As agentic AI becomes more integrated into the platforms of intelligent automation software providers, a research team from MIT has released a series of studies exploring how the technology handles reasoning, negotiation, collaboration and trust-related tasks that will be central to its efficacy in assisting humans in their work and private lives.
The research from the MIT Initiative on the Digital Economy, led by MIT Sloan professor Sinan Aral and postdoctoral fellow Harang Ju, noted that, the Large language models, which are now evolving into agentic AI systems, are designed to make decisions in complex, real-world contexts. While their generative capabilities are well-documented, the reports say, their decision-making processes remain poorly understood.
In an effort to expand that understanding, the MIT team examined issues in four separate studies: how the technology handles exceptions, how AI agents change productivity, performance, and work processes when collaborating with humans, how AI agents handle negotiating, and how human trust in AI search is evolving.
In the collaboration experiment, for example, researchers set up two teams in a collaboration platform mimicking a marketing project, one with all human members, the other with humans and AI agents collaborating. Communication, collaboration, and workflow logs revealed that collaborating with AI agents increased communication by 137 percent and allowed humans to focus 23 percent more on text and image content generation messaging and 20 percent less on direct text editing. Humans on Human-AI teams sent 23 percent fewer social messages, creating 60 percent greater productivity per worker and higher-quality ad copy. In contrast, all-human teams produced higher-quality images, suggesting that AI agents require fine-tuning for multimodal workflows.
“We are already well into the Agentic Age [of AI],” says Professor Aral. “Companies are developing and deploying autonomous, multimodal AI agents in a vast array of tasks. But our understanding of how to work with AI agents to maximize productivity and performance, as well as the societal implications of this dramatic turn toward agentic AI, is nascent, if not nonexistent.”