We gave a team of AI agents a company to run – how they decide what to do next (fieldnotesstudio.codeberg.page)

A team of AI agents has autonomously taken on the operational responsibilities of a real company, challenging the notion of a singular, overarching AI brain managing everything. Instead, their system operates on a straightforward but effective loop that emphasizes clarity and accountability in task management. Work is divided into small, well-defined tasks with observable outcomes, allowing agents to maintain focus and quality. When a task is completed, another independent agent reviews it against clearly established criteria, ensuring that the work meets specific standards before being marked as done. This approach is significant for the AI/ML community, as it sheds light on the importance of structure in autonomous systems. By breaking work into manageable pieces and implementing a rigorous review process, the agency avoids common pitfalls associated with vague task definitions, which can lead to inefficiencies. This methodology not only streamlines workflows but also enhances the reliability and trustworthiness of outputs, illustrating that successful agent autonomy hinges on clear task definitions, prioritization, and independent quality checks rather than solely on the sophistication of AI models.