AI Research

From Skills to Talent: Organising Heterogeneous Age... | AI Research

Key Takeaways

The research paper "From Skills to Talent: Organising Heterogeneous Agents as a Real-World Company" introduces OneManCompany (OMC), a framework designed to m...
We argue that this reflects a deeper absence: a principled organisational layer that governs how a workforce of agents is assembled, governed, and improved over time, decoupled from what individual agents know.
To fill this gap, we introduce \emph{OneManCompany (OMC)}, a framework that elevates multi-agent systems to the organisational level.
OMC encapsulates skills, tools, and runtime configurations into portable agent identities called \emph{Talents}, orchestrated through typed organisational interfaces that abstract over heterogeneous backends.
A community-driven \emph{Talent Market} enables on-demand recruitment, allowing the organisation to close capability gaps and reconfigure itself dynamically during execution.

Paper AbstractExpand

Individual agent capabilities have advanced rapidly through modular skills and tool integrations, yet multi-agent systems remain constrained by fixed team structures, tightly coupled coordination logic, and session-bound learning. We argue that this reflects a deeper absence: a principled organisational layer that governs how a workforce of agents is assembled, governed, and improved over time, decoupled from what individual agents know. To fill this gap, we introduce \emph{OneManCompany (OMC)}, a framework that elevates multi-agent systems to the organisational level. OMC encapsulates skills, tools, and runtime configurations into portable agent identities called \emph{Talents}, orchestrated through typed organisational interfaces that abstract over heterogeneous backends. A community-driven \emph{Talent Market} enables on-demand recruitment, allowing the organisation to close capability gaps and reconfigure itself dynamically during execution. Organisational decision-making is operationalised through an \emph{Explore-Execute-Review} ($\text{E}^2$R) tree search, which unifies planning, execution, and evaluation in a single hierarchical loop: tasks are decomposed top-down into accountable units and execution outcomes are aggregated bottom-up to drive systematic review and refinement. This loop provides formal guarantees on termination and deadlock freedom while mirroring the feedback mechanisms of human enterprises. Together, these contributions transform multi-agent systems from static, pre-configured pipelines into self-organising and self-improving AI organisations capable of adapting to open-ended tasks across diverse domains. Empirical evaluation on PRDBench shows that OMC achieves an $84.67\%$ success rate, surpassing the state of the art by $15.48$ percentage points, with cross-domain case studies further demonstrating its generality.

The research paper "From Skills to Talent: Organising Heterogeneous Agents as a Real-World Company" introduces OneManCompany (OMC), a framework designed to move multi-agent AI systems beyond static, pre-configured pipelines. While current AI agents are highly capable at individual tasks, they often struggle to collaborate effectively on complex, open-ended projects. OMC addresses this by treating an AI workforce like a real-world company, providing a principled organizational layer that manages how agents are recruited, coordinated, and improved over time, independent of their specific internal knowledge.

The Talent-Container Architecture

OMC separates an agent’s "identity" from its "execution environment." An agent is defined as an Employee, which consists of two parts: * Talent: A portable package containing the agent’s role, prompts, skills, tools, and working principles. * Container: The runtime environment (such as LangGraph or Claude Code) that hosts the Talent.
By using six standardized "organizational interfaces," OMC allows different types of agents to work together seamlessly. This means the system can recruit specialized agents from a community-driven Talent Market on demand, allowing the organization to dynamically fill capability gaps as a project evolves rather than relying on a fixed team.

The E2R Tree Search

To manage complex projects, OMC uses an Explore-Execute-Review (E2R) tree search. This mechanism mirrors how a human company functions: * Explore: The system decomposes a high-level goal into smaller, manageable tasks. * Execute: These tasks are assigned to specific employees who carry out the work. * Review: The outcomes are evaluated, and the results are used to refine the project plan or improve future performance.
This hierarchical loop ensures that tasks are completed with formal guarantees regarding termination and deadlock prevention, providing a structured way to handle the uncertainty of real-world project execution.

Self-Improving Organizations

Beyond just completing tasks, OMC is designed to evolve. The system supports "organizational learning" through post-task reflections and performance reviews. Agents can refine their working principles, and the organization can update its Standard Operating Procedures (SOPs) based on lessons learned from previous projects. This creates a feedback loop where both individual agents and the entire organizational structure improve over time, mirroring the way human enterprises grow through experience and performance management.

Performance and Results

The researchers evaluated OMC using PRDBench, a benchmark for project-level software development. In a zero-shot, single-attempt setting, the OMC framework achieved an 84.67% success rate. This performance surpassed existing state-of-the-art methods by 15.48 percentage points, demonstrating that an organizational approach to multi-agent systems is significantly more effective at handling complex, multi-step tasks than traditional, static coordination methods.

Comments (0)

No comments yet

Be the first to share your thoughts!