Summary of Key Points
This article focuses on the transformation of corporate organizations and industrial opportunities in the era of Agents (AI-powered entities), discussing five key issues: 1) how companies can restructure their organizations around Agents (moving towards smaller teams and the rise of super-individuals); 2) the boundaries of human-machine collaboration and who bears the responsibility for outcomes (AI performs the tasks, but humans are ultimately responsible for the results); 3) the core components of future enterprise "operating systems" (context and memory); 4) the directions for AI entrepreneurship (AI integrated with the physical world presents higher barriers to entry); 5) how to manage Agent employees (using standards instead of KPIs). The article illustrates these points through practical examples from various industry experts, highlighting the profound impact of Agents on corporate structures, collaboration models, and business models.
1. Agents Are Here: Companies Need to Become Smaller and More Agile
The emergence of Agents is challenging the traditional hierarchical structure and departmental barriers in companies:
- Startups have an advantage: They can build their organizations from scratch, designing them directly for Agent-based collaboration without historical constraints (such as departmental silos or outdated processes). Small teams of 2-5 people can accomplish what used to require a team of 10.
- Large companies face challenges: They have many historical burdens (old systems and departmental barriers) that need to be addressed before integrating Agents. However, for new projects, they can quickly set up Agent-based organizations.
- The trend towards smaller teams and super-individuals: Previously, tasks were clearly divided among departments (product, development, testing), but now teams are more project-oriented, leading to greater efficiency. There are even "one-person companies" where a single individual uses Agents to complete multiple projects.
Example: Amazon used 1 senior engineer and 8 senior developers to rewrite an AWS product that previously required 1000 people, resulting in improved performance—demonstrating that small teams combined with AI tools can replace large traditional teams.
2. Human-Machine Collaboration: AI Performs the Tasks, Humans Bear the Responsibility
The division of labor between Agents and humans is clear:
- What AI does: Executes specific tasks (coding, customer follow-up, data processing), and can iterate and optimize autonomously, but it needs human guidance on goals and rules.
- What humans do: Set directions (strategies, priorities), validate results, and take responsibility (making decisions that have consequences). For example, sales agents use Agents to contact 4000 customers, but the ultimate customer relationships and strategic judgments are made by humans.
Key conclusion: AI can improve efficiency, but humans are still responsible for the final outcomes because AI lacks autonomous consciousness and cannot assume legal or business responsibilities.
3. The Future Enterprise "Operating System": Context and Memory Are Key
The future enterprise system will not be just tools like Slack or DingTalk for human-to-human collaboration; it will be a platform where humans and Agents coexist, with two core components:
- Context: Includes all company data (documents, meeting records, code, customer information) that Agents need to function correctly. For example, Agents need to understand the company's products and historical decisions to provide accurate advice.
- Memory: Agents' "long-term memory" allows them to remember past tasks and customer preferences, enabling continuous optimization.
System architecture: The foundation is cloud infrastructure (security, permissions), followed by Agents' core capabilities (memory, autonomous iteration), and on top of that, the collaboration ecosystem between humans and Agents (task allocation, permission management).
Challenges: Context and memory have not yet been standardized, leading to fragmented tools—similar to the lack of unified standards for databases in the early days.
4. AI Entrepreneurship Opportunities: Focus on Physical World Integrations
To succeed in the Agent era, avoid pure online or digital initiatives; instead, focus on AI applications that integrate with the physical world:
- Physical AI: Examples include offline monitoring and analysis (e.g., using AI to analyze shopping cart traffic at Target supermarkets), robotics, computer vision—these are deeply integrated with the real world and difficult for large companies to replace.
- High-barrier scenarios in specific industries: For example, legal services in the United States, where state-specific rules and data inconsistencies make it challenging for large models to perform well.
Avoid common pitfalls: Pure online tools (e.g., financial SaaS, general-purpose legal AI) are easily surpassed by large models like Claude for Financial Services.
Example: Clear Company specializes in legal services for the U.S. judiciary; due to the complexity of state-specific cases, it has a strong competitive advantage. In contrast, Harvey, which offers general-purpose legal AI, is more susceptible to substitution.
5. Managing Agent Employees: Use Standards Instead of KPIs
Agents are not humans, so management should use different methods:
- Evaluation of Agents: Focus on standards such as accuracy (no bugs in code), efficiency (speed of customer handling), and error margins (precision of produced parts).
- The core of management: Solve problems related to Agent performance (e.g., insufficient context, poor collaboration, misunderstandings), not human weaknesses (laziness, greed).
Unchanged fundamental principle: Agents also need goals and context, similar to humans, but the management approach is different.
Summary: Agents are tools, not employees. You don't set KPIs for machines; you just ensure they meet the required standards.
Final Insights
The core change in the Agent era is the redefinition of the roles of "humans" and "systems": Humans shift from executors to decision-makers and supervisors, while systems (Agents) become the primary source of productivity. Future companies will be more flexible and efficient, but they still need to address issues such as standardizing context, clarifying responsibility, and refining business models. For individuals, it's important to learn how to leverage AI tools effectively to enhance their strengths and avoid being replaced. For entrepreneurs, integrating with the physical world and focusing on specific industries presents significant opportunities.