AI Agents Are Changing Work Processes and Increasing Efficiency

• AI agents increase efficiency, but they also cause concerns about job security among workers.
• The CoAct-1 system combines programming and graphical interfaces for better automation.
• Human oversight remains important in working with AI agents.

This is reported by ZDNet, VentureBeat.

Key Findings:

• Trust in AI agents among workers is growing, but only for routine tasks.
• New systems like CoAct-1 improve automation by combining code and GUI.
• Utilizing AI agents can boost productivity, but it requires human oversight.

AI agents are becoming an important part of work processes, especially as companies seek ways to automate routine tasks. However, research shows that workers are only ready to trust agents for less complex tasks. According to survey data conducted by Workday, 75% of respondents expressed readiness to work with AI agents, but only 30% of them were willing to accept their recommendations.

Recent research revealed that trust in agents increases with their utilization. While 90% of respondents believe that AI agents can enhance their productivity, many of them also expressed a growing dependence on such systems, which could lead to a decrease in critical thinking and human interactions.

Researchers from Salesforce and the University of Southern California introduced a new system, CoAct-1, which allows agents not only to perform tasks but also to write code to improve speed and accuracy of work. This system combines the benefits of manipulation with graphical interfaces and programming, allowing for a reduction in the number of clicks needed to complete tasks.

CoAct-1 consists of three agents: Orchestrator, Programmer, and GUI Operator. The Orchestrator defines the overall goal, breaks it down into subtasks, and delegates them to the corresponding agents. This helps significantly reduce the number of steps required to complete tasks and increases their success rate.

During successful testing results of CoAct-1 in controlled conditions, real-world work environments may be complex due to outdated software and unforeseen interfaces. This underscores the importance of human oversight in utilizing AI agents for critical tasks.

Tags: Technology/AI