workers express negative sentiments about AI agent automation, primarily due to lack of trust (45%), fear of job replacement (23%), and absence of human qualities in AI (16.3%).
- The research suggests a shift in valued workplace skills, with technical and information-heavy abilities becoming vulnerable, while interpersonal skills like planning, teaching, and communication rise in value.
- The highest concentration of tasks (112 tasks) aligns with workers desiring H2 and experts assessing H2 as feasible, indicating general alignment between preferences and feasibility.
- Cobus Greyling, Chief Evangelist at Kore.ai, shares insights on the intersection of AI and language, including AI agents and development frameworks.
À retenir
So, it seems AI isn’t going to steal all our jobs, just the boring, repetitive ones. Apparently, we humans are still needed for things like “planning” and “talking,” which is a relief for those of us who excel at office gossip and organizing potlucks. But seriously, if you’re still clinging to your “information management” skills, it might be time to brush up on your charm and witty banter, because the robots are coming for your spreadsheets, but they can’t quite master the art of the awkward team-building exercise… yet.
Sources
