Several key parties play crucial roles in shaping the operations of AI agents. These parties include the model developer, the system deployer, and the user. Each of these entities contributes uniquely to the functionality and effectiveness of AI systems.

The Model Developer

The model developer is the foundational party responsible for creating the AI model that powers the agentic system. This entity sets the capabilities and behaviors that dictate how the larger system operates. By designing the underlying algorithms and training the model on relevant data, the model developer establishes the framework within which the AI agent functions. Their expertise is critical in ensuring that the model can perform its intended tasks effectively.

The System Deployer

Next, we have the system deployer, which is the entity that builds and operates the larger system that utilizes the AI model. This includes making calls to the developed model, often through a “system prompt,” and routing those calls to various tools that enable the agent to take actions. The system deployer also provides the user interface through which users interact with the AI agent. Importantly, the system deployer may tailor the AI system to specific use cases, often possessing more domain specific knowledge than either the model developer or the user. This expertise allows them to optimize the AI agent's performance in particular contexts, ensuring that it meets the needs of its intended audience.

The User

The user is the party that employs the specific instance of the agentic AI system. Users initiate the AI agent and provide it with instance specific goals to pursue. They have the most direct oversight of the agent's behaviors during its operation and can interact with third parties, such as other humans or API providers, to enhance the agent's functionality.

Overlapping Roles

It is worth noting that sometimes the same entity fulfills multiple roles. For instance, a single company may both develop a model and deploy it via an API, acting as both the model developer and one of the system deployers. Conversely, multiple entities may share a role; for example, one company might train a model while another fine-tunes it for a specific application, thus sharing the responsibilities of a model developer.

Other Relevant Actors

In addition to these primary parties, other actors play significant roles in the ecosystem of AI agents. The compute provider, for instance, operates the infrastructure such as chips and servers on which agentic AI systems run. Additionally, third parties may interact with the user initiated AI system, further influencing its operations and capabilities. Understanding the interplay between these various parties is essential for grasping how AI agents function and evolve. As the field continues to advance, the collaboration and responsibilities among these entities will likely become even more complex, shaping the future of AI technology.

Bringing It All Together

Our platform brings these key players together in beneficial way, collaboration and effectiveness of AI agents. By integrating the model developers, system deployers, and users, we create a environment that potential of AI technology for all stakeholders involved.

Y. Shavit et al., “Practices for Governing Agentic AI Systems”.