Understanding Multi-AI Agent systems: The future of process automation
A multi-AI agent system involves collaboration between more than one AI agent. A single AI agent does one task properly at a time. However, when you want multiple tasks to happen sequentially, then we may use multiple AI agents.
The question is how do these agents interact with each other. The output of the first AI agent is fed as an input to the second AI agent. The output of the second is fed as an input to the third AI agent and so on. This is when agents work sequentially. However, they can also work parallelly, meaning AI agents give independent responses and results get collected at the end.
Regardless of whether it is a single or multi-AI agent, every AI agent shares the following characteristics:
-
Role: Each agent has a unique role provided to it. Consider an example of a customer query resolution AI agent. The role provided to the agent can be "Senior sales support representative".
-
Goal: Each agent has an objective to fulfill. In this example, it will be providing solution to the customer.
-
Back story: Each agent is provided a backstory as to how it will behave, what is the task that it has to perform. It provides a context to the Agent to work with.
-
Tools and utilities: The agent does not work by itself. It relies on a knowledge base that can be internal to the company or external resources including web search, databases, APIs, etc.
-
Memory: Agents have memory that it can refer to and give an optimum and context-specific response. It remembers the user history and preferences through threads and uses it to optimise the current response.
CrewAI, for instance, provides you a framework to integrate multiple agents and fine-tune the above mentioned characteristics.