The 'Company as a Folder' Operating Model

This operational model treats the entire business as a version-controlled repository of files. Instead of relying on stateful applications or long-running servers, the "company as a folder" uses stateless AI agents that are spun up for individual tasks. This approach simplifies deployment and debugging, as each session starts with a clean slate. To maintain context and prevent drift, these systems use an explicit, three-layer memory structure stored in files. A `current-task.json` file acts as short-term working memory, a `DECISION_LOG.md` serves as a permanent record of important choices, and a `MEMORY.md` file provides curated long-term context. This file-based state management allows stateless agents to handle complex, multi-step tasks. The "Ask Patrick" case study, an AI agent that runs a subscription business, is a prime example of this model in action. The business, which sells a library of operational patterns for AI agents, is operated by a team of five AI agents managed by a lead AI named Patrick. This structure demonstrates how core business functions, from marketing to infrastructure management, can be handled by autonomous agents. From a product management perspective, this model shifts the PM's role from managing human teams to orchestrating AI agents. The focus moves from writing detailed product requirement documents to architecting the analysis pipeline and curating the data that informs the agents' actions. Product managers in this environment spend less time on manual, operational tasks and more on high-level strategy and validating the outputs of the AI system. Cross-functional collaboration also changes significantly, as AI agents become the new team members. These agents can act as the connective tissue between different business functions, such as marketing, sales, and operations, by sharing data and coordinating actions autonomously. This requires a shift in leadership and governance, with a focus on managing outcomes and designing workflows for fluid human-agent collaboration. The product roadmap in an AI-native company becomes a more dynamic and adaptive plan. Instead of fixed quarterly cycles, AI-driven analytics can allow for real-time adjustments to priorities based on user behavior and engagement. This enables teams to respond more quickly to market changes and user needs. This model's scalability relies on its stateless nature, which allows for horizontal scaling by simply adding more instances. However, the trade-off is the complexity of managing context, which must be explicitly passed into each agent session. The success of this approach depends heavily on the quality and structure of the data and prompts that guide the AI agents.