Author: Zhang Feng
I. AI Becomes "Agentic Users," Defining New Boundaries for Human-Machine Collaboration
Recently, Microsoft previewed a new type of AI agent called "Agentic Users" in its product roadmap. These agents will have their own email accounts and can autonomously participate in meetings and handle tasks.This signifies that AI is evolving from a passive tool into an active collaborator with a certain "agent" identity.This transformation is not an isolated event, but rather the inevitable result of the long-term investment of tech giants like Microsoft in the field of AI Agents. Microsoft defines AI Agents as intelligent systems capable of automating repetitive, low-error-rate tasks by writing and executing code, thereby unlocking value in scenarios requiring massive data processing and precise computation, such as finance and education. However, as AI Agents become increasingly autonomous and even begin to mimic the "identity" of human employees, a series of fundamental questions arise: In cutting-edge fields such as quantum networks and digital finance, how will highly autonomous AI impact existing workflows and decision-making mechanisms? Does the "Rotifer Intelligent Agent Autonomous Evolution Protocol" represent a technological concept that foreshadows AI evolving independently, deviating from a predetermined path? Given the current imperfections in digital governance and compliance frameworks, how should we construct rules to ensure the prosperity of the open-source technology ecosystem while mitigating the risk of loss of control? These questions all point to a core issue: we are standing at a critical juncture in the paradigm shift of human-machine relationships, urgently needing to draw a clear blueprint for the upcoming "intelligent agent society." II. The Evolution from Automated Scripts to "Agent Users" The concept of AI Agents did not emerge overnight; its development has closely followed the leap in artificial intelligence capabilities over the past decade, particularly in Large Language Models (LLM). Microsoft research indicates that, by leveraging its ability to extract logical reasoning from data, large language models can support complex decision-making processes, helping to autonomously execute tasks and thus function as intelligent agents in various workflows. This technological foundation has enabled AI to evolve from executing simple, fixed automated scripts (such as traditional RPA robotic process automation) to "intelligent agents" capable of understanding natural language instructions and planning and executing multi-step tasks. Looking back at Microsoft's practical path, this evolutionary trajectory is clearly visible. Early on, AI applications focused on improving efficiency in specific scenarios. For example, in the medical field, intelligent Power Automate RPA processes were used to connect with hospital information systems (HIS), replacing large-scale repetitive administrative work and improving the resource utilization efficiency of medical teams. This can be seen as the prototype of an AI agent—focused on automating specific tasks. As the technology matured, the focus shifted to building more general and autonomous agent frameworks. Microsoft offers open-source tools and SDKs such as AutoGen and Semantic Kernel at the Infrastructure as a Service (IaaS) level, aiming to provide enterprises with readily available and stable solutions for developing intelligent agents. The pinnacle of development lies in the exploration of "embodied intelligence" and universal agents. Microsoft's research team published a forward-looking paper on "Agent AI," the first attempt to pre-train a foundational model for developing general AI agents by integrating embodied data collected from fields such as robotics. From tools for improving efficiency to programmable frameworks, and then to "agent users" pursuing versatility and autonomy, AI agents have undergone a transformation from mere "technique" to "principle" over the past decade, laying the historical and technological foundation for today's widespread applications. III. Technological Breakthroughs, Business Needs, and Ecosystem Competition Jointly Drive the Agent Wave Why has the AI Agent suddenly emerged and become the focus of the industry at this particular moment? Behind it lies the intertwining and resonance of three driving forces: technology, demand, and ecosystem. First, continuous breakthroughs in core technologies are the fundamental driving force. The leaps in large language models in code generation (such as WaveCoder), logical reasoning, and contextual understanding have given AI Agents a "brain." Cloud computing platforms provide powerful computing power and a stable operating environment, while open-source frameworks have significantly lowered the development threshold. For example, Microsoft, through tools such as the Semantic Kernel, allows developers to more easily build intelligent agents that understand semantics and call external tools and APIs. These technological advancements collectively address the key questions of whether intelligent agents "can think" and "how to act." Secondly, the urgent needs of enterprises for cost reduction, efficiency improvement, and digital transformation provide market pull. In an increasingly competitive global market, companies are eager to free employees from repetitive, low-value labor, allowing them to focus on innovation and strategic decision-making. AI Agents excel at this, processing massive amounts of data and performing precise calculations with "high efficiency and low error rates." From risk modeling in the financial industry to process optimization in manufacturing, intelligent agents have become the core engine for enterprises to unleash the potential of data and build intelligent applications. Industry events such as Microsoft AI Summit Taipei, with AI Agents as their central theme, reflect the strong expectation of the business community for a new chapter in human-machine collaboration. Finally, strategic positioning for building the future ecosystem creates competitive momentum. AI Agents are considered the core entry point and operating system for next-generation human-computer interaction. Whoever controls the dominant platform and protocols for intelligent agents is likely to occupy a pivotal position in the future digital ecosystem. Microsoft is vigorously promoting its Copilot and Agent ecosystem and continuously holding the "Microsoft AI Genius" series of developer events, aiming to consolidate its full-stack advantages from development tools to cloud platforms, gather a developer community, and build a thriving ecosystem of intelligent agent applications. This