Over the past two years, we have encountered numerous AI tools and plugins, such as ChatGPT for writing, MidJourney for drawing, Aiva for composing, and Sora for video production. These AI tools have amazed us with their powerful capabilities, outperforming humans in executing specific tasks with remarkable efficiency. However, this is only the prelude to this transformative journey.
Imagine an opportunity where we can create new employees (AI Agents) based on various business scenarios. These agents are not only incredibly diligent and rational but also possess vast knowledge bases. They relentlessly pursue set goals, respond to every task methodically, and have the capability to learn and grow autonomously, always ready to expand their skillsets through new execution tools and plugins. Most importantly, these employees require no salary, never tire, and are proficient in multiple languages.
This describes the future scenario of AI Agents. Whether we believe it or not, the evolution brought by AI in the labor market and social organization is accelerating.
Agent Hospital: Are AI Doctors More Reliable Than Human Doctors?
Recently, there has been an intriguing case of AI Agent application in the medical field called "Agent Hospital." Developed by a research team from Tsinghua University, this hospital simulation system replicates the complete medical process from reception, diagnosis, to treatment, mimicking various consulting and examination rooms of a real hospital. AI Agents, acting as 14 doctors and 4 nurses, completed the entire process from triage, registration, consultation, medical examination, diagnosis, drug dispensing, rehabilitation, and follow-up, interacting autonomously with patients throughout.
Driven by large language models (LLMs), these AI Agents treated tens of thousands of patients within a few days, achieving an efficiency 500 times that of human doctors. The AI Agents' accuracy rates in examination, diagnosis, and treatment were 88%, 95.6%, and 77.6%, respectively. The accuracy rate for diagnosing respiratory diseases reached 93.06%, surpassing human experts' 87%. Although this system has not yet been applied to real patients, its performance on the medical Q&A dataset MedQA has already surpassed existing technologies.
Tsinghua University's research, though merely a technical experiment in AI Agent application, has profound implications akin to a societal fable. All labor-intensive and long-chain complex business scenarios could be fundamentally restructured by the advent of AI Agents. Even knowledge-based industries might be partially or completely replaced in this technological revolution.
Why Can AI Agents Become "Super Employees"?
AI Agents can become super employees because they essentially integrate various AI tools to execute complex workflows. As AI tools continue to enhance and expand, the working capabilities of AI Agents also keep improving. More critically, AI Agents adopt the OKR (Objectives and Key Results) methodology, which helps clarify goals and continuously optimize processes to ensure ultimate success. Observers predict that in future human societies, any individual might train AI Agents to build a vast enterprise, a phenomenon termed as "one-person unicorn."
The workflow of AI Agents includes the following seven key steps:
• Set Goals: Establish a clear goal for the AI Agent as the starting point of its workflow.
• Create Task List: Based on the set goal, the AI Agent automatically creates a task list, clarifying the tasks to be completed, task priorities, and execution order, while pre-setting solutions for potential issues.
• Gather Information: The AI Agent collects the information needed to execute tasks, including web searches, database access, and interactions with other AI models.
• Manage Data: The AI Agent continuously manages and analyzes the collected data, including directly collected information and knowledge and experiences accumulated from internal and external interactions.
• Integrate Feedback: The AI Agent utilizes market data, customer feedback, and information from internal monitoring systems to evaluate its progress towards the goal and adjust task strategies and tool usage accordingly.
• Continuous Operation Until Goal Achievement: The AI Agent operates continuously through constant action, feedback, and adjustments until the set goal is achieved. This continuous automatic operation distinguishes it from traditional software and is a significant feature.
• Adaptive Learning: Throughout the work process, the AI Agent not only executes tasks but also continuously learns from practice, improving efficiency and adapting to new challenges and environments through accumulated experience.
Understanding the working principles of AI Agents reveals that with well-trained AI Agents, not only can enterprises restructure all business scenarios, but individuals can also mobilize significant resources to achieve complex business collaborations. This is epoch-making for enterprises seeking efficiency optimization and individuals aiming to expand their capabilities.
Mastering AI Agents: Everyone Can Become a "Super Employee"
In July 2023, McKinsey released a 76-page report detailing the future impact of AI on the US labor market. The report predicts that by 2030, AI Agents could replace 30% of the total work hours in the US economy, with an estimated 12 million people needing to transition to new careers. This presents challenges not only for individuals but also for policymakers, business leaders, and society as a whole.
The report sincerely outlines various career transition paths, such as an automotive technician becoming a wind turbine technician. AI Agents can provide real-time fault diagnosis and repair methods through simulated wind turbine maintenance scenarios, using on-site data. This enables workers to proficiently use advanced tools and technologies, effectively integrate and utilize complex resources, and master intricate workflows.
Complex Business Environments Are Ready to Create "Super Employees"
Compared to countries like the United States, the Chinese market shows unique strategies and paces in adopting AI Agent technology. On one hand, China's market pragmatism emphasizes rapid implementation and large-scale application. On the other hand, the highly competitive market environment has already helped China accumulate extensive operational data and user interaction experiences. This data includes not only direct business operation data but also user behavior data and data generated in various simulated environments, providing valuable resources for AI Agent model training.
For example, the role of community operations integrates multiple roles such as marketing, customer service, and sales. Community operators need to create engaging content, like posting pictures and videos on social media, be proficient in online sales, understand customer pain points, and drive sales conversions. During the service and sales process, they also need to effectively communicate and collaborate with various internal departments. Additionally, community operators are responsible for collecting and analyzing operational data and reporting to superiors. Such rapidly growing professionals in these environments provide rich learning samples for AI models. These real complex business operation cases give China a potential advantage in implementing AI Agents and training models.