platform-level competition accelerates the process of AI Agent technology moving from the laboratory to industrial applications. IV. Constructing a Three-in-One Intelligent Agent Development System: Framework-Evolution-Governance Faced with the opportunities and challenges brought by AI Agents, we need a systematic solution, not just piecemeal technical fixes. This system should encompass three levels: technical framework, evolution mechanism, and governance rules. First, rely on robust open-source frameworks to lower the application threshold and ensure security and controllability. Enterprises introducing AI Agents should not reinvent the wheel but should base them on proven open-source frameworks. Tools like Microsoft's AutoGen and Semantic Kernel, supported by official teams, provide readily available and stable solutions. These define standard ways for intelligent agents to interact with the external world (such as through the MCP-Model Context Protocol), but we must also acknowledge the current protocols' security shortcomings and actively improve them through community contributions. Enterprises can leverage this foundation, combining their expertise in digital finance, quantum network simulation, and other fields, to develop intelligent agents for vertical scenarios, achieving rapid and secure deployment. Secondly, we should explore controlled autonomous evolution protocols to guide the positive growth of intelligent agent capabilities. Concepts like the "Rotifer intelligent agent autonomous evolution protocol" represent a cutting-edge direction for enabling AI to learn and iteratively optimize itself in specific environments. The key is "controlled." In highly simulated digital twin environments (such as virtual financial markets or quantum computing networks), we can set clear evolutionary goals and boundary rules for intelligent agents, allowing them to autonomously explore strategies through reinforcement learning and other methods. This not only accelerates the growth of AI's application capabilities in complex fields but also confines the evolutionary process within a secure sandbox, providing valuable data for studying its behavioral patterns. Third, establish a forward-looking digital governance and compliance framework to set rules for an intelligent agent society. When AI agents become "proxy users," existing legal and ethical frameworks face direct challenges. Solutions must be developed first. This includes: defining the legal responsibility of intelligent agents (developers, users, or the intelligent agent itself?); establishing auditing and traceability mechanisms for their operations to ensure transparency in decision-making in key areas such as financial transactions; and developing data privacy and security standards to prevent intelligent agents from abusing their authority. The construction of a governance framework requires the joint participation of technical experts, legal scholars, policymakers, and business representatives, and should be integrated into the design of the open-source technology ecosystem to achieve "governance as code." V. AI Agents Are Irreversible, Require Safety, Inclusivity, and Benevolence. The allure of AI agents is irreversible. While actively developing them, we must remain clear-headed and avoid several potential pitfalls and risks. First, be wary of the illusion of "complete autonomy" and adhere to the fundamental principle of human-in-the-loop. No matter how intelligent an AI agent is, its essence remains an extension of human intent and design. The "agent user" described by Microsoft still aims to improve the efficiency of "human-machine collaboration." We must avoid designing or using "strongly autonomous intelligent agents" that are completely free from human supervision and can set their own ultimate goals. Key decisions, especially in areas such as medical diagnosis, financial risk control, and judicial assessment, must retain the final review and veto power of human experts. The technical architecture should include built-in "circuit breakers" and intervention channels. Secondly, it is crucial to guard against the risks of widening the digital divide and ecosystem lock-in. Powerful AI agent platforms and frameworks may be dominated by a few tech giants, potentially preventing SMEs from equally enjoying technological benefits due to excessively high technical and financial barriers, thus exacerbating the digital divide. Simultaneously, over-reliance on closed ecosystems of single vendors carries the risk of lock-in. Therefore, while embracing excellent solutions provided by companies like Microsoft, the industry should actively promote the development of cross-platform interoperability standards and encourage the development of a diverse and open open-source technology ecosystem to ensure a healthy competitive and innovative environment. Thirdly, attention should be paid to the challenges of employment structure transformation and social adaptation. While AI agents automate a large number of tasks, they inevitably impact existing jobs. Society cannot only focus on technology deployment but also needs to simultaneously plan for workforce retraining and education system reform. Future education should emphasize cultivating creativity, critical thinking, and the ability to work collaboratively with AI to help workers adapt to the new human-machine symbiotic work model. Enterprises also need to take responsibility for providing transition paths for affected employees. Fourth, ethical and bias issues will amplify with increased autonomy, requiring continuous governance. Intelligent agents, based on data training and interactive learning, may inherit or even amplify existing biases and injustices in human society. This harm will be amplified when they are given more autonomous decision-making power. Therefore, ethical review and bias detection of AI agents must be carried out throughout their entire lifecycle of development, deployment, and evolution, becoming a continuous governance effort rather than a one-time certification. Looking to the future, the evolution of AI agents is irreversible, ushering in a new chapter of intelligent applications. The success of this revolution depends not only on the elegance of the code and the power of the algorithms, but also on our ability to build a safe, inclusive, and benevolent development framework for it with a high sense of responsibility and forward-thinking wisdom. Only in this way can intelligent agents truly become powerful partners for humanity in expanding the boundaries of cognition and solving complex challenges, jointly moving towards a more efficient and creative future.
Kikyo
Weiliang
Alex
Brian
Catherine
Miyuki